https://hemerotecadigital.uanl.mx/files/original/337/14329/PERSPECTIVAS_SOCIALES._2007._Vol._9._No._1._0002016264.ocr.pdf c3c34531af1385bb7087edc994f88a71 PDF Text Text ��Perspectivas Sociales - Social Perspectives Vol. 9, no. 1, primavera/spring 2007 Publicación semestral de/ Biannual publication ofthe: Universidad Autónoma de Nuevo León, MéJ<jco (Ing. José Antonio González Treviño, Rector; MTS. Graciela Jaime Rodríguez, Directora de la Facultad de Trabajo Social; Dr.Jorge Noel Valero Gil, Director de la Facultad de Economía); University of Texas at Austin, E.E.U.U. (Dr. Lany R. Faulkner, President; Dr. Barbara W. White, Dean School ofSocial Work); University ofTexas atArlington, E.E.U.U. {Prof. James D. Spaniolo, President; Dr. Santos H. Hemández, Dean School of Social Work); University ofTennessee (Dr. John Petersen, Presiden!; Dr. Karen Sowers, Dean College of Social Work); Universidad Juárez del Estado de Durango (C.P. Rubén Calderón Luján, Rector; Lic. Ana Maria Alvarez del Castillo González, Directora de la Facultad de Trabajo Social), Universidad de Colima (M.C. Miguel Angel Aguayo López, Rector; M.C. Sergio Wong de la Mora, Director de la Facultad de Trabajo Social). Editores /Editors México - UANL: Veronika Sieglin (coord.) y Maria Elena Ramos Tovar México - U. Juárez de Durango: Maria Guadalupe Salas Medina Estados Unidos/USA -Austin: Lori Holleran y Dennis Poole Estados Unidos/USA - San Antonio - Cora Le-Doux Comité Editorial / Editorial Board Claudia Campillo Toledano (UANL, México), Guillermina Garza Treviño (UANL, México), Dennis T. Haynes (UT Austin, E.E.U.U.), Lori Holleran (UT Austin, E.E.U.U.), Cora Le-Doux (Our Lady ofthe Lake University), Raúl Eduardo López Estrada (UANL, México), Maria Elena Ramos Tovar (UANL, México), Manuel Ribeiro Ferreira (UANL, México), Veronika Sieglin (UANL, México), José Guillermo Zúñiga (UANL, México) Comité Científico / Scientific Com.mittee Socorro Arzaluz (El Colegio de la Frontera Norte, México), April Brayfield (Tulane University), Krista Brumley (ITESM), Nilsa Burgos (Universidad de Puerto Rico, Puerto Rico), Miguel Ferguson (UT Austin), Victor García Toro (Universidad de Puerto Rico, Puerto Rico), Nirmal Goswarni, (Texas A&M UniversityKingsville), Dagmar Guardiola (Universidad de Puerto Rico, Puerto Rico), Emilio Hemández Gómez (Universidad Autónoma de Baja California, México), Maria de la Luz Javiedes Romero (UNAM, México), Christina Krause (Universidad de Gottingen, Alemania), Gisela Landázurri Benitez (UAM, México), Maria Cristina Maldonado (Universidad del Valle, Cali, Colombia), Freddy Marinez Navarro (ITESM, México), Amparo Micolta Leifü (Universidad del Valle, Cali, Colombia), Benito Narváez Tijerina (UANL, México), Gabriela de L. Pedroza Villarreal (ITESM, México), C~cilia Quaas Femández (Universidad de Valparaíso, Chile), María Imelda Ramírez (Universidad Nacional de Colombia), Alba Nubia Rodríguez Pizaro (Universidad del Valle, Cali, Colombia) Flavio Sacco dos Anjos, (Universidad Federal de Pelotas, Brasil), Verónica Vázquez García (Colegio de Posgraduados, México), Maria Zebadúa (UANL, México) Editores del actual número/Editors of this number Jorge Noel Valero Gil y Lourdes Treviño Comité de redacción Elisa HemándezAréchiga (UANL, México) Coordinadora de difusión y distribución Maria Elena Ramos Tovar (UANL, México): distribución comercial y académica Claudia Campillo Toledano (UANL, México): distribución académica Publicación semestral/semestral publication: correo electrónico/email: veronikasieglin@yahoo.de; vsieglin@hotmail.com; lorikay@mail.utexas.edu; mramor@facts.uanl.mx www_. fts.uanl.rnx\revista.html ISSN: 1405-1133 Impreso en /Printed in Monterrey, Nuevo León, México Primavera de 2007 Tiraje /issue: 1500 Los artículos publicados son responsabilidad exclusiva de los autores / Toe articles published in this joumal are solely the responsability of the authors FONDO UNIVERSITARIO �Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vo/.9, Num. I I FONDO INDICE DE CONTENIDO - TABLE OF CONTENTS UNIVERSITARIO Presentación - Presentation 5 ARTICULOS DE INVESTIGACIÓN RESEARCB ARTICLES Has Mexican growth been pro-poor? Abdelkrim Araar; Jean-Yves Duelos, Mathieu Audet, Paul Malaiissi 17 Pro-Poor Food Taxation and Subsidy Reforms in Mexico Mathieu Audet, Paul Makdissi, Abdelkrim Araar; Jean-Yves Duelos 49 Gender-bias in Education Opportunities for Population Aged 12-18 in Mexico: 1992-2004 Ernesto Aguayo, Joana Chapa, Erick Rangel, Lourdes Treviño, Jorge Va/ero 65 Toe Informal Sector in Mexico: Characteristics and Dynamics Eduardo Rodríguez-Oreggia 89 Access and Use ofHealth Care Services by Mothers and Children in the Texas-Mexico Border Region: Preliminary Findings from the 2006 Rio Grande Valley Health Survey Patricia B. Reagan, José A. Pagán 157 Diabetes and Employment Productivity: Does Diabetes Management Matter? H. Shelton Brown IIL José A. Pagán, Craig Hanis, Adriana Perez 175 �Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vo/.9, Num. I I Pág. 5-14 NORMAS DE PRESENTACIÓN DE ARTÍCULOS GUIDELINES FOR CONTRIBUTORS FORMATO DE SUSCRIPCIÓN SUBSCRIPTION FORM 5 Presentación Este número especial contiene seis de los artículos presentados en el "VID Encuentro Internacional de Capital Humano, Crecimiento, Pobreza: Problemática Mexicana" que fue llevado a cabo en Octubre de 2006 en Monterrey, México. Cuatro de los artículos versan sobre la problemática de México y los restantes dos sobre la región localizada en la frontera Estados Unidos-México en Texas que se caracteriza por altos niveles de pobreza y una alta proporción de población hispánica. Los artículos tratan de pobreza, impuestos y pobreza, salud y pobreza, los sectores formal e informal y la discriminación en educación. Los artículos "Has Mexican growth been Pro-poor?" y "Pro-Poor Food Taxation and Subsidy Reforms in Mexico" aplican nuevas técnicas para estudiar la pobreza en México. El primer artículo trata los cambios en pobreza entre 1992, 1998 y 2004 y utiliza datos de la ENIGH (Encuenta Nacional de Ingreso Gasto de los Hogares) y el segundo trata los efectos de impuestos potenciales en alimentos como cereales, leche, carne, pescado, huevos, aceites, tubérculos, verduras, frutas, azúcar y café utilizando la ENIGH 2004. La técnica del artículo sobre crecimiento se deriva de Duelos y Wodon (2004) y la del artículo sobre impuestos en alimentos se deriva de Makdissi y Richard (2007). Ambos artículos definen funciones de evaluación para hacer el análisis de cambio distribucional. Ambos definen una norma relativa y absoluta para trabajar en pobreza. La norma relativa del artículo sobre crecimiento evalúa si el ingreso representativo de los pobres crece más aprisa que el crecimiento promedio del ingreso y la del artículo sobre impuestos en alimentos se enfoca en el consumo relativo de los pobres respecto al ingreso relativo. La norma absoluta del artículo sobre crecimiento representa el cambio absoluto en el ingreso de los pobres, mientras que la del artículo sobre impuestos en alimentos considera el consumo de los pobres en cada alimento contra el consumo promedio. Ambos artículos confrontan el problema de que México tiene dos líneas de pobreza: una para las áreas ISSN 1405-1133 O 2007 Universidad Autónoma de Nuevo León, University ofTexas of Austin, University ofTexas ofArlington, University ofTennessee, Universidad Juárez del Estado de Durango, Universidad de Colima. �6 / Presentación / Presenta/ion rurales (menos de 15000 habitantes) y otra para las áreas urbanas, determinando los ingresos en referencia a los precios rurales. El artículo sobre crecimiento desarrolla las herramientas estadísticas para probar la signi:ficancia de los resultados. Los autores encuentran que el crecimiento de México fue totalmente anti-pobre (de acuerdo a la norma absoluta) durante el periodo 1992-1998 y pro-pobre durante los periodos 1998-2004 y 1992-2004. Estos resultados aplican tanto para la razón de pobreza (headcount ratio) como para los índices diferenciales de pobreza promedio. Las pruebas de la pro-pobreza relativa entre 1998 y 2004 y entre 1992 y 2004 son concluyentes en el sentido de que el crecimiento en estos periodos fue pro-pobre tanto para la razón de pobreza como para los índices diferenciales. El artículo sobre impuestos a los alimentos determina el impacto de cambios marginales en el impuesto a un bien al nivel de pobreza de un individuo dado un nivel de ingreso. Los autores definen las curvas Dominancia de Consumo pro-pobreza, relativas y absolutas, siguiendo a Makdissi y Richard (2007) y encuentran resultados diferentes para estas medidas. En el caso de las medidas relativas encuentran que la reducción marginal en impuestos en cualquiera de los alimentos estudiados sería pro-pobre. Cuando se utiliza la norma absoluta, este resultado se obtiene solamente para los impuestos en aceites, azúcar y huevos. Incrementos marginales en cualquier impuesto al consumo de cualquier otro de los alimentos sería totalmente pro-pobre. 1 1 La distribución de recursos dentro del hogar se ha vuelto uno de los aspectos más importantes en la investigación del capital humano. Existe considerable evidencia de que los recursos no son distribuidos aleatoriamente dentro de los hogares y que además son distribuidos inequitativamente en muchos países en desarrollo. El artículo "Genderbias in Education Opportunities for Population Aged 12-18 in Mexico: 1992-2004" estudia la existencia de discriminación en educación en México en contra de mujeres jóvenes. Para el caso de México, se piensa generalmente que las niñas - más específicamente las niñas pobres de áreas rurales- son discriminadas en educación dentro de sus familias. También se cree que 15 de cada 100 Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Yol.9, Num. 1 I 7 padres no invierte en la educación de sus hijas porque piensan que las niñas se casarán y, por lo tanto, invertir en su educación seria un desperdicio de dinero. Más aún, los esfuerzos gubernamentales para abatir la pobreza se han enfocado recientemente en disminuir la "supuesta" discriminación contra las hijas. El programa de asistencia social Oportunidades (antes Progresa) entrega apoyos monetarios a las familias pobres con la condición de que los hijos asistan a la escuela y a los centros de salud. Con la intención de disminuir la supuesta discriminación contra las mujeres, las transferencias son mayores para las niñas que para los niños. Los autores no encuentran suficiente evidencia para apoyar tales creencias. Utilizando un modelo de Mínimos Cuadrados Ordinarios robustos y un efecto de Máxima Verosimilitud con efectos aleatorios para los años 1992, 1998 y 2004, no encuentran suficiente evidencia estadística para apoyar la idea de que las familias pobres, ni en zonas rurales ni en zonas urbanas, dan mayor educación a sus hijos que a sus hijas de 12 a 18 años, o viceversa. De hecho, contrario a la creencia común, los autores encuentran que las familias no pobres, según La definición del Comité Técnico de Medición de La Pobreza (2002), invierten más en la educación de sus hijas que de sus hijos. EL artículo "The comparative dynamics of the informal sector in Mexico" contiene tres partes. En las primeras dos se comparan tres periodos de tiempo: 1990-1991, 1995-1996 y 2003-2004 y en la tercera se comparan los años 1991, 1996 y 2004. Los datos utilizados son de la Encuesta Nacional de Empleo Urbano (ENEU) y la sección urbana de la Encuesta Nacional de Empleo Trimestral (ENET) ya que los hogares encuestados en ambas bases de datos son seguidos durante cinco trimestres. En la primera parte, el autor utiliza matrices de transición para estudiar los cambios entre las siguientes ocho categorías laborales: informal dependiente, formal dependiente, empleador, autoempleado, sector público, no remunerado e inactivo. En esta parte se encuentra que las principales opciones para las mujeres, tanto en el sector formal como el informal, son permanecer en el mismo status (categoría) o cambiar a la inactividad. Para los hombres dentro del sector formal o informal, las opciones más importantes son permanecer en el mismo sector o cambiar de un sector al otro. En la segunda parte, el autor utiliza el modelo �8 / Presentación / Presenta/ion Revista Perspectivas Socia/es / Social Perspectives primavera/spring 2007. Vo/.9, Num. I / logístico donde las variables dependientes son las diferentes categorías laborales. Se encuentra que una mayor edad disminuye la probabilidad de estar en el sector formal y que la gente más educada tiene mayor probabilidad de estar en el sector formal o en el sector público. Otro hallazgo importante es que la probabilidad de tener seguridad social aumenta si otros miembros del hogar también tienen este beneficio. Finalmente, en la tercera parte, se utilizan regresiones cuantílicas para determinar los salarios y las categorías laborales. El autor encuentra que las categorías empleador y sector público presentan los rendimientos más altos. El artículo "Access and Use of Health Care Services by Mothers and Children in the Texas-Mexico Border Region: Preliminary Findings from the 2006 Rio Grande Valley Health Survey" de Reagan y Pagán reporta resultados sobre la relación entre los patrones de uso de servicios de salud de las madres que viven en la frontera Estados Unidos-México y sus hijos, utilizando la Encuesta de Salud del Valle de Río Grande 2006 la cual reúne datos sobre madres y un hijo de cada hogar seleccionado aleatoriamente siempre que sean residentes de los condados de Cameron Hidalgo, Start y Willacy. Las comunidades a lo largo de esta fronter; se encuentran entre las de las tasas de pobreza y falta de aseguramiento médico más altas en Estados Unidos. Las estimaciones son realizadas en dos pasos. En el primero, se estiman los determinantes de acceso/uso a los servicios de salud tanto para madres como para sus hijos utilizando años de edad, género, status de salud (reportado por el encuestado), ingreso familiar anual y status de aseguramiento médico en un modelo de ecuaciones probit bivariadas. En el segundo paso se encuentra que los errores de ambas estimaciones, para madres e hijos, están relacionados. Este resultado es importante ya que muestra que las intervenciones que promueven el buen uso del servicio de salud para madres latinas tienen efectos positivos sobre sus hijos. Los autores argumentan que los programas públicos de seguridad social que se enfocan en cubrir a los hijos que no están asegurados, pero que dejan a los padres sin seguro pueden no tener la ventaja de presentar estos efectos que pasan de madres a hijos. El artículo "Diabetes and Employment Productivity: Does Diabetes Management Matter?" de Brown, Pagán, Hanis y Pérez utiliza información de una encuesta llevada a cabo en Brownsville Texas en ' ' 9 un área metropolitana que está localizada en la región fronteriza entre Estados Unidos y México que se caracteriza por altos niveles de pobreza, bajo desempeño en educación y por un 91.3% de población de origen hispánico. Los autores utilizan microdatos del Proyecto de Impacto en Diabetes y seleccionaron un participante de cada hogar. El artículo tiene dos secciones principales. En la primera sección, los autores examinan si un pobre ~ontrol de la diabetes, medido por la interacción de tener diabetes y niveles de hemoglobina glicosilada Hbalc, es la causa de resultados adversos en el mercado de trabajo más que estos resultados fueran causados solamente por presentar la enfermedad. En la segunda sección, los autores examinan si la diabetes afecta o no en la productividad laboral entre niveles de salarios, con o sin control, mediante una regresión cuantílica. En la primera parte los autores encuentran que, en el caso de hombres con diabetes, la enfermedad per se está negativamente relacionada a los salarios. En la segunda parte encuentran que los efectos nocivos de la diabetes son mayores en los deciles más altos de ingreso y que aquellas personas con diabetes controlada no son más productivas que quienes la presentan y no la han controlado. �lO / Presentación I Presenta/ion Presentation This special issue contains six of the papers presented at the "VIII International Meeting of Human Capital, Growth, Poverty: Mexican Problematic" that took place in Monterrey, Mexico in October 2006. Four of the papers are about Mexico and the remaining two about the region located in the US-Mexico border in Texas, which is characterized by high levels ofpoverty anda high proportion ofHispanic population. Toe papers <leal with poverty, taxes and poverty, health and poverty, formal and informal labor sectors and discrimination in education. Toe papers "Has Mexican growth been pro-poor?" and "Pro-Poor Food Taxation and Subsidy Refonns in Mexico" apply new techniques to study poverty in Mexico. Toe first paper is related to the changes in poverty among 1992, 1998 and 2004 and uses data from ENIGH (National Survey ofHousehold Income and Expenditures). The second one is related to the e:ffects of potential taxes on foods such as cereals, milk, meat, :fish, eggs, oils, tubercles, vegetables, fruits, sugar and coffee using ENIGH 2004. Toe technique of the paper on growth is derived from Duelos and Wodon (2004) and the one of the paper on taxes on foods is derived from Makdissi and Richard (2007). Both papers define evaluation functions to make the analysis of distributional change. Moreover, both define a relative and an absolute norm to work on poverty. Toe relative nonn of the paper on growth evaluates if the representative income of the poor grows faster than the mean growth of income, while the paper on taxes on foods looks at the relative consumption of the poor against the relative income. Toe absolute norm of the paper on growth represents the absolute change in the poor's income, whereas the paper on taxes on foods considers the consumption of the poor on each good against the average consumption. Both papers confront the problem that Mexico has two poverty lines: one for the rural areas (less than 15000 habitants) and one for urban areas by assessing ali income measures in reference to rural prices. Toe paper on growth develops the statistical tools to test for the significance of the results. Toe authors find that Mexico's growth was Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Vo/.9, Num. l / 11 absolutely (with absolute nonn) anti-poor during the period 1992-1998 and pro-poor during the periods 1998-2004 and 1992-2004. These results apply to the headcount ratio as well as to the average poverty gap indices. Toe tests for the relative pro-poomess between 1998 and 2004 and between 1992 and 2004 are conclusive in the sense that the growth in these periods was pro-poor for both measures of poverty. Toe paper on pro-poor food taxation assesses the impact ofmarginal changes to the tax on a good at the poverty leve! of an individual, given a certain level of income. Toe authors define the relative and absolute pro-poor Consumption Dominance curves following Makdissi and Richard (2007) and they find different results for these measurements. In the case of the relative measurements, they observe that marginal reduction in taxes on any category of the foods studied would be pro-poor. When the absolute norm is used, this result is only obtained for the taxes on oils, sugar and eggs. Marginal increases in any consumption tax on any other category of foods would be absolutely pro-poor. Toe intra-household allocation ofresources has become one ofthe most important issues in human capital research. There is considerable evidence that resources are not only allocated randomly within households, but also unequally distributed within the family in many developing countries. The paper "Gender-bias in Education Opportunities for Population Aged 12-18 in Mexico: 1992-2004" studies the existence of discrimination in Mexico against young women related to years of education. For the case ofMexico, it is generally believed that girls-more speci:fically poor rural girls- are educationally discriminated within their families. It is also claimed that 15 of every 100 parents do not invest the education of their daughters because they think girls will in get married and, therefore, investing in their education woulol be a waste of money. Furthermore, government efforts to abate poverty have been recently focused on decreasing the "assumed" discrimination against female children. Toe government assistance program Oportunidades (formerly Progresa) gives monetary transfers to poor families conditioned on ~aving their children attending school and health clinics. Intended to reduce such "assumed" discrimination against girls, transfers are larger for girls �12 / Presentación / Presenta/ion than for boys. Toe authors <lid not find enough evidence to support such believes. Using an OLS-Robust model anda ML-Random Effects model for the years 1992, 1998 and 2004, they did not find enough statistical evidence to support the idea that poor families, either in rural as in urban areas, provide more education to their 12 to 18 years old sons or daughters. In fact, contrary to the general belief, they found that non-poor families, as established by the Mexican Technical Committee for Measuring Poverty (2002), invest more in the education oftheir daughters. The paper "Toe comparative dynamics of the informal sector in Mexico" is divided into three parts. The first two compare three time periods: 1990-1991, 1995-1996 and 2003-2004 and the third part compares the years 1991, 1996 and 2004. Toe data come from the ENEU (National Survey ofUrban Employment) and the urban part ofthe ENET (Quarterly National Survey of Employment), since in those surveys the households are followed through five quarters. In the first part the author uses transition matrices to study the changes among the following eight work status: dependent informal, dependent ' formal, employer, self-employed, public sector, not remunerated, unemployed and inactive. It is found that the main options for women working in the formal and informal sectors are to remain in the same status or to move to inactivity. Formen employed in the formal or informal sector, the most important options are to remain in the same sector orto switch from one sector to the other. In the second part, the author utilizes the multinomial logit model having as dependent variables the different work status. He finds that a higher age lowers the probability of being employed in the formal sector and that more educated people have higher probability of being employed in the formal e r public sectors. Another important finding is that the probability of havir;g social security increases if other members of the household also have this benefit. Finally, in the third, quantile regressions are used to study the determinants of wages and the work status. Toe author finds that thi> c;tatus of employer and public sector show the highest returns. The paper "Access and Use ofHealth Care Services by Mothers and Children in the Texas-Mexico Border Region: Preliminary Findings from the 2006 Rio Grande Valley Health Survey" ofReagan and Pagan reports Revista Perspectivas Sociales I Socio/ Perspectives primavera/spring 2007. Vol. 9, Num. 1 I 13 results on the pattems ofhealth care utilization between border-dwelling mothers and their children, using the 2006 Rio Grande Valley Health Survey, which collects data on mothers and one randomly selected child from each household, residing in Cameron, Hidalgo, Start and Willacy counties. Communities along this U.S. - Mexico border region have the highest rates of poverty and uninsurance in the U.S. Toe estimations are made in two steps. In the first one the determinants of health care access/utilization are estimated for both mothers and children utilizing years ofage, gender, self-reported health status, yearly household income and health insurance status, in a model of bivariate probit equations. In the second step, the errors of both regressions, mothers and children, are found to be related. This result is important because it shows that interventions that promote good health care utilization behavior for Latín mothers spillover to their children. The authors argue that public health insurance programs that focus on covering uninsured children but leave their parents uninsured may end up not taking full advantage of health care access/utilization from mothers to their children. The paper "Diabetes and Employment Productivity: Does Diabetes Management Matter?" of Brown, Pagan, Hanis and Perez uses information from a survey in Brownsville, Texas, a metropolitan area that is located in the US-Mexico border region characterized by high poverty levels; low educational attainment and by a 91.3% of Hispanic origin population. Toe authors used microdata from the Diabetes hnpact Project and selected one participant from each household. The paper is organized in two sections. In the first one, the authors examine whether poor diabetes management, measured by the interaction ofhaving diabetes and glycosylated hemoglobin levels (Hbalc), is the cause of adverse labour market outcomes rather than diabetes per se. In the second one, they examine whether or not diabetes a:ffects labour productivity, whether managed or not, across wage levels using quantile regression. In the first part the authors find that in the case of men diabetes per se is negatively related to wage. In the second part they find that the detrimental e:ffects of diabetes are higher at the higher wage quantiles and that those persons with diabetes who do manage it are not more productive than those whose diabetes has not been managed. �14 / Presentación / Presentation Duelos, J.-Y. and Q. Wodon (2004): "What is "Pro-Poor"?" CIRPEÉ Work.ing Paper #0425. Makdissi, P. and P. Richard (2007), "Pro-Poor Indirect Tax Reforms", mimeo. ARTÍCULOS DE INVESTIGACIÓN RESEARCH ARTICLES �Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vol.9, Num. 1I Pág. 17-47 17 Has Mexican growth been pro-poor? Abdelkrim Araar*, Jean-Yves Duclost, Mathieu Audet¡, Paul Makdissi§ ** Abstract This paper propases techniques to test for whether growth has been propoor. We first review different definitions of pro-poomess and argue for the use of methods that can generate results that are robust over classes of pro-poor measures and ranges of poverty lines. We then provide statistical pro-cedures that rely on the use of sample data to infer whether growth has been pro-poor in a population. We apply these procedures to Mexican household surveys for the years of 1992, 1998 and 2004. We find strong statistical evidence that Mexican growth has been absolutely anti-poor between 1992 and 1998, absolutely pro-poor between 1998 and 2004 and between 1992 and 2004, and relatively pro-poor between 1992 and 2004 and between 1998 and 2004. Toe relative assessment of the period between 1992 and 1998 is statistically too weak to lead to a robust evaluation of that period. Keywords Pro-poor growth, Poverty, lnequality. * Département d' économique and CIRPÉE, Pavil/on De S 'eve, Université lava[, Sainte-Foy, Québec, Ganada, GJK 7P4; emai/: aabd@ecn.ulaval.ca; fax: /-418-6567798; f Département d' économique and CIRPÉE, Pavillon de S 'eve, Université lava/, Qu ébec, Ganada, GIK 7P4; email: jyves@ecn.ulaval.ca; fax: /-418-656-7798; phone: 1-418-656-7096 f GRÉDL Université de Sherbrooke, 2500, boulevard de l'Université, Sherbrooke, Qu ébec, Ganada, J/K 2RI; email: maudet@worldbank.org § Département d'économique and CIRPÉE, Université de Sherbrooke, 2500 boulevard de l'Université, Sherbrooke, Québec, Ganada, JIK 2RI; emai/: paul.makdissi@USherbrooke.ca ** We are grateful to Dr. Lourdes Treviño and Professor Jorge Va/ero Gilfor their invita/ion to present this paper al the Eight Symposium on "Capital Humano, Crecimiento, Pobreza: Problemática Mexicana" that took place in Monterrey, Mexico on October 12 and 13, 2006, and to the participants at that conference for valuable comments. ISSN 1405-1133 O 2007 Universidad Autónoma de Nuevo León, University ofTexas ofAustin, Univer.aity ofTexas of Arlingtoo, Uoiv=ity ofTeonessee, Universidad Juárez del Estado de Durango, Universidad de Colima �18 / Has Mexican growth been pro-poor? Resumen Este trabajo propone técnicas para comprobar si el crecimiento ha sido a favor de los pobres. Primero repasamos las diferentes definiciones a favor de la pobreza o "pro pobres" y argumentamos por el uso de métodos que puedan generar resultados que sean robustos eligiendo entre clases de mediciones pro pobreza y rangos de lineas de pobreza. Luego generamos procedimientos estadísticos que se basen en datos muestrales y que permitan inferir si el crecimiento ha sido pro pobreza en una población. Aplicamos esos procedimientos a muestras de hogares mexicanos para los años 1992, 1998 y 2004. Encontramos fuerte evidencia estadística de que el crecimiento mexicano fue absolutamente anti pro pobreza entre 1992 y 1998, absolutamente pro pobreza entre 1998 y 2004 y entre 1992 y 2004, y relativamente pro pobreza entre 1992 y 2004 y entre 1998 y 2004. La evaluación de la medida relativa en el período entre 1992 y 1998 es muy débil estadísticamente para que conduzca a una evaluación robusta del período. Palabras clave Crecimiento pro pobres, pobreza, desigualdad. Introduction It would seem relatively uncontroversial to conceive of the pro-poomess of growth as referring generally speaking to the impact ofgrowth on the wellbeing of the poor and therefore to its impact on poverty. Like many distributive concepts, however, its precise meaning and its usefulness are essentially a matter of judgement1 • There are at least three elements of contention in trying to make the assessment ofpro-poomess operational. Toe first fundamental issue in the definition of pro-poomess is whether it should be absolute or relative. A second issue is what poverty line should be chosen to separate the poor from the non-poor. A final issue is how we should assess in the aggregate the heterogeneous impact of growth among a population of heterogeneous poor, an issue which also addresses what relative normative weights are to be attributed to the different poor individuals. 1 See, among many recen! contributions to that debate, Bourguignon (2003), Bruno, Raval/ion, and Squire (1999), Do/lar and Kraay (2002), Eastwood and Lipton (2001), United-Nations (2000), and World-Bank (2000). Revista Perspectivos Sociales / Social Perspeclives primavera/spring 2007. Vo/.9, Num. l I 19 This paper attempts to address all these three issues and to make it conceptually operational and empirically feasible to test for the pro-poorness of distributive changes. To do this, we first rely on the definitional framework of Duelos and Wodon (2004). Roughly speaking, and according to that framework, a relative definition of pro-poomess judges a distributive change to be pro-poor if the proportional change in the incomes of the poor is no less than sorne norm, often set as the growth rate in mean income or in sorne quantile such as median income. For an absolute definition, the incomes of the poor need to grow by an absolute amount that is no less than sorne norm, this time often set as zero or as sorne proportion of the absolute change in mean or median incomes. These different definitions can also be linked to the usual concepts of absolute and relative poverty. With relative poverty, the poverty line is usually defined as a proportion of sorne central tendency of an income distribution; with absolute poverty, the real level of the poverty line normally remains the same even if the income distribution changes. Toe framework of Duelos and Wodon (2004) also enables to get around the difficult of having to choose l) a poverty line to separate the poor from the nonpoor, and 2) a set ofnormative weights to differentiate among the poor. Toe framework <loes this by investigating how pro-poor judgements can be made robust to wide classes of pro-poor evaluation functions and to ranges of poverty lines.2 This paper then makes it empirically feasible to test for pro-poomess of growth. To do this, we derive the sampling distribution of the various estimators that are needed to test for absolute and relative pro-poomess. This enables us inter alia to draw confidence intervals around the differences that must be signed in order to conelude that a change has been robustly pro-poor - or not. We implement these statistical techniques taking full account of the sampling design of the surveys we use. We apply the procedures to Mexico's National lncome and Expenditure Surveys collected in 1992, 1998 and 2004. We find strong evidence that 2 Many di.fferent approaches have been proposed to separate the poor from the nonpoor and to compute and aggregate index ofpro-poorness. See.for instance, McCul/och and Baulch (1999), Kakwani, Khandker, and Son (2003), Kakwani and Pernia (2000), Raval/ion and Chen (2003), Klasen (2003), Essama-Nssah (2005), Rava/lion and Datt (2002) and Son (2004). �20 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales I Social Perspectives primaveralspring 2()()7. Vol. 9, Num. 11 Mexican growth has been absolutely anti-poor between 1992 and 1998, absolutely pro-poor between 1998 and 2004 and between 1992 and 2004 and relatively pro-poor between 1992 and 2004 and between 1998 and' 2004. The relative pro-poor assessment of the period between 1992 and 1998 is, however, statistically too weak to lead to a robust pro-poor evaluation of that period. Theoretical framework The setting Let y, = (yf ,Y~ , · · · , Y!J E~• be a vector of non-negative initial incomes3 (at time 1) of size n 1, and let y2 = (!A, Yi. ··· ,iln.) be an analogous vector of posterior incomes (at time 2) of size n2• ,., 21 Toe change from yl to y2 will be deemed pro-poor if W (y, Y2, g, z) :SO. Clearly, whether the distributional change will be deemed pro-poor will depend on the way in which z , IT, and IT* will be chosen. To put sorne structure on the form of W in which we should be interested, we need to invoke a few axioms. Toe first one is a focus axiom: W is not sensitive to the values of y 1 that exceed z . Toe distribution of min(y; , z ) is thus sufficient for judgements of pro-poomess. Second, we can postulate an axiom of population invariance. This says that adding a replication of a population to that same population has no impact on W. lt is a common axiom in welfare economics that enables us to make pro-poor judgements even when the absolute population size varies across the distributions. The following draws extensively from Duelos and Wodon (2004). To determine whether the movement from y I to y2 is pro-poor, we :first need to define a standard with which this assessment can be made. First consider the case of a relative standard, which we will take in this paper as the growth in average incomes, denoted by g. Intuitively, for growth to be relatively pro-poor, we wish the poor's "representative" income to undergo a proportional change that is no less than 1+g. That growth g can be negative as well as positive. Relative pro-poomess is consistent, for instance, with the view ofKakwani and Pernia (2000) that "promoting pro-poor growth requires a strategy that is deliberately biased in favor of the poor so that the poor benefit proportionately more than the rich. (p.3)". A third axiom is that of anonymity: this says permuting the incomes of any two persons in any given distribution should not affect pro-poor judgements. Note that this axiom will typically lead to violations of the well-known Pareto effi.ciency criterion; our framework does not not require that none ofthe poor be penalized by a distributional change for that change to have a chance to be declared pro-poor. Were we not to impose this axiom, it would be practically impossible to order the initial and posterior distributions in the presence of a large number of individuals, and we would also need panel data. Toen denote by z > Oa poverty line, defined in real terms. Let W (yi, Y2, g, z) be the pro-poor evaluation function that we want to use. It is defined as the difference between two evaluation functions TI (y1 , z) and TI* (Y2 , 1 + g, z), each for time 1 and time 2, respectively, and which are analogous to poverty indices for each of the two time periods: We may then also impose an axiom of monotonicity: for a given g, if anyone's posterior income increases, W should not increase, and may sometimes fall. Increasing posterior incomes make it more likely that the distributive change will be declared pro-poor. Toe next axiom is a normalization one: if there has been no distributional change, and thus also no change in the mean, then W = O. Relative pro-poor judgements W (Y1, Y2, g, z) = (Y2, 1 + g, z )3 IT (y1, z). Or consumption, wealth, or any other welfare indicator ofinterest. (1) Finally, we can also axiomatize our view of relative pro-poomess. Formally, suppose that y/ (l + g) = y/ (1 + g). Toen, according to relative judgements ofpro-poomess, y and y should be judged equally pro-poor �22 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales / Social Perspectives prim<Neralspring 2007. Vol.9, Num. I I by W regardless of the initial distribution y 1 • Combined together, the axioms that we have invoked until now define what we can term to be a first-order class of relative pro-poor evaluation functions. Denote that class as Q 1 (g, z +). The class Q 1 (g, z +) regroups all of the functions W that satisfy the focus, the population invariance, the anonymity, the monotonicity, the normalization and the relative axioms, and for which z '.S z + . Now let Fj (y) be the distribution function of distributionj. Also define as Qj (p) the quantile function for distribution Fj . This is formally defined as Qj (p) = inf {s?: O¡Fj (s)?:p} forp E (O, l]. With acontinuous distribution and a strictly positive income density, Q(p) is simply the inverse ofthe distribution function, and it is the income of that individual who is at rank p in the distribution. The popular class of FGT indices is then given by: P,(z) 1 .P¡(z;a) = (1- Q,(p)/z)º dp. (2) 0 Pj (z; a= O) is the headcount index (and the distribution function) atz, and Pj (z ; a = 1) is the average poverty gap. Duelos and Wodon (2004) show that a movement from y 1 to y2 will be judged pro-poor by ali propoor evaluation functions W ( ·, ·, g, z ) that are members of Q 1 (g, z +) if and only if (3) A distributional change that satisfies (3) is called first-order relatively pro-poor since all pro-poor evaluation functions within Ql (g, z +) will find that it is pro-poor, and this, for aoy choice of poverty line within (O, z +] aod any W that obeys the above-defined axioms. Verifying (3) simply involves checking whether - over the range of poverty lines [O, z +] - the headcount index in the initial distribution is larger than the headcount index in the posterior distribution when that distribution is normalized by l + g. An altemative and equivaleot way of checking whether a distribu- 23 tional change can be declared fust-order relatively pro-poor is to compare the ratio ofthe quantiles to the ratio of the mean, or again equivalently, to compare the growth in quantiles to the growth in the mean. That this, we check whether, for allp E [O, Fl (z +)], (4) or whether Q2(p)- Q¡(p) > Q1(p) - g. (5) Using (5) is equivalent to Ravallion and Chen (2003)'s suggestion to use "growth incidence curves" to check whether growth is pro-poor. These curves show the growth rates of living standards at different ranks in the population. First-order pro-poor judgements can be demanding in expansion periods. They require ali quantiles of the poor to undergo arate of growth at least as large as the rate of growth in mean income. We may, however, be willing to relax this condition if the rate of growth for the poorer among the poor is sufficiently large to exceed g even though the rate of growth for the not-so-poor may be below g. An axiom that captures this is the distribution sensitivity axiom. It says that the evaluation functions TI should give more weight to the poorer than to the not-so-poor among the poor. Distribution-sensitive pro-poor judgements imply that shifting incomes from the richer to the poorer is by itself a pro-poor distributional change. This axiom is known as the Pigou-Dalton principie in the welfare literature. Adding the distribution-sensitive axiom to the earlier axioms defines a second-order class of relative pro-poor evaluation functions Q2 (g, z+). Formally, Q2 (g, z +) is made of all functions W (·, ·, g, z) that satisfy the focus, the population invariance, the anonymity, the monotonicity, the normalization, the distribution sensitivity and the relative axioms, and for which z:::; z +. It can then be shown that a movement from yl to y2 will be judged propoor by all pro-poor evaluation functions W (, , g, z ) that are members �24 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vol.9, Num. 1/ of Q2 (g, z +) if and only if A ((1+ g)z;a = 1) $ Pi (z; o: = 1) forall z E (o,z+¡. (6) As for (3), a distributional change that satisfies (6) is called second-order relatively pro-poor since all pro-poor evaluation functions will find that it is pro-poor, and this, for any choice of poverty line within [O, z +] and for any W that obeys the above-mentioned axioms for Q2 (g, z +). Verifying (6) simply involves checking whether the average poverty gap in the initial distribution is larger than that in the posterior distribution when that distribution is normalized by 1+ g and this, over the range of poverty lines [O, z +]. As for first-order pro-poor judgements, there are alternative ways of checking condition (6). The cumulative income up to rank p (the Generalized Lorenz curve at p) is given by C;(J¡) = [ Q;(q)dq. (7) Toe use of the Generalized Lorenz curve provides an intuitive sufficient condition for checking second-order relative pro-poorness. A distributiona~hange is indeed second-order relatively pro-poor if for all p E [O, F2 (z + )], 1 ,, (8) 11 o Expression (8) involves computing the growth rates in the cumulative incomes of proportions p of the poorest, and to compare those growth rates to g. For 1 + g equal to the ratio of mean income, condition (8) is equivalent to checking whether the Lorenz curve for y2 is above that of y 1 for the range ofp E [O, F2 ((1 + g)z + )]. Absolute pro-poor judgements Absolute pro-poor judgements are made by comparing the absolute change in the poor's incomes to sorne absolute pro-poor standard. Denote that standard as a. Toe axiom of absolute pro-poomess says 25 essentially that TI* should be "translation invariant" in y and a, or that the pro-poor judgement should be neutral whenever the poor gain in absolute terms the same as the standard a. Hence, this axiom demands that if y + a :: y, then W (y, y, ~ z) :: O. This allows us to _define the class of first-order absolute pro-poor evaluation functions QI (a, z +) as made of ali those functions W ( ·, ·, a, z ) which satisfy the focus, the population, the anonymity, the monotonicity, the normalization and the absoluteness axioms, and for which z S z + . We will later set a to zero for the empirical illustration of this paper. We can then show that a movement from yl to y2 will be judged first-order absolutely pro-poor (that is, P!O-poor by all evaluation functions W ( , , a, z) that are members of Ql (a, z +)) if and only if (9) An equivalent way of checking whether a distributional change can be declared first-order absolutely pro-poor is to compare the absolute change in the values of the quantiles for ali p E [O, F(z + )]: (10) An analogous result holds for absolute second-order pro-poor judgements.These judgements also obey the axiom of distribution sensitivity: they are made on the basis of the class of indices Q 2 ( a, z +), which is defined as for QL (a, z +) but with the additional requirement of distribution sensitivity. We can then show that a movement from y 1 to Y2 will be judged second-order absolutely pro-poor if and only if (z +a)A((z+ a;a = 1) $ zP1 (z; a = 1) for ali z E [o,z+]. (11) A sufficient condition for condition (11) is then to verify whether, for ali p E [O, F2 (z ++a)], the change in the average income of the bottom p proportion of the population is larger than a: (12) �26- / Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Yol.9, Num. 11 Has Mexican growth been pro-poor? Statistical inference In practice, household data surveys are needed to check if growth is pro-poor or not. This forces us to <leal with issues of sampling variability and statistical inference. Indeed, a difference observed in a sample may not be empirically strong enough to be significant from a statistical point of view4 . Each of the conditions noted above takes the form of testing whether N (z) :'.SO or !).s (p) ~ Oover sorne range ofz or p. This therefore involves testingjointly overa set of null hypotheses. For primal tests ofpro-poorness, our formulation of our null hypothesis is thus that of a union of null hypotheses H0 : t.'(z) > Oforsome zE (o,z+] (13) 27 per can be shown to be asymptotically normally distributed, we can use ~ (z)+uA•<•>((B)as altemative lower and upper bounds for one-sided confidence intervals for !).s (z). For instance, an upper-bounded confidence interval ~ (z)+uA•<.->((8~ shows all of the values of r,. for which we could not reject a null hypothesis Ho :t.•(z) ::=; r,.in favor of H1 : !).s (z) :'.S r,.Our decision rule is then to reject the set of null hypotheses (13) in favor of (14) if: (17) For dual tests, we proceed similarly, noting that the signs in ( 15) and (16) are inverted. We thus build a confidence interval t.g(p) - 112,,.(p¡((B) and reject (15) in favor of (16) if (18) 1 1 1 to be testedagainstanaltemativehypothesiswhichisanintersectionofalternativehypotheses H1 : t.'(z) $ Oforall z E (0,z+J. ' 1 ' 1, (14) There remains to define ¿•(z), Ll'(p),u4.<•> ªºd u4•(p)· Let H be the number of sample observations ofincomes from a distribution j, yJ, ...,yf. Then, we have that (19) For dual tests, we use a union of mili hypotheses 1 H0 : t."(p) < Ofor sorne p E (O, 1) (15) (20) to be tested against an intersection of altemative hypotheses where the empirical distribution function is given by H1 : t.' (p) ~ Ofor all p E (O,1]. (16) F;(Y) = ?3(z; O), 1 1 1 Our decision rule will be to reject the union set of null hypotheses in favor ofthe intersection set of altemative hypotheses only if we can reject each of the hypotheses in the null set at a 100 · 0 % significance level. This can be conveniently carried out using a 100 · (1 - 0)% one-sided confidence interval, a devise we use repeatedly in the empirical Mexican graphs below. To see how this can be done, denote by ~'As (z) the sample estimator of As (z) , by As 0 (z) its sample value, and by ª l•(z} the sampling variance of As (z ). Let((B)be the (1 - 0)-quantile of the normal distribution. Given that ali of the estimators used in this pa4 See Araar (2006). (21) and that (22) and H fs1= H - 1 ¿ !/;'. h=I (23) �28 / Has Mexican growth been pro-poor? Revista Perspeclivas Sociales / Social Perspectives primavera/spring 2007. Vo/.9, Num. I / Using a first-order approximation, we :find Í'2(gz; a) - P2(gz; a) function until percentile are given by: Í'2(gz; a) - P2(gz; a) + gzf'fz(gz;a) ({l,i /J,2- µ¡ µ¡) 111 - /J,2 - (26) l\(gz;a)- P2(gz;a) = .A(gz;a)-P2 (gz;a) + : o(n- 1/2). (30) 1 1 where 1 (A 1 :µ 1 ). (32) Suppose that the two empirical distributions come from independent samples, namely, the selection of the sampling units was made independently in each sample. We then have (33) 1. n - 1 {(I(r/' < Q(p)J-v) Q{p) +,J'I [,j' < Q{p)J-C(p)} + o(n-112¡ (36) (37) (38) and E (27) (28) (29) ({l,i: /J,2) _ (A µ¡)] ÍJ = (f>i(z;a)- Pi(z;a)) -a(P2 (gz;a-1) - P2 (gz;a)) (35) C'(p)-C{p) = Therefore, for a> O, we can express .&•(z) - t.•(z) as 1 + o(n-I12¡ (25) where Ff'(gz; a) = O: (gzr1 (P2(gz; a - 1) - P2(gz; a)) > 0 for a > 0 and Ff'(gz;O) = h(gz) > o (the density at gz) for a= O. Hence, we have: 1 Q(p) _ Q(p) = H -' E (1&1' < Q(p)J -v) /(Q{p)) (24) + o(H- 1/2) + a (A(gz;a - 1) - P2(gz; a)) [ 29 lf, however, the two samples are dependent because they come, for instance, from the same panel data, then the variance must be estimated jointly over the two samples and we then have (34) For the dual or percentile approach, :first-order approximations of the sampling distribution of the quantile estimator and of its cumulative where I[!I" < Q(p)) is is an indicator function taking the value of 1 if its argument is true and Ootherwise. Supposing that the number of primary sampling units increases asymptotically to infinity, we can then estimate the sampling distribution of the above estimators in t.•(z) and l •(p) taking full account of the survey design. This is done using the procedure described in Duelos and Araar (2006), pages 284-287, a procedure which takes into account the sampling weights, the sampling design and the number of statistical units (individuals) within each of the last sampling units (each household observation in the sample). Has tbe Mexican economy been pro-poor? We apply the above methodology using Mexican data spanning the last decade and a half. Mexico is a particularly interesting economy over which to test the pro-poomess of growth. Mexico has indeed undergone very significant economic changes since 1990. After the 1994-1995 economic crisis, which culminated in an important devaluation of its currency and was probably the most severe in the country's economic history, rapid growth in exports (facilitated by NorthAmerican Free Trade and other trade agreements) as well as macroeconomic and public sector restructuring led to strong growth. Recent institutional changes have inter afia encouraged competition and growth in transportation, telecommunications, and power generation and distribution. An important issue is whether this relatively recent tidal growth has "lifted all boats". The data used for our application come from the National Income �30 / Has Mexican growth been pro-poor? and Expenditure (ENIGH) Surveys collected in 1992, 1998 and 2004. These sample data are representative at the national level. Toe objective of the ENIGH surveys has been to collect information on incomes and expenditures, goods and services used for self-consumption, and socioeconomic characteristics and labor market activities of all household members. Toe sampling process was stratified and multi-staged, with the final sampling units being households and all their members. As is common in Mexico, we use total income per capita as the 1 measure of living standards for all members of a household. To adjust for temporal variation in prices, we express incómes in reference to the 2004 consumer price index. To correct for spatial variation in prices, we assess all incomes in reference to rural prices. This is done by multiplying urban household incomes by the ratio of the rural to the urban poverty line. Toe rural poverty line in 2004 is often estimated to be around 550 pesos per month per capita. We use the product of household size and household sampling weight as an expansion factor to ensure that our samples are representative of the national distribution of the living standards of individuals. For the estimation of standard errors and thus for statistical inference, we take into account the stratification and multi-stage structure ofthe survey designas explained at the end of Section 3. 1 1 '· ' ' 1 .1 I •f 1 1 ,, We begin our investigation by considering the evolution of the density ofper capita incomes in Figure 1. Toe distribution of per capita income has worsened until 1998 since the density curves have shifted to the left. 1t has however ex.hibited a strong and quick recovery between 1998 and 2004, as shown by the shift of the density curve to the right. Toe estimates of the Lorenz curves and Gini indices presented in Figure 2 and Table 5, respectively, suggest that inequality has decreased between 1992 and 2004. Figures 3 and 4 and the results ofTable 5 suggest that absolute poverty, as measured by the headcou.nt and poverty gap indices, has increased between 1992 and 1998 and decreased between 1998 and 2004. Formal statistical testing for first-order absolute pro-poomess of Mexican growth can be done using the information presented in Figures 5 to 1O. Toe top line of Figure 5 shows the sample estimates of Revista Perspectivos Scciales / Social Perspectives prima,,era/spring 2007. Vc/.9, Num. l / .ó.1(z) = ?¡9911 (z;o: = O) - Pum (z;o: = O) 31 (39) for the difference between 1998 and 1992, whereas the dotted bottom curve is the lower bound of the one-sided confidence interval, (40) Since .!lA{z) - o-¿..<z>((8) > O is verified on Figure 5 for all reasonable poverty lines, we can infer from our data that growth was absolutely anti-poor during the period 1992 and 1998. Toe same result obtains from Figure 6 using differences in quantiles. The sample estimates ofthe difference in quantiles between 1992 and 1998 is shown by the dashed curve, and the upper bound of a one-sided confidence interval is shown by the dotted curve. Since we can see on Figure 6 that ~(p) + ªc.•(p)((8) < o for all percentiles p between O and 0.95, we can again conclude from our data that growth was absolutely anti-poor during the period 1992 and 1998. Opposite results are obtained when comparing 1998 to 2004. Judging from Figures 7 and 8, the change in distribution was first-order absolutely pro-poor. Toe upper bound of the confidence interval for .!l1(z) is everywhere negative, whatever reasonable poverty line is selected, and the lower bound of the confidence intervat for .!l1(p) is everywhere positive, whatever reasonable percentile is selected. Given the conflicting results reported above, it would seem useful to check for pro-poomess over the entire period 1992 to 2004. This can be done using Figures 9 and 10. The distributive change was almost certainly :first-order absoulutely pro-poor. Toe lower bound of the con:fidence interval for P= (z; a= O)- Pi992 (z; o:= O) is everywhere negative, again whatever reasonable poverty line is selected, and the lower bound of the con:fidence interval for the difference in quantiles,Q=(p) - Q1992(p) is everywhere positive, until at least the O.8 percentile. Thus, the anti-poor movement of 1992 to 1998 was outdone by the pro-poor movement of 1998 to 2004 so that the entire period of 1992 to 2004 can be inferred to be overall first-order absolutely pro-poor. Given the robust results obtained for :first-order pro-poomess, it is not �32 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Yo/.9, Num. / I useful to test for second-order pro-poomess since first-order pro-poomess implies second order pro-poomess. This can be seen by noting that P; (z; a = 1) = [ P; (y;a = O)dy. (41) If first-order pro-poomess obtains at order 1, then by (41) second-order pro-poomess also obtains. Toe same relation is obtained by noting from equations (7), (10) and (12) that the Generalized Lorenz curve condition is implied by the quantile condition. Testing for first-order relative pro-poorness can be done using Figures 11 to 20. Figure 11 shows why observing pro-poomess in samples <loes not mean that we can infer it in populations; to go from sample pro-poorness to population pro-poomess, we need to apply statistical inference methods. :ro see this, note that despite the fact that average income fell by about 30% between 1992 and 1998, the sample estimates of P¡998(gz;a = O)- Pim(z;a = o) suggest that the distributive movement during that period is first-order relatively pro-poor since that difference is always negative in the samples observed. But drawing a confidence interval around the sample estimates make it clear on Figure 11 that the observed differences Pi998(gz; a = O) - Pim(z; a = O) are not statistically significant over a wide range ofbottom poverty lines - the upper bounds of the one-sided confidence intervals extend above the zero line for z up to around 600 pesos and p up to around 0.28. Hence, with a conventional level 95% of statistical, the first-order relative pro-poor condition is not satisfi.ed. An analogous result is obtained on Figure 12 from comparing growth in quantiles to growth in average income. Again, for a substantial range of percentiles, the one-sided confi.dence interval overlaps with the zero line. 1,1 '1 Moving to second-order relative pro-poomess <loes not help, as shown by Figures 13 and 14. Toe statistical insignifi.cance now extends overa wider range or z (up to around 900 pesos) and p (up to around 0.4) values. This may seem counter-intuitive at first sight, but it follows from the fact that statistical uncertainty for first-order comparisons at the bottom of the distributions builds up at the second-order since second-order conditions are made of cumulatives of first-order statistics (as discussed above). There is therefore an important lesson to be drawn here. 33 If one were to omit testing for statistical signifi.cance, it might seem that second-order relative pro-poomess over the 1992-1998 period certainly cannot be weaker than first-order relative pro-poomess over the same period. But if one takes into account the effect of sampling variability at the bottom of the distribution, than the evidence for second-order relative pro-poomess is statistically weaker than that for first-order relative pro-poomess. Testing for relative pro-poomess between 1998 and 2004 is more conclusive, as shown on Figures 15 and 16. The confi.dence interval around the sample estimates of P2004(gz;a = O)-P1993(z;a = O) on Figure 15 is always below zero for z up to around 1200 pesos (as opposed to 1800 pesos for the sample estimates), which leads us to infer a robust first-order relative pro-poomess change in that period. A similar result is obtained on Figure 16 from comparing growth in quantiles to growth in average income. For a range of percentiles up to about 0.7, the lower bound of the confidence interval lies above the zero line. Given the above results, it would seem interesting to test for second-order relative pro-poomess for the 1998-2004 period. The results are shown on Figures 17 and 18. We now obtain even stronger (and very strong) evidence of the relative pro-poorness of that period. The confidence interval is always below zero for differences P,oo,(gz;a = 1) - P1998(z;a = 1) and above zero for differences C-.(p)/C,-(p) µ,oo,/µu,~- 2004 1998 2004 1998 This is not surprising given that, as discussed above, if first-order pro-poorness is verified statistically at order 1, then we can expect second-order pro-poorness also to be inferred statistically. Toe results ofthe tests for relative pro-poorness over the period 1992 to 2004 are even stronger. These are shown on Figures 19 and 20. Toe confidence interval around the sample estimates of P2ro1(gz; a = O)- P,998(z; a =O) on Figure 19 is always below zero even as we extend z beyond 3000 pesos. Toe same strong evidence is displayed on Figure 16 from the comparison of growth in quantiles to growth in average income between 1992 and 2004. For a range ofpercentiles up to about 0.9, the lower bound of the confidence interval is everywhere above the zero line. Hence, the period 1992-2004 shows a statistically and ethically very robust degree �34 / Has Mexican growth been pro-poor? Revista Perspeclivar Sociales / Social Perspectives primaveralspring 2007. Vol.9, Num. ¡ ¡ Table 1: Descriptive statistics of relative pro-poorness change in that period. Stali.stics 1 1992 Sample siz.e (bouseholds) 10530 Gini index 0.622 (0.017) Average income pet capita: 2115 (142) [6.52] 1.000 µ.year/ µ.JJm (0.000) Headcount index 0.291 (0.014) Average poverty g¡ip 0.129 (0.()()1)) Conclusion This paper proposes techniques to check for whether growth has been pro-poor. It first reviews different definitions of pro-poorness and argues for the use of methods that can generate results that are robust over classes of pro-poor measures and ranges of poverty lines. It then makes it empirically feasible to test for pro-poomess of growth. To do this, it derives the sampling distribution of the various estimators that are needed to test for absolute and relative pro-poomess. This leads to the convenient use of confidence intervals around the curves that must be ranked in order to conclude that a change has been robustly pro-poor - or anti-poor. 1 1 '1 These statistical techniques are then implemented using Mex.ico's National Income and Expenditure Surveys collected in 1992, 1998 and 2004 and taking fu1l account of tbe sampling design ofthese surveys. We find strong evidence that Mex.ican growth has been absolutely anti-poor between 1992 and 1998, absolutely pro-poor between 1998 and 2004 and between 1992 and 2004, and relatively pro-poor between 1992 and 2004 and between 1998 and 2004. Toe assessm.ent of the period between 1992 and 1998 is, however, statistically too weak to lead to a robust evaluation of this period, and this is true botb for both first and second-order assessments of pro-poomess. Year 1998 10952 20595 0.577 0.524 (0.011) (0.010) 1492 2430 (69) (87) (6.83] 0.705 (0.057) 0.354 (0.022) 0.169 (0.015) (10.88] 1.149 (0.087) 0.106 (0.()()1)) 0.036 (0.004) - ( ...): Standard errors - [...]: Sam.pling design effect Figure 1: Density functions 1 - - 1992 1 ~ ,1 t l •11 º !o----;eoo"J;:"---::i:::----.----,-------. 1200 1800 2-400 3000 · lncome per capita (y) 35 �36 / Has Mexican growth been pro-poor? ,- 37 Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vo/.9, Num. 1 I Figure 2: Lorenz curves Figure 4: Average poverty gap curves:P(z; a = l)for a range of z 45º 1ne 1998 ~ 'ii á al li 11.. c.'-! .2 .4 .8 .6 600 Percentiles (p) ,- 1 11! 1800 2400 3000 Poverly line (Z) Figure 5: 1992 to 1998 is first-order absolutely anti-poor: Pi998(z; a Figure 3: Poverty headcount curves: P(z; a= O)for a range ofz ' ' 1200 1992 2004 = Pim(z; a = O) ---- ---.....• íi~ i '1 1 .ti 11..~ "l 600 o o 600 1200 1800 Poverty line (z) 2400 3000 1200 1800 Poverty Une 2400 - -- NuHhorizontal line - - - - - Difference 1 1·•· ········ Lower bound of 95% confidence interv~ 3000 O) - �38 / Has Mexican growth been pro--poor? Revista Perspectivas Sociales/ Social Perspectives primaveralspring 2()()7. Vol.9. Num. / / Figure 6: 1992 to 1998 is first-order absolutely anti-poor: Q1998(p) - Q1992(p) o---------_-_=.-..-...-..-..-..=...,..,...,-..,-,-...-..-..-..-..-..-..-...-..-..-..-..- - - - - - - - - - -------- -- ------- .... -- . 39 Figure 8: 1998 to 2004 is first-order absolutely pro-poor: Q2()()4(p) - Q1998(p) . ··~- I\ -- -- .... -,,, ~--·..../'\ 1 I ...... '' \ _, g I I \_; 1 1 1 1 1 1 \ í \ 1 1 1 1 ~ .10-----.1-9----.38~----.....f j f - -- -•.7-6------,.95 o+-----...-----.------.---------0 Percentile (p) .19 1--- - - - Null horizontal line - - - - - Difference •······ •· •· Upper bound of 95% oonfidence interval Pi 998(z; a = O) = O) - .J....----~----...------,-----,------, o 600 1--- 1200 1800 Povertyline 3000 2400 NuB horizontal line - - - - - Difference -·····•·· ·• Upper bound of 95% confidence interval .95 1 1 Figure 9: 1992 to 2004 is first-order absolutely pro-poor: P2004 (z; a = P¡992(z;a =0) 0------------------------- 1 .76 ,fjf Pen:entlle (p) Null horizontal fine - - - - • Difference · •·•·•· · ·•· Lower bound of 95% oonfldence interval Figure 7: 1998 to 2004 is first-order absolutely pro-poor: P2004(z; a "? .38 ~t------.----.----...----2400 ---.. o 600 1200 1800 3000 1--- PoYerty line Nul horizontal line - - - - • Difference ·· · ···· ·· · · Upper bound of 95% confidence int~ 1 O) - �40 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales/Social Perspectives primavera/spring 2()()7. Vo/.9, Num. J / Figure 10: 1992 to 2004is first-order absolutely pro-poor: Q2004(p) - Q1992(p) ª o+-----------------------'-•.' \:\ 41 Figure 12: 1992 to 1998 is not statistically first-order relatively pro-poor: Q1993(p)/Q1m(p) - µ,1998/ µ1992 O-f>-------+---------------_.;-.. ... .-·· ............... ..·•·· ...,· ( .. CII .19 1-- .38 .57 .76 Percentile (p) Null horizontal line - - - - - Difference ··• ........ Lower bound of 95% confidence interval 1 ! .95 1 .02 .2 .38 .56 Pen:entile (p) .74 .92 , - - Null horizontal line - - - - • Difference 1 ..... ··· ··· Lower bound of 95% confidence interval Figure 11: 1992 to 1998 is not statistically first-order relatively pro-poor: P1998(gz; a= O) - Pi992(z;a = O) Figure 13: 1992 to 1998 is not statistically second-order relatively pro-poor: Pi998(9z; a= 1) - Pim(z; a= 1) o .. -f""'....._,-,----'-:.:,-..-...-. - - - - - - - - - - - - - - ......... ·.. '' CII '' q ·····......,._.................................__ ,_......._ 1 o '' '' '' '' ' ',, -----------------------1200 1800 Poverty line (z) 2400 , - - NuU horizontal fine - - - - - Difference 1 · ·······.. · Upper bound of 95% confidence interval 3000 8.. ;----..-----..-----,------,,-----, o 600 1200 1800 2400 Poverty line (z) , - - Nu8 horizontal line - - - - - Diffeience 1 ·•·•· · ·· ··· Upper bound of 95% confidence interval 3000 �42 / Has Mexican growth been pro-poor? Revista Perspectivas Sociales / Sacia/ Perspectives primavera/spring 2007. Vo/.9, Num. I I Figure 14: 1992 to 1998 is not statistically second-order relatively pro-poor: C1998(p)/Cum(p) - Figure 16: 1998 to 2004 is first-order relatively pro-poor: C2004(p)/C1998(p) - µ1998/µ1992 /.½J:1J4/µ1995 "' ------------- ---- --, ~ .02+----,.2,------.38.-----.• 56----,J-4-----, .92 1 '1 \ \ ...,q U')-+------,,------.----.------,-----, 1 O Pen::entile(p) lfll 1 .38 .57 .76 .95 Percentile {p) Null horizontal line - - - - - D'rfference 1 --· --...... Lower bound of 95% conlidence interval = O) - Pi998 (z;a=O) 1 .19 ¡--- , - - - Null horizontalline - - - - - Difference 1 ······-···· Lower bound of 95% confidence interval Figure 15: 1998 to 2004 is first-order relatively pro-poor: P2004 (gz; a 43 Figure 17: 1998 to 2004 is strongly second-order relatively pro-poor: P2004(gz¡a = 1) - Pi998(z; a= 1) 111 U') .------.-----.------,,------, .. + o - - - - -600 1200 1800 2400 3000 Poverty llne (z) , - - - Nun horizontal line - - - - - Difference 1 .......... · Upper bound of 95% confidence interval 1-- NuB horizontal line - - - - - Difference ... · ...... · Upper bound of 95% conlidence interval 1 �44 / Has Mexican growth been pro-poor? Figure 18: 45 Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vo/.9, Num. 1 / 1998 to 2004 is strongly second-order relatively pro-poor: C.};(X)4(p)/C1~(p) - JJ,,dXJ4jJJ,1m Figure 20: 1992 to 2004 is first-order relatively pro-poor: C2004(p)/C1992(p) - JúJJxM/1'1992 (\1 "l o +-----..------.-----------,-----,------, .02 .2 .38 .56 .74 .92 Percentile (p) --1· · ·· •· · · ·· · Null horizontal line - - - - - Dlfference Lov,,er bound of 95% confidence interval 1 --1· ·····•···· 1200 1800 Poverty line (z) .38 .56 .74 - - - Null horizontal line - - - - - Difference 1 1······ · ··· · Lower bound ol 95% confidence interval = O) - o~------------------------ 600 .2 Percentlle (p) Figure 19: 1992 to 2004 is first-order relatively pro-poor: P2004 (gz; a P1992(z;a = O) o .02 2400 NuB horizontal line - - - - - Difference Upper bound of 95% confidence interval 3000 1 .92 �46 / Has Mexican growth been pro-poor? References Bour Gui Gnon, F. (2003). "The poverty-growth-inequality triangle," Tech. rep., Paris, Agence francáise de développement. Ravallion, M. and S. Chen (2003). "Measuring Pro-poor Growth," Economics Letters, 78, 93- 99. Bruno, M., M. Ravallion, and L. Squire (1999). "Equity and Growth in Developing Countries: Old and New Perspectives on the Policy Issues," SSRN Working Paper Series 604912. Ravallion, M. and G. Datt (2002). "Why Has Economic Growth Been More Pro-poor in Sorne States of India Toan Others?" Joumal ofDevelopment Economics, 68, 381-400. Dollar , D. and A. Kraay (2002). "Growth Is Good for the Poor," Journal ofEconomic Growth, 7, 195-225. Son, H. (2004). "A note on pro-poor growth," Economics Letters, 82, 307- 314. Duelos , J . Y. and A. Araar (2006). Poverty and Equity Measure-ment, Policy, and Estimation with DAD, Berlin and Ottawa: Springer and IDRC. United-Nations (2000). "A Better World for Ali," Tech. rep., United Nations, New York. ' Eastwood, R. and M. Lipton (2001 ). "Pro-poor Growth and Pro-Growth Poverty Reduction: What do they Mean? What does the Evidence Mean? What can Policymakers do?" Asian Development Review, 19, 1- 37. 1 Essama-Nssah, B. (2005). "A unified framework for pro-poor growth analysis," Economics Letters, 89, 216-221. 1 1 Growth?" Discussion Paper #96, lbero-America lnstitute for Economic Research, Georg-August-Universitiit, Gttingen. Me Culloch, N. and B. Baulch (1999). "Tracking pro-poor growth," ID21 insights #31, Sussex, Institute of Development Studies. 1 '¡ 47 Araar, A. (2006). "Poverty, Inequality and Stochastic Dominance, Theory and Practice: Illustration with Burkina Faso Surveys," CIRPÉE Working Paper #0634. Duelos, J. - Y. and Q. Wodon (2004). "What is "Pro-Poor"?" CIRPÉE Working Paper #0425. 1 1 Revista Perspectivas Sociales / Social P~lives primaveralspring 2007. Vol.9. Num. I / Kakwani, N., S. Khandker, and H. Son (2003). "Poverty Equivalent Growth Rate: With Applications to Korea and Thailand," Tech. rep., Economic Commission for Africa. Kakwani, N. and E. Pernia (2000). "What is Pro Poor Growth?" Asian Development Review, 18, 1- 16. Klasen, S. (2003). "In Search ofthe Holy Grail: How to Achieve Pro-Poor World-Bank (2000). The Qua/ity of Growth, New York: Oxford University Press. �Revista Perspectivas Sociales/ Social PerspeclÍlleS primavera/spring 2007. Vo/.9, Man. J I Pág. 49-63 49 Pro-Poor Food Taxation and Subsidy Reforms in Mexico* Matbieu Audett, Paul Mak:dissi:j: Abdelkrim Araar§, Jean-Yves Duclos,r Abstract This paper uses a methodology developed in Mak:dissi and Richard (2007) to identify pro-poor tax reforms in Mexico using the ENIGH 2004 data set. This method, which is based on stochastic dominance, enables us to identify tax reforms that will be deemed as pro-poor by a wide spectrum of poverty analysts. Keywords Stochastic dominance, Pro-poor, Tax reform. Resumen 1, Este trabajo utiliza una metodología desarrollada por Makdissi y Richard (2007) para identificar las reformas impositivas a favor de los pobres con datos de México utilizando la información de 2004 de la Encuesta Nacional de Ingresos y Gastos de los Hogares. Esta metodología, basada 11 *We are grate.ful to Dr. Lourdes Treviño and Professor Jorge Va/ero Gilfor their invita/ion to present this paper at the Eight Symposium on "Capital Humano, Crecimiento, Pobreza: Problemática Mexicana" that took place in Monterrey, Mexico on October 12 and 13, 2006, and to the participants at that conference for valuable comments. f GRÉDI, Université de Sherbrooke, 2500, boulevard de l'Université, Sherbrooke, Québec, Canada, JlK 2Rl; Email: mathieuaudetX@hotmail.com f Département d'économique, CIRPÉE and GRÉDJ, Université de Sherbrooke, 2500, boulevard de /'Université, Sherbrooke, Québec, Ganada, JJK 2Rl; Email: paul. makdissi@usherbrooke.ca § Département d'économique and ClRPÉE, Pavillon De Séve, Université Lava/. SainteFoy, Québec, Canada, GlK 7P4; Email: aabd@ecn.ulaval.ca 1Département d'économique and CIRPÉE, PavillonDe Séve, UniversitéLava/, SainteFoy, Québec, Canada, GlK 7P4; Emailjyves@ecn.ulaval.ca 1 1 '1 1 1 1 1 ,, 1 1 ' . .. ISSN 1405-1133 C 2007 Universidad Autónoma de Nuevo León, Univeisity ofTexas ofAustin, University ofTexas of Arlington, Uoiversity ofTeooessee, Universidad Juárez del Estado de Durango, Universidad de Colima �50 I Pro-Poor FoodTaxalíon and Subsidy Reforms in Mexico en la dominancia estocástica, nos permite identificar que reformas impositivas podrían ser consideradas por un gran número de analistas como favorables a los pobres. Palabras clave: Dominancia estocástica, Pro pobres, Reforma impositiva. Revista Perspectivos Sociales / Social Perspectives primavera/spring 1007. Vol.9, Num.. J / 51 section will present a brief methodological framework devised by Makdissi and Richard (2007). The third section presents our analysis of indirect tax reforms for Mexico utilizing the ENIGH 2004 database. In closing, we present a brief conclusion and possibilities for future research. Methodological Framework Introduction 1 1 According to Santoro (2006), the economic literature has used three different approaches to analyze the impact of marginal tax reforms. The :first approach is based on the work of Ahmad and Stem ( 1984) and uses a specific social welfare function. The second approach identifies avenues for tax reform based on aversion to inequality and symmetry of the social welfare function. This approach was pioneered by Yitzhaki and Thirsk (1990), Yitzhaki and Slemrod (1991) and Myashar and Yitzhaki (1996). The third approach considers that marginal tax reforms can also be used as instruments for reducing poverty. It is based inter alia on the recent papers of Makdissi and Wodon (2002), Liberati (2003) and Duelos, Makdissi and Wodon (2006). This last approach allows the identi:fication of tax reforms that will be considered poverty-reducing by a wide spectrum of poverty measures. In this article, we check whether sorne indirect tax reforms affecting the price of food could also be pro poor. By this, i t is meant that not only must a tax reform lead to a reduction ofpoverty levels, but it must also be deemed "equitable for the poor", in that its positive impact must accrue disproportionately more to the poor. To accomplish this, we adoptan analytical approach proposed by Makdissi and Richard (2007) . Building on this approach, this article attempts to identify Mexican food consumption goods which, through tax reform, sbould be used to favor pro-poor growth. In doing this, we also follow on the steps of Araar, Duelos, Audet and Makdissi (2007), which asked whether growth in Mexico between tbe period of 1992 and 2004 was pro poor. The remainder ofthe article wil1 be divided as follows. Toe following Let us suppose that the government wishes to put into place a propoor indirect tax reform. Let us consider three possible scenarios the government can face: 1. the governmentruns a budgetary surplus and wishes to implementa marginal reduction in the tax (or a marginal increase in the subsidy) on good i; 2. the government wishes to reduce the budgetary deficit by implementing a marginal increase in the tax (or a marginal decrease in the subsidy) on good i; 3. the government wishes to implementarevenue-neutral indirecttaxre form. It must therefore finance a marginal tax reduction on good i (or a marginal increase in its subsidy) with a marginal increase in the tax (ora marginal decrease in the subsidy) on good j :,=- i. Poverty Measurement We consider that poverty can be measured by an additive poverty index. This class of indices has the following form: P (z) = [ p (y,z)dF(y), (1) where y is real income, z is the poverty line, F (-) is the cumulative distribution of income with support over [O, w], and p (y, z ) is a function that measures the poverty of an individual with an income y usmg a poverty line z . We suppose that p (y, z ) 2: O and that p (y, z ) = O for all �52 Revista Perspectivas Sociales / Social Perspeclives primavera/spring 2()()7. Vo/.9, Num. I / / Pro-Poor Food Taxation and Subsidy Reforms in Mexico y > z. Duelos and Makdissi (2004) use the properties of this function to 53 Budget impact define classes of poverty indices Il5 • These classes are defined by: p(y,z)EC'(z), } n •(z) = P(z) (-l i p(i) (y,z) ~ Ofor i = O, 1, 2, ... ,s, , { p<1) (z,z) = Ofort = O, 1, 2, ... ,s (2) where 8• represents the set of continuous functions that are s-times differentiable on [o, aj. When s = 1, an increase in the income of any one individual will weakly reduce the poverty index. This class of indices is thus Paretian. Toe indices are also symmetrical due to the fact that exchanging incomes between two individuals does not affect poverty. This type ofindices are said to satisfy Pen's (1971) principies for comparing distributions. The poverty indices included in 112 are also convex. This implies that they respect the Pigou-Dalton principie of transfer, which states that a transfer from any one individual to a poorer individual should weakly decrease poverty. In addition obeying the above principles, the poverty indices belonging to IJ3 must also respect the Kolm principie (1976) of transfers, which states that a Pigou-Dalton transfer that takes place at the bottom of the distribution should have a greater impact on poverty then one tak.ing place higher up in the distribution. Thus, a progressive transfer occurring within a lower part of the distribution will reduce poverty even if it is accompanied by a regressive transfer higher up in the distribution. Indices of a class IJswith s greater then 3 can be ethically interpreted by using the generalized transfer principie proposed by Fishburn and Willig (1984). This principie states that the greater the order s, the greater is the sensibility of an index to changes occurring in the lower part of the distribution. The Foster, Greer and Thorbecke (1984) are a particular example of additive poverty measures. Other examples of such indices are given by Watts (1968), Clark, Hemming and Ulph (1981) and Chakravarty (1983). In the case of a revenue neutrality requirement, we must analyze the impact of a tax reform on overall tax revenue. To do this, we suppose that the economy has K consumption goods. Say that the government wants to reduce marginally a tax (orto increase marginally a subsidy) on a good i and fiñance this with a marginal increase in tax (or a marginal reduction of subsidy) on a goodj . Denote by R the per capita tax revenue of the indirect tax system: K R(q) = L tkXk(q), (3) k~l where Xk (q) is average consumption of good k, q is a vector of consumption prices, and tk is the tax imposed on good k (tk < O if k is subsidized). Toe price to the producer is supposed to be constant and is fixed as 1 so that q"' the price of good k, equals 1 + t1c Toe impact of the marginal reform on per capita tax revenue is then: dR = { [x¡(q) + t tk a~q)] [x;(q) + t tkª~!q)] dt;}. dt¡ + (4) Revenue neutrality implies that dR = O. Combined to (4), this leads to: dt · = X;(q) h , "f ( X -( ) ) w ere 'Y J q 1 1 " K ~ + X.W L...r-1 tk Ot; l + X;(q) L r-1 t i, i ,:-,.J< oxk(q). 8t; (5) Wildasin (1984) describes "f as the cost effi.ciency ratio of obtaining one dollar ofpublic funds by taxing good j to subsidize good i. Yitzhaki and Thirsk (1990) and Yitzhaki and Slemro~ (1991) find that if "f is superior to one, it is impossible to have a second-order welfare dominant reform due to the efficiency loss incurred. Makdissi and Wodon (2002) note, however, that in a poverty analysis perspective, it is possible to have a reform that is dominant at ali orders of stochastic dominance even when 'Y is greater than one, so long as that part of the burden is supported by the non poor. Identífying pro poor reforms Toe impact of a marginal change dtk to the tax on a good k will impact �54 / Pro-Poor Food Taxation and Subsidy Reforms in Mexico Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Vo/.9, Num. / / 55 the poverty level of an individual with income y by (13) (6) Using Roy's identity and setting the vector or reference prices to the current price vector, the change in real income produced by a marginal change in the tax on good k is given by (see for instance Besley and Kanbur 1988) and that (14) By using the results of equation (13) and (14), we obtain (15) (7) and wherexk (y, q) is the Marshallian demand ofgood k at the vector ofcurrent prices. If we introduce this result into equation (6), we have (8) We do not wish, however, to determine if a tax reform reduces or increases poverty, but whether it can be considered pro-poor in the sense ofDuclos and Wodon (2004) and Araar et al. (2007). For relative propoomess, this is accomplished by comparing p ( z) for the second period top (y, z) for the first period, using a relative "norm" g. For absolute pro-poomess with an absolute norm of a, we compare p (y - a, z ) with p (y, z). For the purposes of this paper, we suppose that the relative norm g is set to growth in average real income, while the absolute norm a represents the absolute change in average real income. Let: m, 1 1 y•R = _ Y_ l+g y•A = y-a '1 1 '1 ,,,,1 op•R(y,z) &t,. &¡lA (y,z) lJtk = op(y, z) 0y•R ~&t,. = op (y, z) é)y•A ~ot,.· (16) To obtain the impact on total poverty, we integrate (15) and (16) over the entire income distribution. The result is - We now wish to determine how y*R and y*A vary with a marginal variation in t1c Makdissi and Richard (2007) show that 0 &p(J¡, z) [xé(y, q) -yµ Xk] dF (y), -¡¡¡¡- (17) and - 8p (y,z) [ ) [ 0 ~ x.(y) -x. dF (y). (18) Makdissi and Richard (2007) then define relative (CD *R) and absolute (CD*A) pro-poor Consumption Dominance curves as follows: (9) (10) (11) (12) [ { ['t-"-¡}f(,) C[)f"(y) = J CI>f" (z)dx foc, - 1 for s ~ 2, (19) o and { ["l\'1-1]¡<,¡ r.u - 1 CD:,, (y) = f • (20) CDt•(x )dx for s~ 2. Comparing equation (17) to equation (19), and equation (18) to equation (20), Makdissi and Richard (2007) show a series of results. First, a marginal increase in the tax on good i is relatively pro poor for �56 Revista Perspectivas Sociales/ Social Perspeclives primavera/spring 2007. Vo/.9, Num. J / / Pro-Poor Food Taxation and Subsidy Reforms in Mexico all indices P (z) E n• {z) and for all poverty lines z E [O, z+j if (21) Second in a revenue-neutral context, a marginal tax reduction for good i finanded by a marginal increase of the consuroption tax on good/ is relatively pro poor for all indicess P (z) E n• (z) and for all poverty lines z E [Oif+l (22) Makdissi and Richard (2007) propose similar in the absolute pro-poor framework. A marginal increase in a tax on good i is absolutely pro-poor for all indices P (z) en• (z) and for all poverty lines z E [O, z+] if CDf" (y) :5 O, Vy E (o,z+]. (23) Again, in a revenue-neutral framework, a marginal tax increase on goo_d i financed by a marginal decrease of the consuroption tax on good j 1s absolutely pro poor for all indices P(z) e n•(z) and for all poverty lines z e [o,z+Jif (24) 1, An Illustration Using ENIGH Data In this section, we apply the above methodology to Mexican data. The data used for our application is the National Income and Expenditure (ENIGH) Survey collected in 2004. This survey is representative at the national level. Toe objective of the ENIGH surveys was to collect information on incomes and expenditures, goods and services used for self-consumption, as well as socio-econoroic characteristics and labor market activities of all household members. The sampling process was stratified and multi-staged, with the final sampling units being households and all of their members. As is comroon in South America, we use total income per capita as the measure of living standards for all members of a household. To correct for spatial variation in prices, we assess all incomes in reference 57 to rural prices and multiply urban household incomes by the ratio of rural to urban poverty lines. We use as a guide a 2004 rural poverty line set to 550 pesos per month per capita. We weight households by the product ofhousehold size and household sampling weight. To simplify the interpretation of figures and discussions, we normalize income by that rural poverty line so that a household with an income equal to one is at the level of the rural poverty line and a household with an income of 2 has a "real" income equal to twice that line. To simplify the analysis, we suppose that there is no relative econoroic efficiency advantage of taxing one good compared to another (thus we have -r= 1). This study considers tax reforms affecting Mexican food goods. In 2004, all food goods are exempt of value-added taxes (VAT) in Mexico. A few ofthese goods are also subsidized. Table 1 presents the percentage of total food expenditure allocated to various food goods. Toe primary food basket varíes greatly depending on the income quintile. For exarople, households in the poorest income quintile have a greater share of their total food spending on cereals (25.88%) and vegetables (17.03%) than the richest quintile, which spends relatively more on meats (26.23%) and fruits (6.49%). Figure 1 presents the C DR,:s curves for different food categories. The fust conclusion that can be drawn is that a marginal reduction in taxes on any category of food goods would be pro poor and that this conclusion is valid for any poverty index which satisfies the Pen and Pigou-Dalton principies as well as for all poverty lines. Therefore, any marginal increase in tax on any food good would be considered relatively anti poor. Analysts propose from time to time imposing VAT on certain food goods. These recommendations would be relatively anti poor in the Mexican context. A different issue, however, deals with what would happen if a marginal increase in a food subsidy were financed by a marginal increase in the tax on another food good. Figure 1 suggests that increasing marginally subsidies on sugar, oils, cereals, vegetables '\nd tubercules and financing this by a marginal increase in the tax on zoods such as meat, fish, milk and fruits would be relatively pro poor ívr any poverty index which satisfies the Pen and Pigou-Dalton principies as well as for all poverty lines. Hence, it is important to consider the use to which increases in tax revenues are put to know whether a tax �58 / Pro-Poor Food Taxation ami Subsidy Reforms in Mexico 1 59 reform is pro poor or not1 References Figure 2 presents the C [)A:s curves for various categories of food goods and for s = 2. A marginal reduction in taxes on sugar or oil would be absolutely pro poor. This conclusion holds for any poverty index which satisfies the Pen and Pigou-Dalton principles as well as for all poverty lines. Further, a marginal increase in a tax on eggs would be absolutely pro poor up to 1.5 times the current official poverty line. Finally, a marginal increase in consumption taxes on any other category of food goods would be absolutely pro poor. This conclusion is true for the above-stated principles and for all poverty lines. With respect to revenue-neutral reforms, increasing marginally subsidies on sugar, oils, cereals, vegetables and tubercules and financing this by a marginal increase in the tax on goods such as meat, fish, milk and fruits would be absolutely pro poor at the second-order and for all poverty lines. This last conelusion is identical to the one at the end of the last paragraph - indeed, with 'Y= 1, a revenue-neutral reform which is absolutely pro poor must also be relatively pro poor, and conversely. Ahmad, E. and N.H. Stem (1984). Toe Theory of Reform and Indian Indirect Taxes, Joumal ofPublic Economics, 25, 259-298. Conclusion 1 Revista Perspectivos Sociales I Social Perspectives primavera/spring 2()()7. Vo/.9, Num. J / In this article, we use a methodology proposed by Makdissi and Richard (2007) to analyze the possible impact of sorne reforms of the indirect taxation of food goods in Mexico. Results show that a VAT on food goods would be relatively anti poor. However, if this reform were accompanied by increases in subsidies of certain categories of food goods, it might be possible to implement a pro-poor reform both in relative and in absolute terms. Future research could evaluate the whole of Mexico's tax system, both direct and indirect. An analysis of existing tax and transfer programs could also be made by adapting the methodology proposed in Duelos, Makdissi and Wodon (2005). It would also be interesting to study the structure of public utility tariffs using the theoretical framework developed in Makdissi and Wodon (2007) as well as the framework of Araar and Duelos (2007) based on "pro-poor growth curves". Caution should also be exercised when dealing with reforms that suggest increasing taxes on milk. In Mexico, a share ofspending on milk by the poorest ofthe population is indeed subsidized through the long running LICONS program. 1 Araar, A., and J.-Y. Duelos (2007). Pro-poor marginal tax reforms . ' numeo. Araar, A.; J.-Y. Duelos, M. Audet and P. Makdissi (2007). Has Mexican growth been pro-poor?, mimeo. Besley, T. and R. Kanbur (1988). "Food Subsidies and Poverty Alleviation", The Economic Joumal, 98, 701-719. Chakravarty, S.R. (1983). "A New Index of Poverty", Mathematical Social Sciences, 6, 307-313. Clark, S.; R. Hemming and D. Ulph (1981). "On Indices for the Measurement of Poverty", Economic Joumal, 91, 515-526. Duelos, J.-Y. and P. Makdissi (2004). "Restricted and Unrestricted Do~ance for Welfare, lnequality and Poverty Orderings", Joumal of Publlc Economic Theory, 6, 145-164. Duelos, J.-Y.; P. Makdissi and Q. Wodon (2005). ''Poverty-Dominant Transfer Programs: Toe Role ofTargeting andAllocation Rules", Journal ofDevelopment Economics, 77, 53-73. Duelos, J.-Y.; P. Makdissi and Q. Wodon (2006). Socially-Improving Tax Reforms, mimeo. Duelos, J.-Y. and Q. Wodon (2004). What is Pro-Poor? CIRPÉE Working Paper #0425. Fishburn, P.C. and R.D. Willig (1984). "Transfer Principies in Income Redistribution", Joumal ofPublic Economics, 25, 323-328. �60 I Pro-Poor Food Taxation and Subsidy Rejorms in Mexico Foster, J.; J. Greer and E. Thorbecke (1984). "A Class ofDecomposable Poverty Measures", Econometrica, 52, 761-776. Kolm, S.-C. (1976). "Unequal Inequality: I", Journal of Economic Theory, 12, 416-442. Liberati, P. (2003). "Poverty Reducing Reforms and Subgroups Consumption Dominance Curves", Review oflncome and Wealth, 49, 589-601. Makdissi, P. and Q. Wodon (2002). "Consumption Dominance Curves: Testing for the lmpact oflndirect Tax Reforms on Poverty'', Economics Letters, 75, 227-235. Makdissi, P. and Q. Wodon (2007). "Poverty-Reducing and WelfareImproving Marginal Public Price and Price Cap Reforrns", forthcoming in Journal ofPublic Economic Theory. Makdissi, mimeo. :r. et P. Richard (2007). Pro-Poor Indirect Tax Reforms, Mayshar, J. and S. Yttzhaki (1995). "Dalton Improving Tax Reform", American Economic Review, 85, 793-807. Peo, J. ( 1971 ). lncome Distribution: facts, theories, policies, Preaeger, NewYork. Santoro, A. (2006). "Marginal commodity tax reforms: a survey", forthcoming in Journal ofEconomic Surveys. Watts, H.W. (1968). "An Economic Definition of Poverty", in D.P. Moynihan (ed.), On Understanding Poverty, Basic Books, New York. Wildasin, D. (1984). "On Public Good Provision With Distortionary Taxation", Economic lnquiry, 22, 227-243. Yitzhaki, S. et J. Slemrod (1991 ). "Welfare Dominance: An Application to Commodity Taxation",American Economic Review, 81 , 480-496. Revista Perspectivas Sociales / Social Perspecfives primaveralspring 2007. Vo/.9, Num. / J 61 Yitzhaki, S. and W. Thirsk ( 1990). "Welfare Dominance and the Design of Excise Taxation in the Cote d ' Ivoire", Joumal of Development Economics, 33, 1-18. �62 / Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vol.9. Num. I / Pro-Poor Food Taxation and Subsidy Reforms in Mexico Figure 2. Absolute Pro-Poor Consnmption Dominance Curves, :zmi Order. - Table 1 Share of total food expenditure by good and exoenditure auintiles Expenditure quintiles 4 Richest 3 2 Poorest Good 15,90 18,95 21 ,20 23,92 25,88 Cereales 16,47 14,88 14,35 12,77 9,21 Leche 26,23 27,79 25,21 22,99 17,37 Carnes 3,74 2,95 2,34 2,08 2,16 Pescado 2,48 3,28 3,84 4,68 5,58 Huevo 1,27 1,76 1,55 2,24 3,08 Aceites 1,29 1,51 1,62 1,78 Tuberculo 1,92 11 ,17 13,30 13,92 15,18 17,03 Verduras 4 ,56 6 ,49 3,71 3,12 2,27 Frutas 1,18 1,03 1,75 1,42 2,91 Azucar 1,11 1,19 0 ,95 1,12 1,47 Café 12,75 9 ,67 8 ,94 8 ,31 11 ,20 Otros ali - 0,1 - - - - - - - - - - - - - - - - - ~ ~ 0,05 0,4 P8IIC8do -0, 1 0,2 " - 0,15 0,1 0,05 o .....,~:.;;.=...;;.----------------' ~~~~~~~~~@~~~~~~~ <:>- <:), <:), <:>- <:)• " ' "' "' " ' '\,• '\,• '\,• '\,• '\,• -0,2 Fuas -0,25 Ama Café -0,3 0,25 1, --Vermm ~ -0,15 Canes Pescado 0,3 Lecha -Hieo r-----------------------... -----Cereates 0,35 -:_-_- Cere ...,_ al_88.. · c... Figure l. Relative Pro-Poor Consumption Dominance Curves, 2® Order. 0,45 63 Tuben)tjo. -VIIRkns Fnns kDlla CeM -----------------------1 �Revista Per.rpecttvas Sociales / Social Perspectives primavera/spring 2007. Yol.9, Num. I I Pág. 65-88 65 Gender-bias in Education Opportunities for Population Aged 12-18 in Mexico: 1992-2004 Ernesto Aguayo, Joana Chapa, Erick Rangel, Lourdes Treviño, Jorge Valerot Abstract There is considerable evidence that resources are not allocated randomly within households, and that resources are unequally distributed within the family in many developing countries. Such an unequal distribution of goods usually takes the form of a bias against females. For example, girls lag markedly behind boys in schooling in many developing countries even though this gender gap has been declining in recent years. Using an OLS-Robust model anda ML-Random Effects model for the years 1992, 1998 and 2004 of ENIGH, we did not find enough statistical evidence to support the idea that poor families, nether in rural nor in urban areas, provide more education to their 12 to 18 years old sons or daughters. In fact, contrary to the common belief, we found that non-poor families, invest more in the education of their daughters, especially in the urban areas. However, this education discrimination against male children has been decreasing over the years. It is also found that female head of households are more likely to have children with higher levels of schooling and that children having both parents at home or having older brothers or sisters present higher levels of educational attainment. Keywords Education, Discrinrination, Poverty, Intra-family allocation, Mex.ico. f Facuitad de Economía, Universidad Autónoma de Nuevo León, Loma Redonda 1515 Pte., Monterrey, NL, México, 66450'. Corresponding author:jvalero@faeco.uanl.mx ISSN 1405-1 133 (} 2007 Universidad Autónoma de Nuevo León. University ofTexas of Austin, University ofTexas ofA.rlingtoo, University ofTennessee, Univer.;idad Juárez del Estado de Durango, Universidad de Colima. �66 / Gender-bias in Education Opportunitiesfor Population Aged 12-18 in Mexico: 1992-2004 Resumen Hay considerable evidencia de que los recursos no están asignados aleatoriamente en los hogares y de que los recursos se distribuyen desigualmente dentro de la familia en muchos países en desarrollo. Usualmente esta distribución desigual del ingreso toma la forma de un sesgo en contra de las mujeres. Por ejemplo, las niñas van retrasadas respecto a los niños en la escuela en muchos países en desarrollo aunque este rezago por género ha ido declinando en años recientes. Utilizando un modelo de Mínimos Cuadrados Ordinarios robustos y un modelo de Máxima Verosimilitud con efectos aleatorios y con datos de las Encuesta Nacional de Ingresos y Gastos de los Hogares para los años 1992, 1998 y 2004, no encontramos suficiente evidencia estadística que apoye la idea de que los hogares pobres, ni en las áreas urbanas ni en las rurales, provean más educación a sus hijos que a sus hijas. De hecho, contrario a lo que comúnmente se cree, encontramos que las familias no pobres invierten más en la educación de sus hijas, especialmente en las áreas urbanas. Sin embargo, esta discriminación contra los hijos hombres ha ido decreciendo con los años. También se encuentra que las mujeres que son jefas de familia tienen mayor probabilidad de tener hijos o hijas con niveles más altos de escolaridad y que los hijos e hijas que tienen ambos padres en el hogar o que tienen hermanos o hermanas mayores presentan más altos grados de educación. Palabras clave Educación, Discriminación, Pobreza, Asignación intrafamiliar, México. .. Introduction Toe intra-household allocation of resources has become one of the most important issues in human capital research. There is considerable evidence that resources are not allocated randomly within households, 1 See, for instance, Deolalikar (1993) for Indonesia, Parish ami WU/is (1993) for Taiwan, Schultz (1993) on investments in health and education in many groups ofdeveloping countries. Thomas (1990) finds evidence ofnon-random distribution or resources in Brazi/ian households, and, to a lesser extent, Deaton (1987) and Svedberg (1990) find gender bias in Africa. Haddad et al. (1994) provide an overview ofthe literature on within-household resource al/ocation. Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Vo/.9, Num. / / 67 and that resources are unequally distributed within the farnily in many developing countries. 1 Becker (1965, 1981) conceives the family acting as a single decision maker which regards child education as an investment decision. Models of intra-household allocation of goods that follow Becker's approach assume that the allocation is determined in one of the following three ways: i) parents allocate resources based on the differential labor market returns to boys and girls (Rosenzweig and Schultz 1982); ii) parents allocate resources according to their own utility, which depends on the well-being of their children (Behrman et al 1982, Behrman 1988); iii) households allocate resources based on the productivity of individual members (Pitt et al 1990); However, several authors have pointed out the limitations of those approaches and proposed alternative collective models for the analysis of household behavior. Those models assume resources are allocated according to the relative bargaining power ofthe family members (Manser and Brown 1980, McElroy and Horney 1981, Ulph 1988 and 1990, Thomas 1990, Haddad and Hoddinott 1991, Chiappori 1992, Lundberg and Pollak 1993, Wolley 1993, and Echeverria and Merlo 1999). Such an unequal distribution of goods usually takes the form of a bias against females. For example, Bardhan (1984), Behrman (1988), Harriss (1990), Rosenzweig and Schultz (1982), Sen (1984), and Sen and Sengupta (1983) provide evidence, based on mortality rates and human capital investments, that gender bias is important in explaining the household expenditures on health, nutrition and education among children. Pitt and Rosenzweig (1990), Parish and Willis (1993), Quisumbing (1994), and others have also worked on the effect of gender bias on investments in childrend' s human capital. Brinton (1988) developed the concept ofhuman capital system. In this system, social and economic institutions - such as family, educational system and work organization- share the responsibilities ofhuman capital development across the individual' s life cycle. lt is argued that a cross cultural perspective in gender strati.fication theory helps understand gender stratifi.cation in countries with different social, economical and cultural characteristics than American ones. Under this concept, differential parents' investment in sons and daughters is explained by: parents' �68 / Gender-bias in Education Opportunitiesfor Population Aged 12-18 in Mexico: 1992-2004 perception of sex discrimination by employers, parents' control over resources for investment in children, extent of government support, sex preference ofparents, female marriage behavior and degree of flexibility in life cycle timing human capital development decisions. 2 Furthermore, girls lag markedly behind boys in schooling in many developing countries even though this gender gap has been declining in recent years (King and Hill 1993; Behrman 1993). Alderman et al (1996) report that, in 1990, girls tended to receive less schooling than boys, particularly in rural areas, low-income countries, and in South Asia. According to the World Bank (2005), in 1990, secondary school enrollment in low-income countries was 26 percent for girls and 42 percent for boys.3 By 2001, female secondary enrollment had increased to 41 percent as compared to 51 percent for male enrollment.4 Toe existence and sources of gender bias has become highly relevant for the case of Mexico where the government has been implementing social programs aimed at the reduction of gender inequality under the presumption that there is discrimination against girls in education opportunities. Moreover, in 2004, Secretaria de Desarrollo Social (SEDESOL) conducted the Primera Encuesta Nacional sobre Discriminacion en Mexico (First Nacional Survey on Discrimination in Mexico). It is reported that 15% of the respondents think they should not invest in their daughters' education because they will end up getting married. 1, 1 1 ' '1 1 1 111 ' 1 Toe aim of this paper is to determine whether there is evidence for differences by gender in the allocation of household resources. We will focus on child education, as measured by the number of years of schooling completed. Using an OLS-Robust model anda ML-Random Effects model for the years 1992, 1998 and 2004, we did not find enough 1111t¡ 2 For instance, Brinton (1988) found that Japan has a system ofhuman capital development that encourages greater gender stratification (in favor ofmale children) than the American system. 3 Based on gross enrollment ratio which is the ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the /evel of secondary education. 4 Primary male. school enrollment in 2001 was 72 percent for fema/e and 82 percent for Revista Perspectivas Sociales / Social Perspeclives primavera/spring 2007. Vo/.9, Num. / / 69 statistical evidence to support the idea that poor families, nether in rural nor in urban areas, provide more education to ·their 12 to 18 years old sons or daughters. In fact, contrary to the common belief, we found that non-poor families, invest more in the education of their daughters, especially in the urban areas. Fortunately, this education discrimination against male children has been decreasing over the years. It is also found that female head of households are more likely to have children with higher levels of schooling and that children having both parents at home or having older brothers or sisters present higher levels of educational attainment. Toe remainder of this paper is organized as follows. Section II provides sorne background on Mexico's educational gender gaps. Section m describes de data used and section IV specifies the model. Section V presents the results and section VI concludes the paper . Educational Gender Gaps in Mexico lncreasing human capital investments in children is considered to be among the most effective ways of encouraging growth and ofalleviating poverty in developing countries. To stimulate such investments, many governments in Latín America and Asia have initiated programs to provide financia! incentives for families to send their children to school. 5 In 1997, the Mexican government created Programa de Educación, Salud y Alimentación (PROGRESA), which in 2002 became OPORTIJNIDADES. This program provides focalized aid on education, health and food with the objective of forming human capital in the poorest communities and families in Mexico (Parker and Scott 2001 ). Toe aid for education takes the forro of monetary transfers to fam.ilies that are contingent upon their children 's regular attendance at school. Toe transfer amount varíes with the child 's grade leve! and is greatest for children in secondary school. Toe benefit level is also slightly higher for female children who are traditionally thought to have lower secondary school enrollment levels. 5 Such programs exist, for instance, in Bangladesh, Pakistan, Argentina, Chile, Colombia, Brazil, Nicaragua, and Honduras. See Berhman, Segupta and Todd (2001). �70 / Gender-bias in Education Opportunitiesfor Populotion Aged 12-18 in Mexico: 1992-2004 Notably remarkable is the fact, however, that, according to the World Bank (2005), between 1990 and 2001, secondary school enrolhnent in Mexico has been higher for female than for male. Table 1 shows that, in 1990, this ratio was 54 percent for girls as compared to 53 percent for boys. Toe gender gap widened by 2001 when secondary enrolhnent for female was 78 percent; 5 percentage points above that for male. Table 1. School Enrolhnent by Country Group and Education Level Male Primarv schoo,--1 Lowincome Lower middle íncome Middle income U"=r middle income Latinamerica & Caribbean Mexico Secondarv schoo/ 2 Lowincome Lower middle income Middle income Unner middle income Latinamerica & Caribbean Mexico Ten;ia,v school 2 Lowincome Lower middle income Middle income UDDer middle income Latinamerica & Caribbean Mexico 1990 Female .. .. 94 94 90 90 91 86 98 93 87 100 42 58 58 58 .. 53 13 7 10 11 16 17 26 49 50 59 .. 54 3 9 10 15 13 13 Male 82 .. .. 2000 Female 2001 Male Female 71 .. .. 92 82 93 93 92 72 93 92 94 .. .. 100 99 100 47 36 51 41 72 73 78 81 72 70 71 81 87 75 .. 78 83 73 .. 82 89 78 11 7 12 8 .. .. 29 20 21 .. .. .. .. 36 25 20 31 22 22 Literature on sex discrimination in education access in Mexico is limited.6 Lopez (2004) used a probit model to analyze determinants of secondary schooling enrolhnent in Mexico, and, in contrast to the aforementioned data about education enrolhnent, found that being a woman reduces the probability to enroll in secondary school, and this effect is even higher for rural than urban area. A possible explanation ofthese findings could be that she used data from ENIGH 1984, 1989, 1992 and 1994, thereby reflecting previous information. Parker and Pederzini (2001) found that although there seems to be no difference in primary school enrolhnent and overall years of education between male and female children, a lower proportion of women attend secondary or tertiary school. 93 92 96 99 .. 71 Revista Perspectivar Sociales/ Social Perspectives primavera/spring 2007. Yo/.9, Num. / / Table 2. School Enrolhnent for Secondary and High School by Sex in Mexico 1 .. Secondary Schoo/ Male Female HinhSchoo/ Male Female .. .. 39 26 21 Source: World Development Indicators 2005, the World Bank 1 Net emollment ratio. The ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the level of education shown. 2000 2001 2002 2003 2004 61 61 62 63 64 64 65 66 65 67 46 48 48 51 50 53 53 55 53 57 Source: Calculations based on data from Información Estadística. Instituto Nacional de Geografía, Estadística e Informática, INEGI and Proyecciones de la Población de México 2000-2050. Consejo Nacional de Población, CONAPO (2003) 1 Toe ratio of the number of children of official school age ( as defined by the national education system) who are enrolled in school to the population ofthe corresponding official school age. 2 Gross enrollment ratio. Toe ratio of the number of children of official school age (as defined by the national education system) who are enrolled in school to the population of the corresponding official school age. Recently, secondary and high school enrollment in Mexico has been higher for female than for men. Toe gap is particularly higher for high school than for secondary school. This observed gap has increased between 2001 and 2004 as shown in table 2. Data We use data from the Encuesta Nacional de Ingreso y Gasto (ENIGH) for the waves 1992, 1998 and 2004. The ENIGH is a national income Th . ere IS, on the other hand, plentifal literature on sex discrimination in the labor market in Mexico. See,for instance, Camero (1995), Valdez (1995), MayerandCordourier (2001), and Sariñana (2002). 6 �72 / Gender-bias in Education Opportunitiesfor PopulationAged 12-18 in Mexico: 1992-2004 expenditure survey that emerged in 1984. However, it was in 1992 when the survey started to be conducted on a regularly basis (biennially). This database is statistically representative for Mexico and contains detailed information ofhouseholds for severa! measures of income and expenditure, socio-demographic characteristics of every member in the household such as age, education level, and characteristics of the job. The ENIGH also contains information of the physical characteristics of the dwelling. This national survey uses houses as sample units and households as units of observation. Toe purpose to use data for the waves 1992, 1998 and 2004 is to compare the evolution of observable and unobservable factors that might generate differences in education opportunities between boys and girls in three different points in time. We know that the best way to perform this kind of analysis is to use panel data and follow the same individuals through the time. However, we do no have this kind of data source for Mexico. Toe original data contained 50,862 individuals from 10,530 households for 1992, 48, 11Oindividuals from 10,952 households for 1998 and 91,378 individuals from 22,595 households for 2004. We decided to drop domestic workers, temporal visitors and heads of households absent. As a result, the size of the database decreases to 50,378 observations from 10,530 households for 1992, 47,581 observations from 10,952 households for 1998 and 91,450 observations from 22,595 for 2004. Additionally, from the sample of individuals we selected only children between 12 and 18 years old in order to avoid possible bias in the selection of the sample. 7 These restrictions led to a total of 7,623 children in the 1992 ENIGH aged between 12 and 18 with valid responses for all the variables employed in this research. In 1998, a total 6,871 children met the age requirements and had valid responses. Finally, in 2004, a total of 11,109 children aged between 12 and 18 and had valid responses. Before starting the descriptive analysis of the data it is necessary to 1 Women usual/y gel married earlier than men. Therefore, we decided not to include children older than 18 because we wi/1 end with a smaller number o/ women and men in the samp/e. Additionally, because o/the samefact we might end with a biased sample o/more educated women. Revisto Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vol.9, Num. J / 73 first define the variables used in this research. Table 3 below presents a complete list of the variables employed in this research and its definitions. Most ofthe definitions are very transparent and do not need further explanation with the exception of rural and poor. We consider that an individual lives in a rural area if (s)he lives in a town with a population smaller than 2,500 habitants. Additionally, a household is classi:fied as poor if it has a quarterly per capita income lower than $2,170.82 (measured in 2002 pesos) for urban areas. For rural areas, a household is classi:fied as poor if it has a quarterly per capita income lower than $1,615.75 (measured in 2002 pesos).8 Table 4 shows the descriptive statistics of the data. Toe second column presents the mean values for all the individuals in the sample, the third and fourth column show the mean values for boys and girls respectively. The last column contains the difference in means between girls and boys. Toe :figures presented in parentheses represent the standard deviation for each variable and the :figures in the squared brackets (in the last column) represent the t statistic for the difference in means. Table 3. Variable Definitions Variable Education head education Gender Poor hours worked clúldordet Rural -¡:;;e both parents 2ender Xrural 2ender X nnnr State dummies 8 Delinition Years of education of clúld Years of education of the head of the household Dummv variable ,.,,.,~1 to 1 if tbe child is hov ancl zero otherwise Dummy variable equal to I if the clúld lives in a poor household and zero otherwise We~ hows worked bv tbe child Equal to the birth order of the child divided by the total number of clúldren in the familv Dummy variable equal to 1 if the child lives in a rural area and zero otherwise Aoe oí the child Dummy variable equal to I if the two parents are present in the household and zero otherwise F,,,,,.1 to the interaction between e:eoder and rural ~ual to the interaction between 2ender and nonr We include state dummv variables as controls This is the definition employed by the Mexican Technical Committee for Measuring Poverty (Comité Técnico de Medición de la Pobreza (2002)). lt is worth mentioning that, in order to estímate this fine o/poverty, the Mexican Technical Committee for Measuring Poverty uses a different definition o/rural area. They considera household as rural if it is located in a population smaller than 15,000 habitants andfor the construction o/this variable we use the same definition. �74 '••11 / Gender-bias in Education Opportunities for Population Aged I 2-/8 in Mexico: I 992-2004 Toe data shows that girls had 0.25 more years of education than boys in 1992. Additionally, it is possible to observe a statistically significant difference of 9 hours between the number of hours worked weekly by boys and girls in that year. It is also identified a slightly higher proportion (0.03) ofboys in poverty compared to girls. Toe rest of the variables in the sample of 1992 do not present any statistically significant difference between the two groups analyzed. On the other hand, in 1998, girls had, on average, 0.20 years of education more than boys. In addition, the data for this year shows that, compared to girls, a slightly higher proportion ofboys lived in households where the head ofthe farnily is a man. Moreover, boys worked on average 1.87 hours per week more than girls. The rest of the variables for 1998 do not present any statistically significant difference between boys and girls. For 2004, the data shows that the difference in years of education between girls and boys is equal to 0.15 in favor of girls. Toe data shows that, on average, boys work 5.72 hours a week more than girls and that boys are 0.12 years older than girls in the year 2004. No other statistically significant difference between the two groups analyzed was found in the 2004 sample. Table 4. Descriptive Statistics 1992 Variable Ecilcllli111 head ewcelion headgoruler Po11e lullrswcd:ed clild order Model Specification Age bmhpareits 1998 Ewcllli111 head ecbcawm headgondeJ: hwtswllked chldorder The aim of this study is to explain the differences in schooling between female and male children within a family and, specifically, if there exists gender bias during the allocation process. In order to do this, we estimate equations of the form Bays 652 (235) (231\ Girls 6.TT <2321 5.01 (3.74) 0.86 4.99 (3.711) 0.87 5.D3 (3.711.) 0.86 (034) (034) (035) 0.39 {0.49) 1030 (19.89) 0.56 037 {0.48) 14h6 (22.36) 0.56 m211 0 .45 {0.50) 14.87 (2.03) 0.89 {031) 0 .41 {0.49) 5.59 (15..51) 0.57 (027) 0 .44 {OJIJ) Rucol Age bmhp1mts 0.45 {OJ!I) 14.83 (2.01) 0.89 {031) 14.79 (199) 0.89 (031) 7.D8 6.98 (235) (2311) 5.43 536 7.18 (232l 552 (433) (434) (431) 0.84 (1]36) 0.48 {0.50) lhl (10.08) 0.59 {028) 0.41 {0.49) 14BI (199) 0.87 0.84 ID36l 0.49 2.53 O.!tl ID3l!l 0.48 {0.51J) 0.66 (12.66) (6211) (034) 1 11 Ali 6.64 (021) Rucol Pom, 1 75 Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vol.9, Num. J / (tUD) 0.59 t1128'l 0 .42 (.tl .49) 14.82 (199) 0.87 (.tl34) 0.61) (0211) 0 .41 {0 .49) 14.78 (1 .911\ 0.86 l{J3l!l Diff,rence in Mems 015* í4J31 Ofil4 [Dhl] -0.007 [-093] 0.03 [:237] -9.06· [-20.70) 0.009 n .451 -O.DI [-118) -0.08 [-1.74] -0.005 [-0.69) 0.20• [3.55] 0.15 [1.47] -OD2• 12J01 -0.002 [-0.17] . ¡ .97• [-7.81) 0.003 [D.59] -0.02 [-1.77] -OD4 í-0.771 -0.005 [-Oñ8) 21114 (1) Ecilc111i111 head ecbcalion where Edu¡¡ is the number of years of formal education completed by child j in family i. X¡ is a vector of variables which are common to all family members (we include characteristics of the head of family such as gender and education, as well as family characteristics such as the presence of both parents in the household, rural or urban location of the household, the number of children, and whether the family is classified as poor according to their income). We also include dummy variables to control for cultural and other (regional) unobserved differences between the states of residence of the families. Z¡¡ is a vector of variables which vary across family members (such as gender, age, child birth order, and head gondeJ: Pom, hiairsvllked chldord,r Ruc,t Age bclhpue,u 7fil 7.56 (233) 6.n (233) (4ñl) OBI {039) 031 {0.46) 7.84 (17.'ITI Ofil (.tl28) 0.29 {0.45) 14.84 (1.99) 0.82 t1138'l 6 .73 (4h15) OBI (.tl39) 031 (.tl.46) 10h2 (19.33) Ofil (.tl.111.) 019 {0.45) 1490 (2.01) 0.82 (038') 7 .11 (233) 6.n (4Sl) 0.82 {039) 031 {0.46) 4.90 /14.117'1 0.63 {028'1 029 l{J.45) 14.77 () .97) 0.83 ID38'l 0.15• [233) -0.008 [-0.09) 0 .006 [093) 0 .004 [OJO) -5 .n• f-18341 .O .0009 f-0.18) .0D04 [-0.52) -0.12• [-3.45) 0.009 fl 351 �76 / Gender-bias in Education Opportwlities for Population Aged 12-/8 in Mexico: 1992-2004 hours worked, if applicable). Interactions between gender and rural/urban status and between gender and poverty status are included to investigate to which extent gender bias is influenced by each of these conditions. Toe error term is assumed to have two components: one common to ali children within a family, 8¡, and another which varíes independently across siblings, uif. First, following Parish and Willis (1993), the models are estimated applying heteroskedasticity-robust methods (Eicker 1967, Huber 1967, and White 1980). That is, we deal with the issue that errors in the equations are not independent because of the common unmeasured family effect, O¡, by estímating robust standard errors. 9 Unobservable preferences, however, may influence both the family characteristics and the allocation of resources to cbildren. 10 Toe instrumental variable approach normally used to solve _this problem is not feasible in this case because all of the exogenous variables are contained in the model, leaving no instruments available to identify the family effect. Instead, we could estímate fixed- and random-effect models that control for the possible correlation between the regressors and the disturbance. A limitation of the :fixed-effect model is that we cannot estímate fJ1 , the coefficients of the variables common to ali siblings. Additionally, as Griliches (1979) emphasizes, the within estimators are not necessarily closer to the "true" estímators because differentiating may exacerbate the effects ofother potential econometric problems such as measurement errors in explanatory variables or endogeneity involving the individual error component. 0n the other band, one can think ofthe unobserved effect to be uncorrelated with all explanatory variables, whether these variables are fixed 9 We do not know whether robust standard errors wi/1 be larger than usual standard errors ahead oftime. However, as an empirica/ matter, the robust standard errors are oftenfound to be larger than the usual standard errors (Wooldridge 2003, p. 261). ° 1 For e:xamp/e, high fertility families may choose to invest less in the education of each child, leading to a negative correlation between di and the number of siblings, which, in turn, leads to correlations with related variables such as relative birth order. Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vo/.9, Num. J / 77 across a family or not. In this sense, a child 's education is explained by his own characteristics as well as bis family ' s background, but there is an unobserved effect that varies randomly within and across families. Thus, we can include in a child' s education equation a variable such as head of family 's education even ifit does not change across siblings. But we are assuming that head of family' s education is uncorrelated with the unobserved effect, which contains other family and child characteristics (see Wooldridge 2003, pp. 469-71). Therefore, we estímate also equation 1 assuming random e:ffects with the following speci:fication in the error terms µ.¡i ~ N(O, a2,.) o¡~N(O, <ra) (2) (3) E(µu o;)= E((o¡ ok) = O(i ;,k) (4) E(µij JJ.;s) = E(µ;i /.4.i) = E(µ;i /.4.s) (i ;,k; j ~) (5) Cov (X;, o;) = Cov (Z¡i, o;) = O (6) Notice that in the random effect model /31 represents the mean value of ali the intersections and 8¡ represents the (random) deviation ~om the mean value of the individual intersection. However, 8¡ is not directly obse~able and for that reason the error wij (equal to µ¡j + 8i) is heteros~edastik ( cr 2w = cr 2µ +cr 20) and, therefore, is not appropriate to use Ordinary Least Squares (OLS) to estímate this equation. Results Table 5 reports the coefficient estimates and standard errors for the two models considered (OLS-Robust and ML-Random E:ffects) for the years 1992, 1998 and 2004. Each year was estímated independently. Toe OLS-Robust model presents less signi:ficant coefficients than the ML-RE ~odel. However, both models report similar results. As noted, we include m our regressions children between 12 and 18 years old having at least one parent at home. In order to investigate whether families discriminate am?ng the education given to their daughters and sons, depending on their rural or urban status and their economic (poor or non-poor) condi- �78 / Geruler-bias in M=tion Opponunitiesfor Population Aged 12'-18 in Mexico: 1992-2004 tion, we include the.interaction bétween the variables gender and rural and gender and poor. With the inclusion ofthese interaction variables, the coefficient ofthe variable gender inquires only into the existence of gender discrimination on education within non-poor urban families (i.e. poor=0 and rural=O). A negative sign indicates that, after controlling by other individual and family characteristics, non-poor urban boys expect to achieve less years of schooling than non-poor urban girls. In our regressions, the coefficient for the variable gender is negative and significant in all cases. In 1992, the OLS-Robustmodel reports that, within non-poor urban families, boys have in average 0.3564 years of schooling less than girls. In 1998, this number decreased to 0.3003 years and by 2004, education discrimination against boys within nonpoor urban families decreased to only 0.1654 years. Toe ML-RE model reports slightly smaller coefficient estimates than the OLS-Robust model. In 1992, non-poor urban boys had 0.3382 years of schooling less than non-poor urban girls; in 1998, this number decreased to 0.2646 years; and, by 2004, this estímate decreased to 0.1458 years. 1 1 Toe sum of the coefficients of the variables gender and the interaction variable gender x rural allow us to inquire whether non-poor rural families (i.e. poor=0 and rural=1) discriminate among the education given to their female and male children (the sum of the coefficients and standard errors are reported in table 6). Toe OLS-Robust model estimates that, in 1992, non-poor rural boys obtained 0.3847 years of schooling less than non-poor rural girls. However, in 1998, gender discrimination on education within non-poor rural families became statistically not significant. Toe ML-RE model reports very similarresults. In 2004, the ML-RE model reports, with statistical significance, that non-poor rural boys get 0.2505 years of schooling less than non-poor rural girls. In general, we can affirm statistically that non-poor families, especially in the urban areas, discriminate against their male children on the education provided. Fortunately, such education discrimination against male children seems to have been decreasing. Gender discrimination on education within poor families can be Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Vol.9, Num. J / 79 estimated by adding the coefficients of gender and the interaction between poor and gender for the case of urban families and by adding the coe:fficients of gender, the interaction between poor and gender, and the interaction between rural and gender for the case of rural families (table 6). It is generally believed that girls within poor families, especially in the rural areas, are relatively more discriminated against. Por example, Oportunidades (formerly Progresa), the Mexican government assistance program, is intended to alleviate discrimination against girls in the poor families by offering larger monetary transfers to families with girls attending school. However, contrary to our findings for non-poor families, we did not find enough evidence to claim that poor families, both in rural and urban areas, discriminate on the education given to their male or female children in any of the years of the period under study. It is noteworthy that discrimination against girls is not even evident in 1992, before Oportunidades was created. Toe lack of evidence supporting gender bias against female children on education suggests the need for a review of assistance programs favoring the investment in human capital for girls and their impact on a possible education gender gap. Children in the rural areas used to achieve less years of schooling than children in the urban areas. However, in the 2004 regressions, the coefficient of the variable rural (for girls) and the sum of the coe:fficients of the variables rural and gender x rural (for boys) became statistically equal to zero (except for girls in the random e:ffects regression). In 2004, rural and urban children get, on average, the same years of schooling (table 6). Poor children complete less years of schooling than non-poor child.ren. In both models and in ali years poor girls are about 0.7 years less educated than the non-poor ones (reading the coe:fficient of the variable poor). Similarly, poor boys are about 0.45 years less educated than the non-poor ones (reading the sum ofthe coefficients of the variables poor and gender x poor) (table 6). Toe education level and the gender ofthe head of the household are also related to the education attainment of children. Female and more educated head ofhouseholds are more likely to have children with higher levels of schooling. Children having both parents at home or having �80 / Gender-bias in Educa/ion Opportunitiesfor Population Aged 12-18 in Mexico: 7992-2004 older brothers or sisters present have also higher levels of schooling. Parish and Willis (1993) found this last result in their study for Taiwan and described that family credit constraints, when all children are young, force the older ones to leave school and help with the family income. However, we found that, the larger the family size the fewer years of education a child will have. Toe coefficients of the variable age can not be interpreted in this regression but it was introduced to control for the fact that older children have more years of schooling. The variable hours worked was introduced to the model to check whether children leave school to work. We found a negative and significant relation between hours worked and schooling but the coefficient estímate of the gender variable was only modestly modified when we introduced hours worked to the model. Finally, we included 31 state dummies to control for cultural differences among the regions of the country. Concluding Remarks There is considerable evidence that resources are not allocated randomly within households, and that resources are unequally distributed within the family in many developing countries. Such an unequal distribution of goods usually takes the form of a bias against females. For example, girls lag markedly behind boys in schooling in many developing countries even though this gender gap has been declining in recent years. For the case ofMexico, it is generally believed that girls-more specifically poor rural girls- are educationally discriminated within their families. It is also claimed that 15 of every 100 parents do not invest on the education of their daughters because they think girls will get married and, therefore, investing in their education wil1 be a waste of money. Furthermore, government efforts to abate poverty have been recently focused on decreasing the "assumed" discrimination against female children. Toe government assistance program Oportunidades (formerly Progresa) gives monetary transfers to poor families conditioned on having their children attending school and health clinics. Intended to reduce such "assumed" discrimination against girls, transfers are larger for girls than for boys. We did not find enough evidence to support such believes. Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vo/.9, Nwn. / / 81 Using an OLS-Robust model and a ML-Random Effects model for the years 1992, 1998 and 2004, we did not find enough statistical evidence to support the idea that poor families, nether in rural nor in urban areas, provide more education to their 12 to 18 years old sons or daughters. In fact, contrary to the general belief, we found that non-poor families, as established by the Mexican Technical Committee for Measuring Poverty (2002), invest more in the education oftheir daughters, especially in the urban areas. Fortunately, these education differences have been decreasing over the years. We also found that female head of households are more likely to have children with higher levels of schooling and that children having both parents at home or having older brothers or sisters present higher levels of educational attainment. �,. J 00 Table 5. OLS Robust and Random Effect Models: 1992, 1998, 2004 1992 1 OLS-Robust Std. Err. Coef. 0.0773 •• -0.3564 gender 0.1252 .. -0.7123 rural 0,1339 gender x rural -0.0282 -0.0853 0.1353 head gender head education• 0.1334 0.01000.0574 0.0319 • chlld order 0.0250 .. -0.1223 no. of chlldren 0.1118 .. poor" -0.6311 0,1295 • gender x poor 0.2422 0.1479 • both parents 0.2729 0.0178 .. age 0.5861 0.0020 .. -0.0145 hours worked educationª Random Effects Std. Err. Coef. 0.0507 .. -0.3382 0.0742 •• -0.7190 -0.0752 0.0856 -0.0723 0.1093 0.0068 .. 0.1313 0.0195 •• 0.0487 -0.1180 0.0154 -0.6725 0.0702 0.0803 .. 0.3255 0,1165 •• 0.2383 0.0106 .. 0.5751 0.0011 •• -0.0121 N 1998 Random Effects OLS-Robust Std. Err. Std, Err. Coef. Coef. 0.0613 .. 0.0717 •• -0.2646 -0.3003 0,0836 .. 0.1146-0.5554 -0.5807 0.0978 • 0.1569 0.1590 0.1709 0. 1067 .. -0.5509 -0.5141 0.1467 0.0068 .. 0.1483 ·0.1502 0.0082 0.0325 •• 0.0235 .. 0.0880 0.0871 0.0278 •• -0.1045 0.0177 -0.1149 0.0778 .. 0.1042 .. -0.5389 -0.4907 0.0900 .. 0.1925 0.2007 0.1344 0.1753 •• 0.11550.6566 0.6839 0.0112 .. 0.0157 .. 0.5513 0.5526 0.0026 .. 0.0048 • -0.0087 -0.0088 2004 Random Effects OLS-Robust Coef. Std. Err. Coef. Std. Err. 0.0426 •• -0.1 458 -0.1654 0.0695 0.1245 0.1294 0.0619 0.1794 -0,1047 0.0719 -0.1137 0.1608 0.0924 -0.1346 0.1396 -0.1262 0.0050 •• 0.0983 0.0997 0.0091 0.0475 0.0146 0.0196 -0.0149 0.0155 .. -0.1355 -0.1431 0.0402 0.0604 .. 0.1084 .. -0.6034 -0.6092 0.0691 .. 0.1505 • 0.2109 0.2538 0.0951 .. 0.1430 • 0.3043 0.3196 0.0234 •• 0.0089 .. 0.7166 0.7278 0.0011 .. 0.0029 .. -0.0189 -0.0219 ..__ ~ ~ ";, o- ~- ¡;· ~ ª- .;;· " f i ~: ~ ... ;t, ~ etate dummies IS" 1 6· 0.3651 .. -1.8711 constan! slgma_u slgma_e rho R•2 .. 0.46 79.57 7,623 F or LR-Chl2 observations rou -1 .7145 1.0251 1.4019 0.3484 0.2719 •• 0.0261 0.0162 0.0147 4426 7,623 4,119 -1 .1291 0.3144 .. 0.40 72.10 6,871 - -1.1560 1.0583 1.4965 0.3334 - 0.2971 .. 0.0299 0.0190 0.0161 3420 6,871 4 040 .. -2.7095 0.4575 •• -2.5970 1.1065 1.3991 0.3848 - 0.43 75.54 11,109 0.2276 .. 0.0210 0.01 42 0.0120 6930 11,109 7,176 .. " :,.. 1.... ';-' o;; ¡;· ~ ~- .... 'O ;e .:.., a) years of education; b) less than 1,615 pesos of 2002 per capita for rural households and less than 2,170 pesos for urban households ~ ~ *") 95% significant; *) 90% significant. The sarnple includes boys and girls between 12 and 18 years old living with at least one of their parents. Source: own estimations with data from ENIGH 1992, 1998 and 2004. ~ e;· s fr Table 6. OLS Robust and Random Effect Models: Coefficient Interactions 1992 educatlon• non-poor urban famllles non-poor rural familias poor urban familias poor rural familias 1 OLS-Robust Coef. Std. Err. 1998 Random Effects Coef. Std. Err. OLS-Robust Coef. Std. Err. 2004 Random Effects Coef. Std. Err. OLS-Robust Coef. Std. Err. Random Effects Coef. Std. Err. -0.3564 0.0773 •• -0.3382 0.0507 •• -0.3003 0.0717 •• -0.2646 0.0613 •• -0.1654 0.0695 •• -0.1458 0.0426 •• -0.3847 0.1283 •• -0.4134 0.0839 •• -0.1294 0.1669 -0.1056 0.1024 -0.2791 0.1640 • -0.2505 0.0687 .. -0.1142 0.1192 -0.0127 0.0726 -0.0995 0.1173 -0.0721 0.0755 0.0883 0.1408 0.0651 0.0633 -0.1424 0.1243 -0.0879 0.0794 0.0713 0.1269 0.0869 0.0798 -0.0254 0.1485 -0.0396 0.0691 -0.7123 0.1252 •• -0.7190 0.0742 .. -0.5807 0.1146 •• -0.5554 0.0836 •• 0.1794 0.1245 0.1294 0.0619 • -0.4098 0.1227 .. -0.3964 0.0812 .. 0.0657 0.1506 0.0248 0.0603 rural vs. urban (glr1s) rural vs. urban (boys) poor vs. non-poor (glrls) -0,7405 0.1109 •• -0.7942 0,0728 •• -0.6311 0.1118 .. -0.6725 0.0702 .. -0.4907 0. 1042 •• -0.5389 0.0778 •• -0,6092 0.1084 .. -0.6034 0.0604 .. poor vs. non-poor (boys) -0.3889 0.1095 .. -0.3470 0.0700 •• -0.2900 0.1192 .. -0.3464 0.0753 •• -0.3555 0.1303 •• -0.3925 0.0591 •• [..__ ~ ~ li· l... ~ ¡¡. ~ 8:-, ~ _e; -f ..__ a) years of education **) 95% significant; *) 90% significant. The sample includes boys and girls betwee n 12 and 18 years old living with at least one of their parents. Source: own estimations with data from ENIGH 1992, 1998 and 2004. 00 t.,l �84 / Gender-bias in Education Opportunities for Population Aged 12-18 in Mexico: 1992-2004 Refereoces Alderman,H.; J. R. Behrman; D. R. Ross, andR. Sabot(l996). "Decomposing the Gender Gap in Cognitive Skills in a Poor Rural Economy," Journal ofHuman Resources 31 (1):229-54. Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vo/.9, Num. 1 / 85 Población de México, de las Entidades Federativas, de los Municipios y de las Localidades.[Online] <http://www.conapo.gob.mx/OOcifras/5. htm> [September 2006] Deaton, A. S. (1987). "Toe Allocatioo of Goods Within the Household: Adults, Children, and Gender." Living Standards Measurement Study, Working Paper No. 39, World Bank, Washington, DC. Bardhan, P. K. (1984). "On Life and Death Questions: Poverty and Child Mortality in Rural India," In Lan<J, Labour and Rural Poverty: Essays in Deve/opment Economics, ed. P.K. Bardhan. Columbia University Press: NewYork, NY. Deaton, A. S. (1989). ''Looking for Boy-Girl Discrimination in Household Expenditure Data." World Bank Economic Review 3 ( ): 1-15. Berhman, J. R. (1988). "Intrahousehold Allocation ofNutrients in Rural India: Are Boys Favored? Do Parents Exhibit Inequality Aversion?," Oxford Economic Papers 45 ( ):32-54. Deolalikar, A. B. (1993). "Gender Differences in the Returns to Schooling and Schooling Enrollment Rates in Indonesia." Journal ofHuman Resources 28 (4) 899-932. Behrman, J. R. (1993). "Investing in Female Education for Development," Journal ofEducational Planning and Administration 7(4):393412. Eicker, F. (1967). "Limit Theorems for Regressions with Unequal and Dependent Errors." Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability 1(1): 59-82 Behrman, J. R., R. Pollak and P. Taubman (1982). "Parental Preferences and Provision for Progeny," Journal ofPolitical Economy 90 : 52-73. Griliches, Z. (1979). "Sibling Models and Data in Economics: Beginnings ofa Survey." Journal ofPolitical Economy 87(5): S37-S65. Berhman, J.R.; P. Segupta and P. Todd, (2001). "Progressing Through Progresa: An ImpactAssessment of a School Subsidy Experiment," Peno Institute for Economic Research Working Paper O1-033. University of Pennsylvania. Haddad, L. anq J. Hoddinott (1991). "Gender Aspects of Household Expenditures and Resource Allocation in the Cote D'Ivoire." Applied Economics Discussion Paper Series No. 112, Unifersity ofüxford. Camero, C. (1995). "Analisis de Segregacion y Diferencias en el Rendimiento a la Escolaridad de Acuerdo al Sexo". Un Estudio para el Mercado Laboral del Área Metropolitana de Monterrey". Undergraduate Thesis. Facultad de Economia. Universidad Autonoma de Nuevo León. Comité Técnico para la Medición de la Pobreza (2002). "Medición de la pobreza. Variantes metodológicas y estimación preliminar''. Documentos de Investigación 1. Secretaria de Desarrollo Social. México. Consejo Nacional de Población, CONAPO (2003). Proyecciones de la Haddad, L. J. Hoddinott, and H. Alderman (1994). "Intrahousehold Resource Allocation: An Overview." Policy Research Working Paper No. 1255, World Bank, Washington, DC. Harriss, B. (1990). "Toe Intrafamily Distribution of Hunger in South Asia." In The Politica/ Economy ofHunger, Entitlement and We/1-Being, ed. J. Dereze and A. K. Sen. Oxford University Press: New York, NY. Huber, P. J. (1967). "Toe Behavior of Maximum Likelihood Estimates Under Non-Standard Conditions." Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probabi/ity 1(1): 221-33. �86 / Gender-bias in Education Oppórtunities far Population Agea 12-18 in Mexico: 1992-2004 87 Instituto Nacional de Geografia, Estadística e Informática, INEGI. Información Estadística. [Online]. <http://www.inegi.gob.mx/est/contenidos/ espanol/rutinas/ept.asp?t=medul 7&c=3285> [September 2006] Quisumbing, A R. (1994). "Intergenerational Transfers in Philippine Rice Villages: Gender Differences in Traditional Inheritance Customs." Journal ofDeve/opment Economics 43 ( ): 167-95. King, E.M., and M. A Hill (1993). Women 's Education in Developing Countries: Barriers, Benefits, and Policies. Baltimore and London: Toe Johns Hopkins University Press, published far the World Bank. Rosenzweig, M. R. and T. P. Schultz (1982). "Market Opportunities, Genetic Endowments and the Intrafamily Distribution of Resources: Child Survival in Rural India." American Economic Review 72(4): 802-15. Lopez, G. (2004). "Determinantes de la Matricula en la Escuela Secundaria en Mexico". El Trimestre Economico. January-March 2004, Vol. LXXI (1), No. 281, pp. 215-245. 11111 Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vo/.9, Num. l / Mayer, D. and G. Cordourier (2001). "La Brecha Salarial y la Teoria d_e Igualdad de Oportunidades. Un Estudio de Genero para el caso Mexicano". El Trimestre Economico. January-March 2001, Vol. LXVIIl (1), No. 269, pp. 71-107. Parish, W. L. and R. Willis (1993). "Daughters, Education and Family Budgets Taiwan Experiences." Joumal ofHuman Resources 28(4):86398. Parker, S. and C. Pederzini (2001). ''Education Di:fferences by Gender in Mexico." In The Economics ofGender in Mexico: Work, Family, State and Marice!, ed. E. Katz and M. Correia. Toe World Bank, Washington D.C. Parker, S. and J. Scott (2001). "Evaluación del Programa de Educación, Salud y Alimentación (PROGRESA) a Partir de: Indicadores de Seguimiento y Evaluación 1998-2001." Mimeo. Centro de Investigación y Docencia Económicas (CIDE). Pitt, M. and M. R. Rosenzweig (1990). "Estimating the Intrahousehold Incidence oflllness: Child Health and Gender lnequality in the Allocation ofTime." Jntemational Economic Review 31 ( ):969-89. Pitt, M., M. R. Rosenzweig and N. Hassan (1990). ''Productivity, Healtb and Inequality in the Intrahousehold Distribution ofFood inLow-Income Countries." American Economic Review 80 ( ): 1139-56. Sariñana, J. E. (2002). "Rendimiento de la Escolaridad en Mexico: Una Aplicacion del Metodo de Variables Instrumentales para 1998". Gaceta de Economía. ITAM. Spring 2002, Year 7, No. 14, pp. 85-125. Schultz, T. P. (1993). "lnvestments in the Schooling and Health ofMen and Women." Journal ofHuman Resources 28 ( ):694-734. Sen,A. K and S. Sengupta (1983). "Malnutrition ofRural Children and the Sex Bias." Economic and Po/itical Weekly 18 ( ): 855-64. Sen,A. K. (1984). "FamilyandFood: Sex Bias inPoverty."InResources, Values and Development, ed. A. K. Sen, Basil Blackwell: Oxford. Svedberg, P. (1990). "Undemutrition in Sub-Saharan Africa: Is There a Gender-Bias?" Journal ofDevelopment Studies 26 ( ): 469-86. Thomas, D. ( 1990). "lntra-Household ResourceAllocation: An Inferential Approach." Journal ofHuman Resources 25 () 635-63. Valdez, F. J. (1995). "Condiciones del Mercado Laboral para Hombres Y Mujeres en el Area Metropolitana de Monterrey, El Asunto de la Discriminación Salarial por Sexo". Undergraduate Thesis. Facultad de Economía, Universidad Autónoma de Nuevo León. White, M. (1980). "A Heteroskedasticity-Consistent Covariance Matrix Estimator anda Direct Test for Heteroskedasticity." Econometrica 48(4): 817-30. �88 / Gender-bias in Education Opportunitiesfor PopulationAged 12-18 in Mexico: 1992-2004 Wooldridge, J. M. (2003). Introductory Econometrics: A Modern Approach. Thomson-Southwestem. Revista Perspectivas Sociales/ Social Perspectives primaveralspring 2007. l'o/.9, Num. / / Pág. 89-175 89 Toe Informal Sector in Mex.ico: Characteristics and Dynamics Eduardo Rodriguez-Oreggia* World Bank (2005). World Development Indicators. Abstract This work analyzes the dynamics ofthe informal labor market in Mexico using information from the National Survey of Urban Employment (ENEU) and the urban section of the Quarterly National Employment Survey (ENET). In the first part, it compares three periods: 1990-199 l, 1995-1996 and 2003-2004 to study the changes in eight of the labor market categories using transition matrices. The categories include dependent informal, dependent formal, employer, self-employed, public sector, not remunerated, unemployed and, inactive. In the second part, it uses a multinomial logit model with the mentioned categories as the dependent variable and it finds that age increases the probability of being in the informal sector and that education increases the probability of being employed in the formal or public sectors. Toe third part uses quantile regressions to explore the determination of salaries as function of the labor market categories. It is found that the categories employer and public sector have the highest returns. Keywords Informal sector, Transition matrices, Quantile regressions, Labor force. Resumen Este trabajo analiza la dinámica del sector informal en México utilizando información de la Encuesta Nacional de Empleo Urbano (ENEU) y la parte urbana de la Encuesta Nacional de Empleo Trimestral (ENET) * Instituto de Investigaciones sobre Desa,w l/o Sustentable y Equidad Social, Universidad Iberoamericana, D.F Pro/. Paseo de la Reforma 880, Lomas de Santa Fe, C.P. 01219 México, D.F Te/: 59504339 59504000 x7679 Fax: 59504195 Email: eduardo.rodriguez@uia.mx T'he author wishes to thank financia/ support from the Secretariat ofLabor and Social Security. through the Labor Research Grants. ISSN 1405- 1133 C 2007 Universidad Autónoma de Nuevo León, Univemty ofTexas ofAustin, University ofTexas ofArlington, University ofTennessee, Universidad Juárez del Estado de Durango, Universidad de Colima �90 / The Jnformal Sector in Mexico: Characterislics and Dynamics comparando primero tres periodos: 1990-1991, 1995-1996 y 2003-2004 para estudiar mediante el uso de matrices de transición los cambios en ocho categorías del mercado de trabajo: informal dependiente, formal dependiente, empleador, auto empleado, sector público, sin pago, desempleado e inactivo. En la segunda parte se utiliza el modelo multinomial logit, usando como variable dependiente las categorías mencionadas de la fuerza de trabajo, encontrando que la edad incrementa la probabilidad de estar en el sector informal y que la educación incrementa la probabilidad de estar en el sector formal o en el sector público. En la tercera parte se utilizan regresiones cuantílicas para estudiar la determinación de los salarios como función de las categorías en la fuerza de trabajo encontrándose que las categorías de empleador y sector público tienen los más altos retornos. Palabras clave Sector informal, Matrices de transición, Regresiones cuantílicas, Fuerza de trabajo. Revista Perspectiva< Sociales I Social Perspectives primavera/spring 2007. lf>l.9. Num. J I 91 pared to those in the formal sector, and to know the causes that pull or push workers to enter the informal sector. It is well known that many individuals move to the informal sector due to the lack of opportunities in the formal sector. There are structural reasons to explain the size ofthe sector, mainly associated with the fixed costs that taxation and excessive regulations represent, and inadequate and weak links that prevail between the benefits of social security and the value workers give to those benefits (Garro, Meléndez y RodríguezOreggia, 2005). An important variable highly correlated with informality is the percentage of working population not covered by mecbanisms that protect them against risks, i.e. social security. The proportion of formal to informaljobs experienced a fall since the end ofthe 80's and until the mid 90's, followed by a slight recovery in the second half ofthe 90's. The crisis periods of 1994-95 and first halfofthis decade, have increased both informality and unemployment, as it can be observed in Figure 1.1 Figure 1.1 Introduction lnformality in Mexico 8 According to data from the Encuesta Nacional de Empleo (National Employment Survey), in the last few years about 60% ofMexican working population has been working with no social security, considering it as a proxy for informal labor. This meaos that only about 18 out of 40 million people have social security as a work benefit. 7 1.2 6 5 0.8 4 0.6 3 0.4 2 ECLAC (2005) has estimated that seven out of ten jobs in Latin America are created in the informal sector, being Mexico a generator of more informal than formal jobs. Toe rate of job generation occurs mainly in micro business, the larger employers in the country. On the other hand, the demographic pressure on the labor market has caused a higher growth rate on the supply than on the demand side of the labor market. According to INEGI, about 1Omillion workers are expected to join the market in the next years, added to the fact tbat nowadays there are about eight million inactive individuals waiting their turn to join the market. Hence, the creation of formal jobs becomes a main challenge for the economy. It is in this context that becomes essential to identify informal workers, to know their characteristics and disadvantages com- 02 o ~------------------_.,¡o 1987 1990 1993 1-- 1996 1999 Unemployment ~Formal /informal ratio Source: World Bank (2004) Recently, given no improvements in the labor market in Mexico (see World Bank report "Doing Business in 2004: Understanding Regulation"), unemployment increased but salaries <lid not decrease as ex.pected. This migbt indicate a change in the way the labor market is working, which could be associated to a drop in inflation, whicb meaos that the wages reduction mecbanism through price increases is disap- �92 / The lnfonnal Sector in Mexico: Choracteristics and Dynamics pearing. It is possible that macroeconomic shocks could be associated with unemployment in the long-run, or with an increase in informality. There are implications of importance for public policies regarding social protection and the financing of these schemes. Since this study focuses on the evaluation and examination oftrends in the informal sector and its comparative dynamics, the estimation of labor participation in the formal and informal sectors, and the differences in wages and occupations are important to identify groups that are socially vulnerable. It is helpful to focus public policy mak:ing on those who have no social security coverage or have a precarious job. According to INEGI, it has been estimated that by the next decade, more than ten million people will join the labor force, thus high grow rates are necessary in the economy, in order to be able to offer them a job. Moreover, INEGI estimates that about eight million people are currently unemployed and looking to join the labor market any moment. Toe importance of this study is based on what a hypothetical severe crisis of job generation could represent, given that the informal sector would be the most viable option to generate an income. One ofthe main challenges here is the incorporation of the informal sector to the formal one and for the generation of more formal employment in a sustained dynamic it is vital to learn the cbaracteristics and evolution of the informal sector. If we consider that in the last years the growth rate of informal jobs has been higber than the one for formal jobs, this perspective should direct efforts on poverty fighting and employment to consider workers in the informal sector. It is known that for sorne people informality represents a convenient decision, meaning flexibility anda certain status (according to Maloney, 1999), but on tbe other band, informality represents important social costs. Informality traps workers and companies in low productivity operations, and it represents several causes that can difficult the implementation of public policies. Toe informal sector represents no security bene:fits, and most regulations are evaded, such as tax pay. Toen again, a strict regulation and taxation, in a country with lax enforcement, could make informality to generate faster. Revista Perspectivas Sociales / Social Perspecti:ves primaveralspring 2007. Vol 9, Num. J / 93 Tbis study aims to examine and evaluate the trends of the informal sector, and analyze its characteristics compared to the formal sector. Probabilities of cbanging labor sector will be measured overa period of time by the use of transitional matrices between a formal and informal job. Toe probability of incorporating a new labor sector will be determined, subject to a set of socio-demographic variables, as well as the previous labor status. Wage incentives to education and job category according to income leve! of the worker will be calculated. Toe use of a dynamic view of sector movements will give a broader perspective on probabilities, progress, and fall backs that have taken place in moving from one sector to another. Additionally, the wage impact and choice of sector can obey a high heterogeneity, which will be considered through a methodology that allows the identi:fication of individuals with similar characteristics in the income distribution curve. Toe justification of this study is given by the possible public policíes that can be focused on the reduction of informality. Steps needed to achieve so require a precise diagnose of those in the informal sector and the reasons of being there, together with the identification of winners and losers and their tendencies along time. For society as a whole, the benefits of reducing informality could be more, since not only more people would have social bene:fits, but they would contribute with their taxes to the social development of the country, reducing poverty and increasing economic development. Property rights would increase, allowing reduction ofsorne costs, and access to services and credits would be granted and would take the advantages of market expansion. Labor Markets and Informality: Different Theoretical Versions The debate on how to classify and analyze the informal sector has included different perspectives. Several definitions have been coined but ' two have been widely spread: in the seventies, Tokman (1978) generated a Iiterature stream in which the informal sector is measured as a segment of the urban economy made of small businesses (less than five employees), and characterized by easy access, simple technology and a low capital-work ratio. Tbis definition has been extensively used in studies in Latin America. However, another de:finition has now been �94 / The Informal Secwr in Mexico: Characteristics and Dynamics widely considered since the mid eighties, in which info~ality is classified as those activities which avoid government regulation (De Soto, 1986; Portes and Sassen-Koon, 1987). International Labor Organization (ILO) has favored the use of business size to define informality; however, it is conceptually more appropriate to use the definition based on avoidance of regulations. B! using the size-of-business definition, the problem raises on the dete~nation of top number to define informality. Typically, five ~mploye~s _is the limit number, but there is no technical reason to establish that llllllt. There would not be a reason either to worry about size if the normative perspective shows interest in identifying workers wi~ a job w~ere regulations are avoided and as a consequence have no social secunty. '• .,, Aditionally, studies on informal workers frequently include self-employed and whereas self-employment is one ofthe main_characteristics in LatinAm.erica work conditions for self-employed and informal workers ' . usually di:ffer, as well as the regulations each one follows, so the ~alys1s on informality should bear that in mind. It should also be cons1dered that labor policies focused on a specific group end up a:ffecting the labor market as a whole. Generally, there are diverse streams ofthe literature trying to explain theoretically the existence of the informal sector in the labor market. Those streams shall be analyzed in the following subsections. Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vol.9, Num. I / 95 phasized that high labor regulations, especially in Latín Am.erica, are an incentive to create informal jobs. Informality exists because the private sector has incentives to avoid regulations, due to the lack of appropriate reinforcement of these. Toe suggestion of a dualism in the labor market was first established by Lewis (1954), who proposed a growth model focused on developing countries. There are two sectors in these economies, a capitalist one, modero and formal, anda subsistence economy, informal and traditional. Basically, workers have different wages depending on the sector of the economy where they work. Here, workers who do not find a job in the modero or formal sector have to take one in the informal one. However, when the formal sector expands, there shall be transferences to this sector from the informal one. From a Human Capital perspective, (Schultz, 1961; Becker, 1962; Mincer, 1962), the approach would indicate that in order to have a duality in the labor market, different salaries between both sectors would have to exist among workers with similar personality and labor characteristics. In this case, workers with low salaries, bad job conditions and in the traditional or informal sector are perceived as low-productive and not willing to learn the abilities needed in the modero sector ofthe economy, where wages are relatively higher and better working conditions prevail, and the chances to enter the sector depend on capabilities (Doeringer y Piore, 1971). Classic theory points out that the informal sector is a clear example of market economy: perfect, but segmented and with no links to the modern sector. Toe modern sector is considered unable to generate enough employment, so the excluded workers have to find a second best option, which is the case of the informal sector. The size of the informal sector therefore refl.ects the extent of the ine:fficiencies of the market and the reforms that have to be done, such as the increase in flexibility and efficiency in the labor market (World Bank, 1995). Most literature is based on Harris and Todaro's (1970) model. It was first used to explain the migration caused when the expected salary in urban areas evens the marginal return of agricultura! labor, which is why migration from rural areas to urban areas happens when any of the following is true: an increase in urban salaries; a reduction in urban unemployment; a reduction of agricultural productivity. By adjusting the model to the informal and formal sector, the informality can be seen as part of the market that is sent outwards, towards what is not covered by social benefits, since the remunerations are above the point of equilibrium in the formal sector. Others, (see Inter-American Development Bank, 2004), have em- From a competence theory perspective the informal sector is the non- Labor Segmentation and lnformality �96 / The Informal Sector in Mexico: Characteristics ami Dynamics regulated part ofthe economy, under a frame where similar activities are also regulated. Toe economic activity moves freely from the formal to the informal sector, and the form the later one operates mak:es it impossible to apply legal action and regulation, hence the legal framework explains the formation of the informal sector. According to Portes, Castells and Benton ( 1989), companies in the modem sector aim to reduce costs through avoiding regulations and sub hiring workers not covered by social benefits. .... Here, the informal sector is in disadvantage in a dual market. Ozorio, Alves and Graham (1995) have argued that "workers protected in the formal sector have higher average wages, vacations, pensions and legal protection at work. Opposed to those who cannot :find a job in these companies and have to take their second best alternative in the informal sector, small companies oras self-employed, getting involved in intense activities of workforce and no social benefits. Based on this, thé regulatory or union intervention, for instance, pushes the formal salaries above the Jevel, so the immigrants, the young, the unemployed, "queue" to get a formal job, so the rest of the waiting time is spent on the informal sector. Segmentation by Wage Difference In the measurement of labor segmentation between formality and informality, comparing salaries has been the key point (Rosenzweig, 1989). However, there are two disadvantages to this approach (Maloney, 1999): the are characteristics not observed by the worker that can be correlated to tbe sector chosen and wages; the value of characteristics not observed might not be reported. Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Yol.9, Num. / / 97 ences between formal and informal sector are due to the segmentation of the market. First, within a neoclassic model, it is assumed that all individuals have the same level ofinformation. Nevertheless, ifwe think that informal workers have less information, this could be the reason why lower mini.mal wages are negotiated compared with formal ones. Secondly, ifwe assume that labor regulations lead to the redistribution of rather higher salaries than lower ones, we are dealing with salary redistribution from the highest to the lowest salaries. It would be expected that the less educated workers compete for a formal job, since they benefit from the redistribution, while the most educated would prefer an informal job, and so a company that meets regulations would end up choosing more productive workers among those with less education, not among those who are more educated. Since there are non-observables that affect productivity, following this hypothesis there would be a wage premia to formality among less educated workers, but a salary to informality in those with more education. Thirdly, based on the efficiency wage models by Stiglitz (1974), and Shapiro and Stiglitz (1984), we have that employers are willing to pay salaries above the optimal market level and so induce workers to make an higher productivity effort. If we assume that a formal job is also a strategic way to increase work effort (such as in the number of working hours), we would expect to find formal workers in companies that benefit more from lower levels of shirking. Here the difference in wages between formal and informal workers is given by the difference between employers in their capacity to monitor, instead of focusing on intrinsic characteristics of workers. Segmentation by Mobility between Labor Sectors 0n the other hand, based on Rosen (1986), it was found that in theory informal wages should be higher than formal wages, in order to make up for the benefits not received. However, they could also be lower given the avoidance of tax payment. As for social security, there would be a compensatory difference to the informal sector for not having the benefits derived from social security, which are present in the formal sector. There are other elements that do not mak:e it clear that wage differ- Maloney (1999) offers an alternative perspective to segmentation based on the difference between formal and informal wages. He propases an analysis ofworkers' labor mobility, using panel data through the conditioned probability offinding a worker in a given sector (formal/informal) at the end of a period t+1 having he started in a different sector in t. He argues there are reasons to believe that informality is preferred by a segment of workers, because of the characteristics of this sector (flexibility �98 / The Informal Sector in Mexico: Characteristics and Dynamics of hours for instance), to the inference given by labor regulations, and to the relatively low productivity of the formal sector in underdeveloped countries. .,, Among other :findings, Maloney mentions that many young workers move to the formal sector after spending sorne years in the informal sector. He also :finds that a significant number ofmiddle aged workers move from the formal to the informal sector. This evidence is interpreted as proofthaty;orkers are free to choose the sector they want to work in. For young workers, Maloney interprets that the informal sector is preferred over the formal because of the bene:fits and the training, hence they later on move to a formal job. Once they acquire sufficient experience and expertise, those workers prefer to leave their formal jobs to set up their own businesses. However, Maloney groups together in the informal sector those working in business with less than five employees and those who are self-employed. Additionally, it is not possible to leam from data if choosing the sector has been a self-made decision. Other Considerations Focusing on the labor market in Mexico, we face a difficulty adjusting these theories to practice. For instance, Maloney (1999) has found that labor segmentation is not necessarily true, but a preference for one or another. Krebs and Maloney (1999) also suggest that maybe workers move to the informal sector because they are more attracted to it than to the formal sector, and not necessarily because the informal sector pays salaries above the market optimal salary and leads to the creation of informality. Garro, Meléndez and Rodríguez-Oreggia (2005) found that value ofbene:fits is observed only in large companies. In addition, Rodríguez-Oreggia (2005) found regional differences, since the value ofsocial security could be higher in areas near the border with the USA .1 Based on Tokman (1978), the high trend of informality in Latin American countries exists because those countries are new to development and process of industrialization compared to more developed coun1 Jt was Jound that in Brazil segmentation may happen in the lowest income /evels, while in the highest levels informality is chosen because of convenience (I'annuri-Pianto and Pianto, 2004). Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vol.9. Num. J / 99 tries, and besides they are marginal in the global economy. According to Tokman, this context makes it difficult to generate more jobs, having a low investment and a large number of immigrants moving from rural to urban areas, which tends to create more informal jobs. However, this analysis gives no room to government intervention in public policy making. From the perspective of globalization, Portes (1994) points out that it has hada relevant role in the growth of informality, since intemational competence takes companies to reduce labor costs in order to survive in the market by the use of outsourcing. Goldberg and Pavnick (2003) find that at studying the opening oftrade related with informality, institutions are highly significative, since companies tend to respond to opening by a reduction of formal jobs when operating in a strict labor environment. Reinforcing laws can also be an important cause of informality. Itzigsohn (2000) mentions, when comparing informal sectors in the Dominican Republic and Costa Rica, that the later one is lower due to a higher reinforcement of regulations, as a result of an important difference in the setting up and running of institutions. Auerbach, Genoni and Pagés (2005) consider it difficult to implement mandatory saving programs such as retirement programs in developing countries that do not apply enforcement, as it is the case in Latin Arnerica. Auerbach, Genoni and Pagés (2005) also found that for several Latin American countries, there are similar pattems of contribution between employed and self-employed, between individuals and household characteristics, which suggest there are demand factors that explain the decision on contribution (formality), as well as the fact that sorne workers, such as those in small companies who eam less than the minimal wage, have no room in this formal market. In this paper, the dynamic of informality regarding other labor status for individuals in Mexico is analyzed. First, probabilities of being in a sector will be calculated using transition matrices. Toen, the decisions of joining a given sector will be analyzed depending on a set of sociodemographic factors and previous labor sector. Finally, wage incentives to education and to labor activity will be measured, depending on the �]00 / The Informal Sector ín Mexíco: Characterístícs anti Dynamics Revista Perspectivas Sociales I Social Penpeclives primaveralspring 2007. Vol.9. Num. 11 income level of the worker. This section aims to set evidence on the dynamics of switcing from formal to informal jobs by using Markov transition matrices, which describe a worker's probability of moving from one job category to another for a given period. In Table 3.1 each row describes the probability of being in a given sector depending on the sector from the previous period. So, the probability of a worker who started in period (t-1) as formal (formal=O) and at the end ofthe period stays as formal would be q, while tbe probability of a worker who started the period as formal, but ended up as informal (informal=!) would be 1-q. It is important to note that the possibilities are given within each line, so the probabilities should add up along each line. Markov Transition Matrices between Labor Sectors Data In this section, we shall determine the mobility between sectors, using transition matrices (Markov's), considering the sector where the worker started and ended up at the end offive quarters using data from the National Urban Employment Survey (known in Spanish as Encuesta Nacional de Empleo Urbano or ENEU) and the urban part of the National Quarterly Employment Survey (Encuesta Nacional de Empleo Trimestral, ENET). In order to analyze the transitional dynamics between sectors, we need panel data, which will allow us to follow up the same worker in a given period of time. Through tbe Encuesta Nacional del Empleo Urbano (ENEU) and the urban part of the Encuesta Nacional de Empleo Trimestral (ENET), it is possible to follow up a twenty percent of workers along five consecutive quarters. This means we have a rotation panel, where the fifth of all household selected are substituted each quarter. While longer time periods are needed in order to know what happens in a worker's life cycle, it is possible to identify in this period those workers who move between sectors, as well as their characteristics. Labor Market Transitional Dynamics '• ... Transition matrices can be used in situations where there is a set of defined status or conditions, and there is a transition from one to another, and where being in a state depends on the previous one. They also require information on the previous state or condition. Due to the fact that elements in the matrices represent conditional probabilities, they should be read as an application of conditional probability to related events. ............. , 101 These matrices, called Markov, arrange in a matrix of ten the probabilities for a worker who started in a labor sector to be in the same or another one at the end of the period. Let us state there are only two sectors in the labor market: formal, denoted Oand the informal, named 1. A degree one Markov's model would specify that the probability for a worker to be in the current period t in a given sector would depend on the prevailing state in the previous period, t-1. The stochastic process describing the dominant sector is defined by two transitive probabilities calledp and q. Ifwe arrange probabilities between sectors, we have: Table 3.1 t Fonnal=O t-1 Fonnal=O a Informal= ! 1-o Informal= ! 1-a p Three different periods of time are used. First of all, we use from the third quarter in 1990 to the third quarter in 1991, which represents a growth and relative bonanza period in the Mexican economy. Secondly, we use from the third quarter in 1995 to the third quarter in 1996, which corresponds to a period of crisis in the domestic economy. Finally, we use from tbe third quarter in 2003 to the third quarter in 2004, a period of stability in the economy. We tried to use the panel of the first periods in the survey, but many errors were found when trying to give a sequence (i.e. in dramatic changes of age, education, gender) so we chose the use of the period mentioned above in the early nineties. Categories used derive from the labor market structure itself, which allows us to follow the Mexican employment survey. Thus, we suggest the use of the following categories: Informal Employment: if the worker is in the prívate sector and has �l 02 / The Informal Sector in Mexico: Characteristics and Dynamics no social security in his main job. Formal Employment: if the worker is in the private sector and has social security from his mainjob. Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vol.9, Num. J ¡ 103 centage of workers who moved to another sector or stayed in the same at the end of the period under study. The reason this happens is due to the nature of the employment surveys used, which ask the worker about their labor conditions on the week previous to the interview, which derives in information on the initial and final sector. Public Sector: if the worker has a job in the public sector. Panel Results 2003-2004 Employer: if the individual states he is an employer. Self-employed: if worker states he works on his own. Table 3.2 shows the general transitions between labor categories in the 2003-2004 panel. Without payment: if the worker states he is working but not receiving Table 3.2 Transition Matrix 2003-2004. Men and Women payment. Informal Em~~... .,, Unemployed: if the individual falls in the definition of open unem- ployment by INEGI. 2 Inactive: if the individual does not participate in the labor market. lofcnnal Einploymeot Fonoal Ein ploym eot Ptablic Scdor Einploy... Sclf-anplayed Wíhoat paymtnt Uoemployed Iaac:tn,·e Total Results S2.96 Formal Public Employer S.lfWiihoot Uoemployed In adive Em~~... Sector anel~ed ~mt 14.15 1.52 1.82 11.87 1.57 2.37 13.14 Total 100 9.64 76.IM 1.92 0.95 3.18 0.35 2.39 5.53 100 2.22 8.65 13.8 1 ll.9S 3.16 6.27 4 .01 J.54 87.82 1.73 1.74 1.34 0 .4 7 44.97 5.66 2.32 ! MI 31.24 S6.4J 12.32 0.30 2.05 2.61 38.66 0. 84 O. S4 1.0 1 0.73 3.70 4.S4 14.74 29.15 100 100 100 100 2.3.62 6.58 1S. 2S 2 1.67 3.45 19.13 6.22 1.36 11.47 1.24 0.36 3.08 9.24 S.91 13.86 O.Sl 2.34 Z.7S 12.97 24.51 78.35 3Z.S6 100 100 100 1.6.5 1.89 Source: Author's calculations using data ofENET 2003-2004. The question we want to answer is: Knowing that a worker starts on a period in a given labor category, what is the probability of changing to a different category or staying at the current one at the end of the period? Or, what percentage of workers remained in the same labor category or moved to a different one? Using the methodology and data previously presented, the transition matrices have been calculated using the panels that start in the third quarter and end in the sameoneofthefollowingyears 1990-1991, 1995-1996 and 2003-2004. Matrices are presented in a general way, separated by men and women, and separated by different age groups in the Appendix. The results are interpreted technically speaking as probabilities, however, it can be suggested that each cell in the chart represents a per12 year-old and above individua/s who during the week when the survey was done did not w ork any hours p er week but were looking for a job or tried to take over an activity on their own. 2 Toe reading of the data in table 3.2 shows that for this period, a worker who started as informal employment in the third quarter in 2003 hada 52.96 percentage probability ofremaining in that category. Switching to the right side in the same group, we find that the same worker who started the period in informal employment hada probability of 14.75 percent of ending up as formal employment; a 1.52 percent probability of ending up in the public sector; 1.82 percent probability of ending up as employer; 11.87 as self-employed; 1.57 without payment; 2.37 as unemployed, and 13.14 as inactive. Workers who started in formal employment hada probability of76.04 percent of ending up in the same group, followed by a 9.64 probability of moving to informal employment, or to inactive in a 5.53 percent. Toe lower probability of movement is 0.95 percentage probability ofmoving toward employer. Workers in the public sector show a higher probability of staying in the same group at the end ofthe period, with an 87.82 �104 Revista Perspectivas Sacia/es I Sacia/ Perspeclives primavera/spring 20<J7. fü/.9, Num. JI / The Informal Sector in Mexico: Characteristics and Dynamics percent, whereas probabilities of moving to di:fferent groups are really low. For example, switching to the inactive group is 3.70 percent. Those who started as employers have only a probability of 44.97 percent of ending up in the same category, followed by 31.24 percent to self-employed, 8.65 to informal employment. Self-employed individuals have a 56.43 percent probability of staying in the same group at the end ofthe period. Moving to inactive represents 14.74 percent, and informal employment 13.81 percent. On the other hand, those who started without payment have a probability of staying in the same group of 38.66, followed by probability of moving to inactive in 29 .15 percent. ...... For individuals who start as unemployed, the probability of ending up unemployed is 12.97, while the probability of moving to inactive represents 24 .51 percent, informal employment in 23 .62 percent, formal employment in 21.67 percent. Individuals who started as inactive have a 78.35 probability of staying in this group, followed by a probability of moving to informal employment in 6.58 percent or self-employed in 5.91.Generally speaking, the informal employment group and inactive seem to be the groups with more frequent movements after staying in the same category at the end of the period. By the analysis oftransition matrices by gender, types ofmovements ofmen and women can be identified in the different groups, as it is shown in tables 3.3 and 3.4. In table3.3, menshowahigherprobabilityofstaying in the initial group if they are in the public sector, with lower movements to other categories. However, for the other groups, being in the informal employment or self-employed category seem to be in general the prevalent second options after the option of staying in the same group. ~ !nfonnal Formal Poblic Emeloyn,ent 56.76 1:.m2~ ... Sector Emp loyer S.lf- Widlout eme•!!!'.!!! 2:!I!!!eol Unanp loyed loadive 16.39 1.76 2.47 13.04 1.40 2.83 5.35 100 10.53 77.84 1.71 1.28 3.86 0.2.5 2.31 2.21 100 2.94 9.77 17.22 23.75 4.25 6.68 5.09 6.25 86.73 L67 1.76 0.83 0.85 46.53 8.24 1.67 2.22 31.75 61.11 15.42 0.20 0.64 1.30 33 .75 0.92 0.51 1.30 1.61 1.90 2.44 3.98 16.67 100 100 100 100 26.39 11.96 21.19 23.46 6.67 27.4 9 6.45 3.07 12.77 1.76 0.84 5.61 12.61 5.60 18 .27 0.88 2.38 1.61 12.90 4.22 2.44 15.54 65.26 10.63 100 100 100 Employment Public Sed« Employer S.lf-anployed Widlout payment Uoemploycd Inadn-e Totol Table 3.4 Transitioo Matrix 2003-2004. Womeo lnfonnal Informal Eme~ent 46.38 Employmeol Formal Ernployment Public Sector Employer Self- -loyed Widlout pa:ymeol Uoemployed lnadive Total - Formal Public Eme!!!x!!!ent 11.92 1.10 Employa0.69 S, lf. Widlout payment ~12l'.ed 9.86 1.86 Uoemployed lnadive Total 1.59 26.60 100 7.89 12.49 2.32 O. J I 1.83 O.SS 2.57 12.04 100 1.46 2.72 8.59 7.07 2.01 4.08 2.34 2.41 88.98 2.04 1.70 1.55 0.07 ]6.73 1.70 2.59 0.69 28.57 49.25 11.03 0.4 2 9.52 4.61 40.69 0.76 0.68 0.57 0.34 5.61 15.65 31.23 34.31 100 100 100 100 19.37 5.60 10.18 18.92 2.86 11.97 5.86 1.05 10.37 0.45 0.28 0.92 4.05 0.00 2.34 3.73 13.06 1.19 1.43 38.29 5,96 10.08 80.71 100 100 100 51.32 Source: Author's caJculatioos using data ofENET 2003-2004. . It seems like in the case ofmen, movement are more often in catc·gones where the worker has no access to social security, while for \\ ,men changes at the end of the period tend towards inactivity. Transition matrices based on age group are shown in the appendix. For those under age, the most frequent movements are towards informal employment and self-employment. In the group of older workers inactivity is the most frequent group at the end of the period. ' Panel Results 1995-1996 Total Employmcnt Formal Table 3.4 shows the transitioo matrix for womeo in the 2003-2004 period. Womeo show a higher probability of staying in the initial group than meo if the initial group is public sector, without paymeot, unemployed or inactive. Public sector shows the highest probability of permaneoce, as it happens amoog meo. However, wheo switching, the first option at the eod of the period is being inactive, followed oot too close by informal employmeot. Table 3.5 shows the general transitional matrix using the 1995-1996 panel. Table 3.3 Transition Matrix 2003-2004. Men Informal 105 Source: Author's calculatioos using data of ENET 2003-2004. Toe general trend between changes ofcategory is very similar to that found in panel 2003-2004, being informal employment and inactivity the ones with the most movements. . As read in table 3.6, roen show similar trends compared to 2003-2004, smce the most probable categories to move to are informal employment �106 Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vol.9, Num. / / / 11,e Informal Sector in Mexico: Characteristics ami Dynamics and self-employment. Table 3.5 Transitioo Matrix 1995-1996. Men and Women lnfcnnal Em2~eat Informal Jlmployment Formal Jlmployment Public Se<lor Employer s.Jf•..,pJoyed Without Publi< Formal Eme~e111 Sedo!" Jlmployer Self. ~l'!!:ed Withoot e~mt Ina<tive Total 2.98 12.17 100 Unemployed 50.27 16.61 2.34 2.29 11.70 1.65 10.45 75.53 1.99 1.31 3.33 0.43 2.29 4.62 100 3.94 4.21 14.57 29.35 100 100 100 100 25.38 79.48 34.00 100 100 100 2.69 8.51 12.49 11.61 3.13 6.18 4.01 4.27 86. 15 1.88 1.90 1.20 0.79 49.01 7.47 2.41 2.22 26.52 54.96 12.16 0.23 2.24 2.54 37.57 O.SS 1.43 2.05 1.42 Inadive 23.56 6.12 Total 13.96 17.69 3.18 18.38 6.17 1.31 12.04 2.33 0.43 3.77 9.91 5.19 12.66 2.33 2.39 2.84 12.64 1.90 2.34 paymeat Unemployed Generally speaking, meo and women show a higher probability of staying in the public sector, if individuals started in that group at the beginning of the observation, and the chances of moving to another category are really low. Women have a higher probability of staying in the inactive and without payment categories compared to men. Panel Results 1990-1991 Table 3.8 shows the general traosition matrix using the 1990-1991 panel. Table 3.8 Transition Matrix 1990-1991. Men and Women Source: Author's calculations using data ofENEU 1995-1996 37.68 25.66 12.12 63.77 - Wthout 5.16 8.46 12.85 l.54 payment Unemployed bia<tive Total 19. 18 5.06 11.42 Infonnal Table 3.6 Transitioo Matrix 1995-1996. Men .... Formal Public Informal Emeloymmt Jlme~ml s.- Infonnal Employment Formal Employment Public:Sect..Employer Self•employed Wíthout payment Uncmployed Inadive Total Employer Self• emel'!!:ed Without ~ent Uncmployed lnadiw Total 17.80 2.23 3. 11 14.26 1.77 3.03 S.64 100 11.34 76.23 1.95 1.65 4.08 0.53 2.27 1.95 100 2.59 9.09 15.23 26.39 3.84 6.92 5.55 5.90 84.76 1.76 2.00 2.78 1.30 51.03 10.00 3.13 3.35 26.86 59.14 15.28 0.22 0.62 1.50 26.39 1.14 1,55 2.81 2.17 4.14 18.75 100 100 100 100 18.27 6.23 26.38 6.04 3.75 1.36 6.96 12.23 7.26 17.63 2.94 3.83 1.94 15.17 63.42 100 100 100 52.17 28.38 10.57 18.97 2.53 13.85 2.45 1.39 13.21 4.80 3.01 IU6 Source: Author's calcuJations using data ofENEU 1995-1996 For women in table 3. 7 similar trends can be read compared to results in 2003-2004. Most individuals tend to move to inactivity, followed by informal employment. Table 3.7 Transitioo Matrix 1995-1996. Women . ' ' Public: Informal Fonnal Eme~mt F.me~mt Sedor Informal Employment Formal Employmeat Public: Se<lor Employer Self-employed Wilhout payment Uncmploycd Inadive Total 2.53 Employer Self• emel<!!:ed Without ~eot 0.87 7.29 1.44 Uncmployed lnadive Total 2.89 23.39 100 47.00 14.58 8.54 74.06 2.07 0.57 l.76 0.38 2.32 10.30 100 2.80 4.73 7.74 4.80 2.29 1.35 1.34 3.52 87.79 2.70 1.74 0.48 0.19 35.81 3.08 2.08 0.89 24.32 47.71 10.72 0.25 12.84 4.34 42.72 O.SI 0.68 1.34 1.44 5.28 17.57 32.70 34.24 100 100 100 100 15.69 5.28 16.76 2.61 11.42 6.38 1.08 10.47 0.00 0.26 0.99 6. 12 4.80 8.34 1.33 2.13 3.63 11.70 1.36 1.76 42.02 82.49 53.79 100 100 100 9.61 107 Source: Autbor's calculations using data ofENEU 1995-1996 lmonnal Employment Formal Employm,nt Pnblic Sed..Jlmployer Self-anployed Formal F.me~ent Em2~"11 Public Employe, ... Self• Witbout emel<!!:ed ~ 11.09 1.62 Unemployed lnadive Total 4.47 2.60 1.77 15.11 100 4.73 1.90 6.21 0.71 1.35 9.21 100 13.66 15.36 13.24 l.54 63.43 6. 11 4.62 3.38 1.92 37.30 6.59 1.54 4.50 24.92 43.66 9.85 0.56 1.25 1.63 28.31 0.96 0.47 O.SS 0.92 9.81 6.11 16.l6 44.92 100 100 100 100 26.48 5.85 22.39 7.76 2.98 10.67 1.83 0.59 3.22 10.96 4.43 10.47 1.37 2.44 2.11 731 1.07 1.23 25.11 77.58 38.49 100 100 100 Source: Author's calculations using data of ENEU 1990-1991 From this evidence, the categories to which most individuals move to at the end of the period are informal employment or inactive. In other periods studied, it was found that informal employment is a category with a high tendency to move to, and here formal employment has a great relevance. This possibly suggests that as from the crisis in the mid nineties, the creatioo of employment has not beeo enough to cover the need for this type of job, so activities where workers have no social security have covered this lack of opportunities, generating a differeoce in the labor dynamics ever since. We can see in table 3.9 that for meo, the category informal em~loymeot represents the first movement option within a period, while mdividuals in informal employmeot and self-employment represent a second place in number of movements. Public sector has more permanence probability, even though in this case it is lower than the previous periods, after inactivity. �108 / The informal Sector in Mexiw: Characteristics ami Dynamics Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vol.9, Num. I / Table 3.9 Transition Matrix 1990-1991. Men Public Infonoal Formal Employment Employmeot Sector 27.55 4.96 Employer 3.54 Informal Employmeot Formal Employmeot 39.80 13.41 65.74 4.95 2.55 Public Sector EmployaSelf-employed Wilbout paymeot Uoemplaycd 7.09 8.41 14.29 16.00 17.06 16.11 16.30 10.00 61.23 6.30 5.28 2.00 3.04 38.70 8.54 1.00 23.08 10.28 16.64 34.62 7.69 13.92 5.25 34.00 13.22 3.08 2.25 6.19 IDadive Total Selfempl~ed represented the one with more movement. Wilbout Uoemploycd Total lnadivc ~eot 1.84 1.98 6.66 100 0.42 1.52 3.54 100 6.67 0.34 26.09 0.70 48.45 0.93 18.00 29.00 0.84 0.53 1.01 3.00 3.72 3.15 5.20 ?l.00 100 100 100 100 16.15 6.53 16.39 6.15 2.89 1.60 8.46 55.14 10.47 100 100 100 13.67 7.85 0.77 3.75 1.48 Toe formal employment category has a higher permanence percentage compared to informal employment, however, informality continues as the second option after formality for men, whereas inactivity is the category that takes this place for women. 0n the other hand, an important percentage of workers move to the informal sector as an altemative to unem.ployment, for both men and women. In the case of women, table 3. l Oshows that inactive is the category with more frequent movement, which is consistent with subsequen~ periods. After this one, formal and informal employment are the opttons with more movement in the period. Women have a higher probability of staying in the initial sector than men, when in the public sector, unemployed or inactive. This is certainly notan alarming indicator of the limited opportunities in the labor market regarding formal jobs. Notwithstanding, this transition analysis does not answer other relevant questions: what are the characteristics of workers who remain in the same or move to a different labor category?, how do these movements take place among education and age groups?, what aspects of household infl.uence this dynamics? This is analyzed in the following section, where a multivariable analysis will be applied to the labor dynamics discussed in this section. Table 3.10 Transition Matrix 1990-1991. Women Analysis of Formal vs. Informal Sector Choice Source: Author's calculatioos using data ofENEU 1990-1991 .... ..., 109 Public Formal Informal Employment Employment Sedar Informal Employmeot Formal Employmeot Public Sector Employer Self-employed Witbout paymeol Uoemployed Inadive Total Employa- Self!!'.!J!I~ Wilbout e~eot Uoemploycd lnadive Total 32.91 21.41 3.35 0.48 5.27 1.12 1.28 34.19 100 8.78 58.68 4.14 0.20 1.97 1.48 0.89 23.87 100 2.27 8.96 9.03 0.89 8.58 8.96 5.13 3.56 66.71 4.48 2.87 4.00 0.25 25.37 1.44 1.78 1.26 14.93 31.01 6.22 0.88 5.97 3.49 28.00 1. 13 0.00 0.41 0.00 18.92 31.34 46.61 55.56 100 100 100 100 13.48 4.25 6.79 14.61 4.60 12.09 7.87 2.63 8.42 0.00 0.33 0.59 3.37 4.10 5.23 2.25 2.23 2.68 8.99 0.78 0.89 49.44 81.08 63.31 100 100 100 Source: Autbor's calculations using data ofENEU 1990-1991 The labor dynamics presented in the previous chapter will be analyzed based on worker characteristics (age, education, gender, marital status), characteristics within households (head of household, percentage of member ofhousehold over 65, percentage ofmember ofhousehold under 12, another member of the household has social security at work) and their labor background (labor category in the last five quarters). With this data, a multinomial probabilistic model will be used to determine the effect over choosing among the categories presented in the previous section. Results of the Analysis of Transitions Toe percentage of workers who remained in informal employment after five quarters has increased as from the early nineties, moving from 32.91 percent (39.80) forwomen(men) at47 (52.17) percentin_the 1995-1996 period, remaining in 46.38 (56.76) in2003-2004. Categone_s wh~r~ most individuals tend to move are informal employment and macttv1ty for women and formal employment and/or self-employment/employer for men. Nevertheless, before the crisis, the formal employment category Having a dependent variable that changes value according to the category in the labor market and that can be determined by the probability of being in the final sector of a period compared to the other sectors, a Multinominal Logit model shall be used (McFadden, 1974) so that: p .. g = e x',;/J+/1' ,.y.J m iex',P, �11 O / The Informal Sector in Mexico: Characteristics and Dynamics Where Pij is the probability of i to be working in sector j, ~ is the set of parameters to be estimated and x is the set of matrices of characteristics to be included in the regressions, such as socio-demograpbics of workers, among others. Multinominal logit estimators refl.ect labor decisions based on the utility an individual obtains after making a decision. These decisions are derived from the hypothesis of maximization ofutility. This unit derived from a specific category, which depends on certain characteristics. .... In our case, the dependent variable consists of labor categories used in the previous sections: informal employment, formal employment, self-employment, employer, public sector, without payment, unemployed and inactive. Dependent variables are dummies for: age group [18-25, 26-35, 36-45, 46-55 and 56-65]; schooling [no education, incomplete elementary, complete elementary, incomplete middle-school, complete middle-school, incomplete bigh-school, complete high-school, college]; bead of household, married/free union; regions [capital, north, center, south]; if another member of the household is in the informal sector. Additionally, variables regarding household are included: the proportion of individuals under 12 years old ofthe total in productive-age in household, the proportion of individuals 65 years old or more in productive-age in household. Controls are included for the labor categories where the individual was at the beginning of the period, i.e. the first period of the sample. Data Data used in this section comes from the micro-data bases ofthe Encuesta Nacional de Empleo Urbano (ENEU) and Quarterly (ENET), and their panels, as it was presented in the previous section. Results Results 2004 Table 4.1 shows the results for the Multinomial Logitusing 2004 data for individuals who remained in the panel as from the third quarter in 2003 Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. VoL9, Num. / / 111 and controls were included for their labor category for the previous year. Toe base category is formal employment, so coefficients are interpreted as the effect of the variable under study being in a given category compared to the same variable in the formal employment category. Toe table includes coefficients, standard error in parentheses and the Relative Risk Ratios, RRR, in brackets. RRR can be read as the times it is probable to :find a given category in an individual with that characteristic compared to the base labor category, which is formal employment. An individual who only di.ffers from reference (Formal Employment) in gender (man) finds it l. 13 times more probable to be in the informal employment category. This ratio is higher also in cases where the category is public sector, self-employed and unemployed. Toe ratio is lower for men in case of employer, without pay and inactive. Married individuals, ceteris paribus, have a higher probability of being in the informal employment than in the formal employment, same as public sector or unemployed. Ratio is higher for employer, self-employed, without pay and inactive, meaning there is a positive relation. �-~-f Table 4.1 Multinominal logit 2003-2004 Public sector Employer Self-employed lY,itb.9\U. pay Unemployed Inactive '~ 1235582••· (.0632848) [1 131516] - '.!32514. . (. 1059136) [1 131516] 5621431 ·-(. 1429743) [ 7925387] -078931'.! (.0738958) [1 754428] -4838133--· ( 123854) [ 6164283] 0687595 ( 1099858) [1071179] -1 25793••· ( 0658946) [ '.!842418] ;¡'%, ~ !l.. - 1788485••· ( 0668017) [ 83623'.!6] 1559969 ( 1083468) [ 8362326] 4796922•-( 1372577) [1 168823] 1959639··· (.0750395) [1 615577] 149328 ( 1297121) [1 161054] .479972••· ( 13183) [ 6188007] 2796715·•· ( 0687749) [1 322695] Informal employmcnt MM N Marned " r ., 5· f~Q Age ~ - 1118694 ( 0767866) [ 894161) 3425026"" ( 1358072) [1408468) 1 09871'.!.... ( 1286507) [3 0003) 6721591 ••· ( 1007733) [1 958461) -4116796.... ( 1474174) [ 6625365) - 2487833" ( 13633) (.7797489) - 3596465··· ( 0808787) [ 697923] !. 1324644 ( 0875955) (1 141638] 6995023··· ( 147904'.!) ['.! 012751] 1 634863 .... ( '.!332844) (5 1J8754) 1 02804""" ( 1088006) ['.! 795581) .2779048" ( 1572801) (1 320361] - 335'.!532• ( 177218) (.715157] - 0538705 ( 0920139) [ 9475548] j 46-55 0231561 ( 1051873) [1 023426] 811314••· (. 1700751) ['.! 250864] 1 656515""* (2471382) [5 '.!41016] 1 27463••· ( 1223251) [3 577379] 5617314•·· (. 1834461) [1 753706] -118619 ( 21 13611) [ 888146'.!] 498897*. . ( 072018) [1 646904] 56-65 1860573 ( 1440111) [l 204491] 67495'.!3··· ( 2324899) [l 963939] '.! 247978 .... (.'.!733977) [9 46857'.!] l 747498". . (. 1535038) [5 74022] l 329405••• (.2324432) [3 778796] -1410177 (.3159522) [ 868474] 1 69721··· ( 139493) [5458699] -.4366875 (3879232) [ 6461733) - 163562 ( '.!994869) [ 8491138) -4098096"" ( 1859243) [ 6637766) -.0518675 ( 2819455) [ 9494547) 1176196 ( 405738) [11'.!4816] -4608826"" ( 1823496) [ 6307267] 26-35 36-45 ~- l ~ ¡r Education -369732Incomplete Pnmary ( 1772711) [ 6909195] "'~- Table 4.1 {continuation2 Multinominal logit 2003-2004 Inform•l emplc,yment ¡; Public sector Employer Self-employed ~Plf'f Unemployed lna.ctive f Complete Pnmary -6790627- • ( 167TT79) [ 5070921) -337150 ( 3583721) [ 7138012) • 3348596 (2873794) [7154385] - 6300496•· · ( 1761844) [ 5325654] - 3209738 ( 2694943) [ 7254423] -4401258 ( 3934853) [ 6439554] -6818371••· ( 1718098) [ 5056871] Incomplete Secondary · 7517217··· ( 200614) [ 471554) 0200524 ( 4340471) [1 020255] -646519· ( 3780042) [ 5238662] -837'.!321••· ( 2210067) [ 4329071 ] • 8521782"" ( 3902086) [ 426485] · 005438 ( 4416985) [ 9945768] - 9113593··· ( 214427) [ 4019775] r' CQmPl,i;Smrulw!. 1101128•( 1669471) [ 3324958] • 18812 ( 35030S8) [ 8285152) .5779925•• ( 2909869) [ 5610'.?35] -.9312454•·· ( 1764938) [ 3940626] -.6616562•· ( 2737207) [ 515996] -.4393083 ( 383215) [ 644482] -1.081792*"* ( 1715066) [ 3389874] l¡· Incompleto Uppcr ~~ -1 29602••· ( 19248'.!4) [ 2736185] 329277 ( 835158) [1 389963] -65'.!9882· · ( 3592899) [ 5204881] -1 102891-• (131306) [.33191] - 6478195•• ( 2131306) [ 5231853] • 178383 ( 4089715) [ 8366'.?2] .J 036343 ( 1973039) [ 3547495] Complete Upper Secondary -1 380219-· ( 172590'.?) [ 2515235] 5831264 ( 3483814) [1 791631] - 3988183 ( 2983068) [ 6711126] -1 160119••· ( 1828231) [ 313449] • 699239S•• ( 2773919) [ 4969631] -4015135 ( 3868643) [ 6693063)[ -1 120307••· ( 1748244) 3261797) lI College d,¡ree -1 439456••· ( 1705893) [ '.?370567] 1 142926• ( 3432555) [3 135931] 1926313 ( '.!843266) [J '.!12436] -1 146978-• ( 1802991) [ 317595] -76'.?0834··· ( 2781357) [ 4666931] 1965438 { 376226'.?) [1217189) -1 027446·•· ( 1730049) [ 3579197] ~ [ !)! §: J· 8 :"' ~ ,... !O 1 Labor category previous year ::::: Informal Employment 3 160829•(.0693153) (23 59014] 1 562321••· ( 1740637) [4 769881) 2 385948-• ( 1970453) [10 86936] 2 882188-· ( 1049738) [17.85329] 2 902877-• (279157) [18 2265] 1 508603-( 1488968) [4520411) 2 343755••· ( 0916393) [10.420'.?9] Pubhc Sector 1771718••· ( 1697315) [5.880947] 6 705229•·· 1503777) [816 6653] 2 217043• ( 3248958) [9 18015] 2 289527• ( 2022247) [9 870272] 2 959607••· ( 4296024) [19 2904] '.! 183148-• (2469714) [8 874194] 2 687454••· ( 1607584) [14 69422] Employe< 24:!03•·· ( 18100'.?4) [11 24924] 2 200259••· ( 3043855) [9 027355] 5 950851-• ( '.?066178) [384 0799] 4 610838·•· ( 1680658) [100 5684) 4 575318-• ( 3670471) [97 0589] 1 140155 ( 4782778) [3 127254] 2 532994••· ( 2184972) [12 59115] - ..., �i i • .,,. Table 4 .1 {continuation2 Multinominal logit 2003-2004 Informal employment Public sector Employer Self-employed Wtlul.ll1 pay Unemployed Inactive Self-F.mployed 3 168421-· ( 1082537) [23 76993] 2 835144ººº ( 1869111) [17 03285] 45192!7··· ( l86ll79) [91 76461] 5 546004-( 1202551) [256 2116] 4.774825·( 2805938) [l 18 4896] 2 183407-• (2116457) [8 876495] 3 690996·· · ( 1186653) [40 08473] Witbout payment 3 176886°0 * ( 2197094) [23 97198] 2 610251 ••• ( 3735604) [13 60247] 4352711 ••• (3405058) [77 68882] 4 490301-• ( 2308433) [89 14828] 6 979993 - · ( 3141721) [1074 911] 1 530828··· ( 4617627) [4 622003] 4 105723º*" ( 2113926) [60 68661] Unemployed 2 174583º*" (.1378685) [8 798518] 2 389586 ••• ( 2234844) [10 90898] 1 83264:! .... (4208708) [6 250381] 2 603837-· ( 1893378) [13 5155] 517511•• (35057) [4.560857] 2 574778- · ( 1801676) [13.1284] 2756072• ( 1459951) [15 73791] Inaetivc 2 653449(.0919166) [14 20294] 2 680063°* 0 ( 1587108) [14 58601] 2 s21ss••• ( 2500473) [12 44788] 3 674983... ( 1164468) [39 44798] 4241957. . . ( 2006571) [69 5438] 2 423985··· ( 1463325) [11 29077] 4 990086•( 0896634) [146 949) • 0929447 ( 1476269) [.9112438) -166108( 0823218) [.8469548] -2 228866• ( 1885292) [ 1076S04] - 3938235··· ( 1492315) [ 6744731] -1 045635°0 * ( 0803677) [ 3S14684) - 2210S11 ( 325352) [ 8016757) • 2107779**" ( 0736071) [ 8099539] Head Yo iru:uu,,~ over 65 Yo 11Wllllm. under 12 Another household m embcr seeunty Constont • 2495684"" ( 1202173) [ 779137) -0135706 ( 3224408) [ 9865211) 2107735 ( 4914487) [1 234633) -1 143406 (7885924) [ 3187314] 3368437 ( 3589111) [1.40052] -4905436 ( 5334553) [.6122935) 4808828 ( 5152431) [1 617S02) 4947565°•• (1251027) [1 640099] 0292068 (2051505) (1 029638] 0069353 ( 2146133) [1 006959) 3008946( 1383974) [l.351067] 1378865 ( 2392527) [1 147845] 1637335 ( 265096) [1 1779] 2765628( 1330035) (1 31859] • 2489609•·· ( 0694183) [ 7796104] -0824195 ( 1149966) [ 9208856] - 6236342. . . ( 1596813) [ 535993) -.3S50128°•• ( 08292) [ 7011645] -6621545•·· ( 143132) [ 515739) - 0767638 ( 1184458) [ 9261086) -0766255 ( 0694773) [ 9262367) • 6581702*. . (:!00651\) -4 502194••· ( 4067386) -6 159899•·· ( 4246546) -2.993007- · (:!325962) 4 058183••· d4 l47722) -2,258856··· ( 4231077) • 8332486·( 2086273) Table 4.1 ( continuation} Multinominal logit 2003-2004 Informal employment Public sector Employer Sclf•employcd ~ pay Uncmploycd Inactivc -.5013145- • (.0978262) [ 6057339) .0653421 ( 1661823) [1 067524] -0909832 ( 1934615) [ 913033) -.2445535( 112079) [ 7830541) .1520279 ( 1963018) [1 164193) • 56058••· ( 1624596) [ 5708779) • 1196453 ( 1026503) [ 88723511 Center • 1364405 ( 0976821) [ 8724582) 1370532 1678939) [1 146889] . 3636801( 1922111) [1 438614) -.0122647 ( 1123815) [ 9878102) 1895284 ( 1960359) [1 208679) -.4672034 ••• ( 1643211) [ 6267526) -.025535 ( 1033873) [ 9747882) South 0193145 ( 1108338) [1 019S02] 5210719· · · ( 183623) [1 683832] 411917•• ( 2122752) [1 509709) 2739427•• ( 1255882) [1 315139] 7062672· · · ( 2096087) [2 026413] -5005121•• ( 1980618) [ 6062202) 197131* ( 1166717) [1 217904) N Log likelihood Pseudo R2 i R'¡¡ ., s· f~- ?R a. ~· t. f.s Reeion Norlh ~ t Household w.llh, 1ocial ' ir;;- 26407 -26291.748 0.4377 Note: Standard errors in parentheses, RRR in brackets. ***, ** and * denote significance at 1, 5 and 10% respectively. Base total category is formal employment. base categories: women, 18-25 years old, no education, capital. i¡· ¡r [ !l il i 1 ~- lf j· 8:--> ~ ,-.. -'º i ._ ' ... UI �116 / The Informal Sector in Mexico: Characteristics anti Dynamics Por age groups, it can be observed that there is a general trend to increase probabilities as the worker gets older, with important variations though. Age g:roups are not significant in the informal employment category. Within public sector, the [RRR] ratio increase with age groups, but tend to decrease in the older group. Por employer and self-employed categories, the ratio of age groups increases faster than other categories. In the category without payment, younger groups have a lower probability ofbeing in that sector than informal employment, but this increases with age. Por unemployed, the ratios are lower than one, which means there is a negative relation, additional to the fact that they are only significant to younger groups. Por the inactive group, the ratio is higher than one in older groups, so it is probable to fall in this group as the person ages, rather than being in the formal employment, which is the case for younger workers. Regarding education groups, we find important variations in categories and levels. Por instance, an individual in informal employment has less probabilities of being in this section having more education than in the formal employment category. Por public sector, the education levels are not significant. For tbe employer category, education is only significative in mid levels, while in the case of self-employed it is lower than one and decreasing. Por those in without payment, the lowest education levels are not significant and all of them are below one. Por the unemployed education is not significant, and for the inactive it is decreasing and lower than one. Also in table 4.1, we present the results for the category in which the individual was in the previous year. In ali cases, being in the same category the year before gives a higher probability of being in the given category. Toe second category with more probability after the one under study is usually self-employment. This indicates this category has a higher dynamism and it is easier to move to any other category from there. Within variables that correspond to households, head ofhousehold has a negative relation in ali cases, which indicates that being head of the household represents less probabilities ofbeing in any other category different from formal employment. Toe proportion of over 65 compared to those in productive age within the household is not significant in any case. RevistaPerspectivas Sociales I Social Perspectives primuvera/spring 2007. Vo/.9, Num. I / 117 For the variable that considers under 12s in the proportion ofproductive age in the household, it is significant and positive only in the cases of informal employment, self-employment and inactive, so there are more probabilities of being in those categories than in formal employment if there is a higher proportion of individuals under 12 in the household. The variable that considers if another member of the household has social security derived from work is always negative, although it is not significant in the public sector, unemployed and inactive categories. This is, an individual has less probabilities of being in the informal employment, employer, self-employment or without pay categories than as in formal employment if someone else in the household has social security at work. Regarding controls per region, the central region is only significant in increasing probabilities ofbeing employer and reduce being unemployed, while in the north it is significant in reducing the probabilities of being in informal employment, self-employed or unemployed. Toe south region has a significant influence in increasing the probabilities of working in the public sector, being employer, self-employed, without payment or inactive, while it reduces the probability ofbeing unemployed. Results 1996 Table 4.2 shows the results for the third quarter 1996, using only individuals observed in the five previous quarters. Results are very similar in trends to those obtained for the 20032004 period. The only age group that shows positive significance in all categories is the one for 56 to 65 years old, which meaos the older group has more probabilities than the younger workers of being in any category before formal employment. This trend occurs in an increasing way as , moving along age groups, except for the unemployed group, where only the oldest group is significant. Trends in the effects of education level are similar to those in 20032004. While ali education groups are significan.t and negative for the informal employment category, which meaos the more education the lower is the probability of being in the informal employment group instead of �j _;_,_.I ~ s· > ~ e- (1) ~- ..... "' -'i'"> .., º t:r' '="> (1) Q ::t. e- º º'="> = S' o oc::B"''"'B ~ = "2.. ll) ..... ll) 00 ' ~ ºººc::g".., 8 '< o :j' ~ "'Ij ~ f"2...P.g-i ~ ~ - · -o (1) (1) ::-. (1) <¡:¡ o oc.. "' '"'º'"' ¡¡; ....,ll)~g_..§e(l)P...,.ll)cn(l) o. e o. t:r' ::t. ,e - :'-' (") s· 0 .., 0p <-· e= ll) ¡:.¡ o. u e¡t.o- g ::t. (") o ::t. .., O"' e= t:,<(1) t:!" t::". ¡:;. . (") ..... (JQCI> ·uo~Ot:r'g ..... s· '<e.~ ¡:l" 0 (JQ u (1) ,. ,. Q cnQ..u "'"" r::r - · e: º c= \ . a(") t:r' !Si..... ~ ::t. o ... . (1) ol:S :::1 (") o -· ~ ll) t:, en f""'t- '-< :::1 • ll) ¡;- ~- :;, S' "O l ~ s· ~ ~· ~i:l !:l. ~~- l f ::! ¡:¡· g<¡gQ¡;.~ (t) - · ~ (1) &' ~ gi g (") <. (1) ll) ~ = ...., (1) 1 =::. (1) en - · S. ,..,. - · o.o "'pi::! ll)ªC!:1(1). [~ocnO...,. ~~ ~ § *~ ª~ ~ < . '< .., Table 4.2 Multinominal logit 2003-2004 Informo! employment f. s Public sector Employer Self-employed Without pay Unemployed Inactive Man 13403. . (.0641922) [I 143431] - 2493554• ( 1031152) [.779303] 7920747••· ( 1355725) [2 207972] 1699946( 075975) [I 185298] • 0667076 (1235706) [ 9354687] 1648315 (. 1052591) [1179194] • 9394299••· ( 0680743) [.3908506] Married - 2309129••· ( 0653545) [ 7938086] 1063348·( 1001353) [l 112194] 3496961 - • (.1238861) (1 418636] 0691782 (.0731601) [1 071627] 3100591-• (.1209743) [l 363506] .4447951• (.1139979) [.6409556] 3975055·( 066736) [l 488108) 26-35 -.0027778 ( 0724234) [ 9972261] 4219397•( 1174755) (1 524917) 1429459••· ( 186258) (4 176438) .8580667*•• ( 0921129) [2 358596) .0166911 ( 1351575) (1 016831) - 0064051 ( 1241225) [ 9936154) 0132522 ( 0775561) [1 01334) 36-45 1093551 ( 0855584) [1 115558) 8469724*. . ( 132261) [2 332574) 1 919877••· ( 1957536) [6 820121) 1.231443· · · ( 1023956) [3 42617] .416:r749· · · ( 1528015) (1 516303) 2001683 ( 1526348) [1 221608] 3546639··· ( 0898622) (1 425701] 46-55 1023999 ( 1092043) [1 107826] 8938785••· ( 1600287) [2 444593) 2 432474••· (2106202) (11 38701] 1466706••· ( 1210324) (4 334932] 1 00456•·· ( 1810748) [2 730706) 3027238 ( 1937034) (1 353541) 1 027045••· ( 1097381) [2 7928] 3379784•• ( 1504771) [140211] 1 060758•-( 221087) (2 88856] 2 656349-· ( 2445:?03) (14 24419] 1935741••· ( 1556116) (6 929176] 2 097706••· ( 21761:?4) (8 147461] 4764521• ( 2630868) (l 610351) 2 115947••· ( 1450753) [8 297438) - 523567·· · ( 1681234) -2503935 ( 327269) [ 5924037) [ 7784944] - 3548177 ( 2535207) [ 7013013) - 4062075•• ( 1756248) [ 6661719] - 1979232 ( 257252) [ 8204329] -.3906169 (.2860994) [.6766393] -.4822626·•· ( 1723505) ¡ 6173849 l Age 56-65 ir. a--... [ ? l· lI ......S· 8 :--, ~ ,-.. .'<> i-... - Education Incomplete PrimBJY 'O �r --- I ~- ' •lz N o Table 4.2 {continuation2 Multinominal logit 2003-2004 Infonnal employment Public sector Employer Self-employed \\'..il!ulli!. pay Unemployed lnactiv e -,802:?474•·· ( 1651849) [ 4483203] Complete Primary • 840885··· ( \6\5412) [ 4313286] -0243154 ( 3089737) [ 9759778] -4504748° ( 2434301) [ 6373255] - 6773404··· ( 1688155) [ 5079662] - 3942854 ( 2466515) [ 6741616] • 7773852°ºº ( 2758062) [ 4596062] Incomplete Secondary -.8996849"º " ( \847673) [ 4066978] 0449586 (3522767) [l 045985] - 6253745º º (3\39048) [ 535061] - 7960135... ( 2021691) [ 451 l'.?38] • 8448205• •• ( 326978) [ 4'.!96345] • 6067105° ( 3\76012) [ 545141'.l] .9443374•( 1966189) [ 3889372] Complete SecondNy -l.251588·( 164370\) [ 2860501] 119'.l'.!54 ( 3101343) [l 126624] -3740375 ( 252358\) [ 687951 l] -919085-· ( \73548\) [ 3988838] -5105518ºº ( 25\7194) [ 6001643] -6746635.. ( 2747439) [ 5093278] -1 086925°•• (. 1683068) [ 33725'.l] • 3078927 ( 2857861) ( 7349942] -5106979º ( 2983714) 16000767] -824'.!97•( 1882995) [ 4385432] -694674°•• ( 2597627) ( 4992372] -6009115( 2774897) ( 5483116] -9882492( 1705631) t 3722278] -.6623648° 0 ( 2736743) ( 5\56305] • 9908084º º" ( 1687471) ( 3712764] 00 Incomplete Upper Secondary -1 4'.!0595··· ( 1857385) [ 2415701] 4999741 (.3336024) (1 648679] • 0063101 ( 2879705) ( 9937097] -1 130225° ( 2050444) 131'.!9604] Complete Upper Secondary -1.424967°0 • ( 1684653) ( 2405165] 5353066° ( 3091443) (1 707972] ., 1739471 ( 2582667) [ 8403413] -1 109943•• · ( 1796176) ( 3195776] 9296877••· ( 304524) [2 533718] 2601547 ( 243364) (1 29713\] •1 069073° ( 1748991) [ 3433267] - 6847:?72••· ( 258209) [ 5042278] Informal Employment 2 926175··· ( 0669322) [18 65614] 173942°( 1543922) [5 69404] 2 197798··· ( 1714247) [9 005166] :? 745609··· ( 1027163) [15,5741] 2 S0764S••• (2457423) [12 27599] 1693734( 1409376) [5.439757] 2 330342-( 0955303) (10 28146] l 934641°0 º ( 1510443) [6 921556] 6728015-· ( 1429209) (835 4868] 2 441038... (251067) [1148496] :? 741:!9:?··· Public Sector :? 347588· ·· ( 4230985) (10 46031] 2 :?18518··· ( 23129) [9 194248] 2 863074• ( 1518417) [17 51529] -1 452148· ·· ( 1659877) [ 234067] Colleg e degree 00 ' i ~ e l ~ o ";¡· ~ ~- ~i:! "ii1 ª'a· l f ¡:¡· Labor cate¡:;ory l!revious year ( 1714905) [15 52252] Table 4.2 (continuation} Multinominal logit 2003-2004 Infonnal e1nploymen1 Public sector f.. Employer Self-employed ~ pay Unemployed Inactiv e 2 372053• (,167539) [10 71938] 2 250209- • ( 2716477) [9 489718] 5 616081 - • (, 1802121) [274 8102] 4.37283- (, 1581038) [79 26761] 4 273258(.3156596) [71 75505] 2 14915(.2964243) [8 577569] 2 495029-• (.2088918) [12 12209] 2 977643- · Self-employment (. 1094 113) (19 64147] 2897158000 ( 1819474) (18 12256] 4 524843• ( 164623) (92 28145] 5 49082· (.1195404) (242 4558] 4 332364(,2492398) [76 12401] 2 846125· · · (.1758146) [17 22091] 375444• (, 1222085) [42 7 1029] 2844849W ithout peym ent (.1960622) [17 19897] 2 348123- • ( 3599089) [10 46591] 4 086159- ( 3045688) [59 51088] 4 356376•( 2081808) (77 97401] 6 495209° 00 (.26919) (661.9623] 2 199857**" (.3373788) (9 02372] 4118135• (.1891782) (61 44454] Employer Unemployed 2 160395- · ( 1129135) (8 67456] 2 679134·•· (.1836577) (14 57246] 2.339923° ** (.2592491) (10 38044] 2716213( 1527924) (15 12295] 2 84597--• ( 3045455) (17 21826] 2995811 ( 1527488) (20 00158] 3 139553- • (. 1245092) [23 09355] 2 686553• (.1525367) (14 68098] 2 626683•-(,2132109) [13 82783] 3 684491•-- Inactive 2 54007·•· ( 0888057) (12 68055] 4.016097• (.2259405) [55 48413] 2 815168- • (, 1392017) [16 66261] 5 227348- • (.0921762) [186 2981] (.1139906) [39 82484] [ [' ~ ? 1 ¡· lI ...s· 8 :"' ~ ,-. ~ Household - 2406196°• (. 1166083) [,7861406] -2070892 ( 1396738) (,8129471] - 2497451 . . . ( 0843404) [ 7789994] -2 3 l 3558° 0 0 ( 1776901) [.0989087] -480839600 • (. 1356735) [ 6182641] -1 14435800• (.0835412) [ 3184284] - 1044743 % members over 65 ( 3004036) [ 9007979] 6022129 ( 4240661) (1.826155] - 7416206 ( 5962135) (,476341 3] -0515407 (3375906) [.949765] · 9044664· ( 5190615) [ 4047578] - 1963177 ( 4938199) [.8217511] - 2827827 ( 3004301) [ 7536836] 2839528° % memben under 12 ( 1440132) (1 32837] 0994479 ( 2297827) (110456 1] 1792917 ( 2081896) [1 19637] 114916 ( 160593) [1 121779] -4308198 ( 4132936) [.649976] 2829398 ( 2726587) (1.327025] -1403192-· ( 2222686) [ 2458111] Head i,. • 3316258°•• ( 0755677) ( 7177559] l ::::: Ñ �122 / The Informal Sector in Mexico: Characteristics and Dynamics -~" "'"'"' º"'"' r-'..,_;.,.__, - " \O ("I 00 cor- - .~.,..,.._ : _..,.., r-"'"' P\ .., o"' ,· ....:, ...... Results 1991 ,,..., C"I .. ("1 - r - - , .~- ~g~ co - r¡;;; ~ ~ ( ' I O\ ~ ~~ - ,· oco ..._;,,.,__. In categories regarding household, being head of the household reduced the probability ofbeing in other categories di:fferent from formal employment. For the ratio of over 65 to members in a productive age, no significant e:ffect can be established over probabilities, except for those without payment, where the probability of being in that category is reduced compared to formal employment. For the ratio of members of the household under 12 to those in a productive age, this is only positive and significant for the informal employment group, but negative for the inactive one. Having another member of the household with social security from work decreases the probability of being in the informal employment, employer, self-employed and without pay categories compared to formal employment. This means that if someone else in the household has social security at work, it is more probable that the individual is in the formal employment category. For regional controls, it was found that the north region reduces in a significant way the probability of being in any category compared to the formal sector. The central region reduces the probability of being in the public sector, unemployed or inactive, while the south is only significant in reducing the probability of unemployment. -r- °'.., ""co ,,.,...., O\...,.. C"l tr\ - 123 Within the categories individuals were five quarters back, the same category where they currently are has a higher effect on the probability of being in any category different from the reference, except informal employment, where the main effect comes from being self-employed before. The second most important effect usually comes from being in the self-employed group, and before that one the category "without payment". Self-employment is the category that introduces a noticeable dynamics in the other categories, having a positive significant e:ffect on the probability of being in other categories rather than formal employment. ..,,.• _,..- : Revista Perspeclivas Sociales I Social Perspeclives primaveralspring 2007. Vo/.9, Num. 11 .....,;.,__. 00 ' ' " " ' Table 4.3 shows the result for the third quarter in 1991 with individuals who were observed since the third quarter in 1990. Being a man reduces the probability of being in the public sector �j í _:;_i og.s,~g ~ '""8~oe; e; en ("I) ::l. ("I) 8.oS~o.. ¡i ]§si.g ....., "O ~ oo-o¡;. - 0 o's • ~("I)~ ::10C1S-oo = s· g- . a s ~ o- g s ~ ..... ~ l~ IS ... ;¡· l3g"Q5-:§ ("I) ~ - o. ¡g· ~º~ogS ::::1 • ....., ~ ("I) ..... -g_ s· o'=.',~ ..... = ,.. ~ er s· s ("l)l::$t:d0"0'="' ~ ("I) ("I) s ~ -g_ s s· º ei;:i. ~ "Oe (Je¡ (Je¡ ~ ~ i:;· a- s 8.&.o fl ~aº~~ ........ s ("I) ("I) §o. s· g g t:t> ......... e: en o = ("l)=-ct ("I) ("I) ~ ~ 9 s. o ~ (le¡º "O o- c:s .... 'ciO~o .:=~ Table 4.3 Multinominal logit 2003-2004 lnfonn1l employmont Public sector 8 en ("I) ~ . "l"j e; ....o 0--o i:: O- o ,.. ¡;. -o .... ~ ("I) !, Employer Self-employed Without pay Unemployed Inactiv e O 1437° .03712- l 03575••• .3013679- • 1258618 1481047 •l 198245• • • (O 0993) [O 6898] (02010) [2 8112] (.0962275) [1 351707] (, 1703257) [1134125] ( 1764722) [1159634] ( 0758059)) [ 3017232] .O 1333° ( 0797) [ 8751] O 1399 · (O 0945) [1 1502] O 38::!8. . (O 1632) {14664] 2197091•• ( 0873313) [1 245714] 1955833 ( 166544) [1 21602] -4153888( 2026015) [ 6600836] 5340878- · ( 0714524) [1 705891] 26-35 O 0061 ( 0857) [1 0064] 04297 ... (0.1083) [1 5368] 14360• .. (0 2447) [4 2039] 6523519- · ( 1099868) [1 920051] - 1100319 (. 1745224) [ 8958056] - 243319 ( 2053572) [ 7840214] - 0714288 ( 0821747) [ 9310625] 36-45 O 1077 (0 1034) [11137] O8::!S6..• (O 1242) [2 2834] 19156(0 2564) [61914] l 01::!973- • (, 1217635) [2 753777] 5358198- • (.202564) [1 708849] -6521551•• ( 310843) [ 5209219] 3243943( 0975099) [1 383193] 46-55 O 1959 (O 1278) [1 2164] 10363- · (0 1515) [2 8189] 2 !764•·· (O 2707) [91420] 1153271 00• ( 1404527) [3 16854] 1164912( 2462'759) [3 205641] - 0279133 ( 3481833) [ 972472'1] 1173337( 1194848) [3 232764] O 2810° (O 1686) [1 3245] O 9581••· (O 2104) [2 6069] 2475 1•(O 3059) [118837] 1487235- • ( 1706247) [4 424843] 2145638- · (2988511) [8 547489] -1243548 ( 4916216) [ 8830665] 2102426- • ( 1488007) [8 186007] -O 3348* 0 (O 1659) [O 7154] 04070 (O '.l977) [1 5023] -O 1926 (0 317:?) [O 8247] - 424104 00 ( 1669511) [ 6543558] - 3239114 ( 3165842) [ 7233143] • 3394435 ( 4572393) [ 71'.l1666] - 260642 ( 1615828) [ 7705567] MBJTied 9 "' ¡. ( 0798) [11541 Man 8 ("I) AJ:! i J ..... [ 1 l I s· l)Q 8 ~ ~ ,-. ~ 56-65 f ..... Education Incomplete Primary Ñ u, �j ¡.-~J .... N O'I Table 4.3 ( continuation) Multinominal logit 2003-2004 Public sector Employer Self-employed :'Nj\h9utpay Unemploy ed Inactive Complete Primary -O 5970• • (O 161J) [ 550409] O 321216 (O 28871) [1 3788] O0602 (O 3063) [1 0620] • 6245539·· · ( 1627935) [ 5355003] - 2899847 ( 3039101) [ 748275] -8930429** (45 11611) [ 409408] -4071885•· · ( 1566505) [ 6655188] Incomplete Secondruy -O 6051 ••• (O 1963) [O 5460] 798427•• (O n213) [2 2220] O 3127 (O 3726) [1 3671] • 7352079•·· ( 2108914) [ 4794058] - 1540917 ( 3971959) [ 8571934] - 2695767 ( 5049765) [ 76370:rl] -.3279083° ( 1985577) [ 720429] Complete Secondary -O 8943(0 1676) [O 4088] O 5219* (O 2915) [1 6852] O 1435 (O 3193) [1 1543] -903601•• · ( 172674) [ 4051082] -5701557" ( 3182133) [ 5654374] - 818931* ( 4484815) [ 44090:rl] -6819047••· ( 1622482) [ 505653] -O 8118••· (0 1893) [O 4440] O 6033• (O 3117) [1 8282] -O 1264 (O 3759) [O 8812] -1 078198··· ( 2100542) [ 3402079] - 2962451 ( 355739) [ 7436051] -4860207 ( 471!742) [ 6150691] -4078772• • ( 182624) [ 6650605] ~1 0112••· (0 1758) (O 3425] O8335·-· (O 29 15) [B015] 04388 (O 3265) (1 5509] -9934341º** ( 1825335) [ 3703029] -41 19593 { 3215768) [ 6623512] - 5320881 ( 4481725) [ 5S7m2J • 5769:r78••• ( 1648429) [ 5616211] -O 9115•• · (O 1721) (04019) 1 0383··· (O 2892) p 8244] O9375• (0 3096) (2 5537] - 9926435**• ( 1786203) [ 370595] • 5148065 ( 332576) [ 5976162) -4591702 ( 4443953) [ 6318077] -4272589-· ( 16633) [ 6522947] 1 3094• • · (0 1905) [3 7041] 1 455812•• · ( 1078687) [4 287965] 1 379322-• ( 2598976) [3 972206] 1 006191*( 2275091) [2 735163) ( 0981333) [3 471207] [6 6116) O 8798... (0 1396) [2 4104] O7609º"" (0 1293) [2 1402] 3 8907••· (O 1058) [48 9466) 1 1809··· (O 2149) [3 2576] 1 22121•.. ( 1436612) [3411903] 1074132•(3675579) [2 927451) ( 2830378) [3 459667) lnfonnal cmployment Incomplete Upper Sccondary Complete Upper Secondary College degree ' ;;l "' ~ e, 3 !l.. ~ s " ;¡· f~- r~ a~- l f ¡;¡· Labor cate2ory l!revious year 1 8888••• Infonnal Employment (O OT/6) Public Sector 1241172••· 1 244502• • · 1 365473•·· ( 119261 l) [3 917576] Table 4.3 (continuation) Multinominal logit 2003-2004 ""~~ Public sector Employer Self-employed Withoutpay Unemployed Inactive Employer 10782(O 1798) [2.939S] 1 5143- • (O 2082) [4 5464] 3 8164••· (O 1737) [454404] 2 540719·•· ( 1495935) [12.68879] 2 490773••· (.4332993) (12.07061] 7742863 ( 6111054) [2 169044] 1179639 ... { 2121529) [3 2532] 1 4 903••· (O 1525) [4 3944] 2 5553• (O 1673) [12 8754) 3 265903- Self-Employmcnt l S315**0 {O 1076) (4 6254] ( 1036439) [26 20375] 2 213625-• ( 2810443) [9 148825) 1 2625:rl... (.3083453) [3 534343] 1 936095•{ 1147464) [6 931627] Without paymcnt 14229• • (0.3230) [4 1492] 20173- • (O 3779) [7 5182] 2 8045--• (O 5299) [16 519] 3 019291- • (2989903) [20 47677) 4 817182• ( 315237) [123 6162) 1 505483. . ( 64 12422) [4 506329] 3 059394( 2545801) [21 31463] Unemployed 1 2604-• (O 2112) [3 5268] 14507-• (O 2897) [4 2661] 1 3722•• (O 5389) [39441] 1 721074• ( 2584943) [5 590529] 986087 { 6253818) [2 680724] 2 135449••· ( 3238812) [8 460843] 1 769354••· { 2073573) [5 867062] l 3601 - • 17000- • 1 7074- • 2 074775••· 2 471455••· 1 83 125- • 3 3284 6••· (O 0961) [3 8969) (O 1238) [5 4742] (0 2224) [5 5149] ( 1139817) [7 962755] { 2165449) (11 83967] (.2030375) [6 241684] (0819341) (27 89535] Infonnal ,mploym, nt Inactive l¡· [ ' [ ~ ? l· l .f j· "' 8 N Household ~ ~ -0.3819°•• (O 0940) [O 682S] -O 3247(O 1131) [07227] % memben over 65 -O 3654 (O 3724) [O 6938] -O 2357 (O 4187) [O 7899] % manbersunder 12 O 0443 (0 1585) (1 04S3] O2352 (O 1943) [l 2652] Head -O 0998*•• -2553925•• ( 1041497) [ 7746124] -3 9367••· ( 365084) [ 0195 112] - 6950763•·· (.2543271) [ 4990364] -1411064( 0956443) [ 2438838] (O 7856) [O 5959] 2868503 ( 3948603) [1 332225] 1331704 ( 5571117) [1 142445] 1711459 (76592) [I 186664] -1 0 1434••· ( 326732) [ 3626418] -O 4466" {O 2373) [O 6397] - 1446657 ( 1662225) [ 8653115] -1 225626• { 6546017) [ 2935738] -Ol n049 ( 4735507) [ 9829422] -1 805081••· (2539623) [ 1644611] (O 1980) [O 9049] -O 5175 J :::: Ñ -.J �J28 / The Informal Sector in Mexiro: CharacJeristics anti Dynomics Revista Perspectivas Sociales / Social Perspectives primavero/spring 2007. Vo/.9, Num. I / 129 Education levels generally reduce the probability ofbeing in a category different from formal employment, as the individual moves on in those levels. The exception is found in the public sector, where the more educated increases the probability of being in this category compared to formal employment. However, for the employer category, education levels are not significant in the highest, while for the group without payment only complete mid-school is significant, and for the case of unemployed complete elementary and mid-school are significant. As for the position of the previous year observed in 1992, the main effect in all categories comes from the category where the individual was the previous year. However, different from the previously analyzed years, the main dynamics comes from the category without payment, while the same dynamics are introduced by the self-employment category for the years 1996 y 2004. Being head of household reduces the chances of being in a category different from formal employment. Tbe variable of the ratio of individuals over 65 to those in a productive age in the household is only significant for the inactive group, where the probability of being in this sector compared to formal employment is reduced. Toe variable of the ratio of under 12 to those in a productive age is only significant for the categories without payment and inactive, where the probability of being in those sectors is reduced compared to formal employment. The variable of other members of the household with social security from work is significant for the informal employment, employer, self-employed and without payment groups, where the probability of being in those categories compared to formal employment is reduced. .... .......,....... Variable for the north region is only significative to reduce the probability of being in the informal employment, public sector and unemployed categories. Toe center variable is only significative to reduce the probability in the informal employment, public sector and unemployed categories. The southem region variable is significative to reduce the probabilities of informal employment and unemployment. General Findings �130 / The Informal Sector in Mexico: Characteristics and Dynamics Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vo/.9, Num. J / From the two analyses in this section, we can get two important conclusions. First, there seems to be a change in the labor dynamics around 1994/95. Before those years, it seems like the labor dynamics carne from the workers in the category without payment. After that period, the main source of dynamics among sectors comes from the self-employment category. are relevant to consider that could be more attractive in the categories used. Wage Premia and Income Levels In this section, we develop a quantile regression, ora regression by level of income on the wage logarithm per hour based on a number of factors. Toe use of this type of analysis by quantiles allows us to leam how wage premia vary to education, and to the other labor categories that are used as part ofthis work in regards to a specific income leve! ofthe employee. This means we can leam about the wage premia received by a worker in informal employment based on a set of characteristics, whether the individual earns a low, medium or high income. Age has an important role on the decision to join different labor categories. As we move forward in age groups, there is less probability of being in the formal employment sector generally. This has implications on the future of retirement programs. For instance, it was found that the younger workers have a higher probability ofbeing in the formal employment category, but as they grow older they change to a job that has no social security, so they probably will not meet the requirements to rate and access a retirement in the future. We can also observe the incentives that younger workers have to increase their human capital if future perspectives do not look promising. As for education levels, the higher education is, the more probable it is to find the worker in the formal employment or public sector categories. ................ In the variables regarding household, there is no big relevance as for the number of individuals over 65, although there is evidence regarding those under 12 in the household reducing the probability ofbeing in the formal sector. Nevertheless, we noticed a strong effect in cases where another member of the household has social security to have a higher probability ofbeing in a formal job, which is congruent with other studies done for Latinamerica (See for instance Auerbach, Genoni and Pagés, 2005). This indicates a possible use of the social household networks to have more information on the labor market, since there it is possible to find a high quest for sorne given employments through relatives, and results in something productive. It has been documented that those individuals who are looking for a job through relatives usually get higher job offers than those who use different sources of information (see Calvó and Ionnannis, 2005). Having analyzed all characteristics of workers within the labor dynamics, it is important to wonder ifthose wage rewards between formal and informal workers [the different labor categories used in this book] 131 Koenker and Bassett (1978) proposed this type of regressions given the fact that the Ordinary Least Squares (OLS) are based only on the measure of the conditional distribution of the dependent variable. Quantiles allow us to leam the effects of the independent variables on the conditioned distribution, together with the media (Koenker, 2005). Quantile Regression Aregression of the OLS is based on the mean of the conditional distribution of the dependent variable used in the analysis. This proxy assumes that possible differences in terms ofthe impact ofthe exogenous variables together with the conditional distribution are important. Nevertheless, this could mean insufficiencies in sorne research agendas. If exogenous variables influence parameters of the conditional distribution ofthe dependent variables in others more than the mean, then an analysis that ignores this possibility would be seriously weakened. (See Koenker and Bassett 1978). Unlike OLS, the regression models by quantiles allow a wide characterization of the conditional distributions of the dependent variable.3 3 See Abadie et al. (2002) for a recent extention on quantile regressions, considering instrumental variables. �132 / The Informal Sector in Mexico: Characteristics and Dynamics Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. 10/.9, Num. / / Given a wages equation, the regression model by quantiles could be expressed like this: In w, .. xifJs + ua When Quants(ln Wi I xi) = xifJs (1) Where x¡ is the vector of exogenous variables and fJs is the vector of the parameters. (In wlx) denotes the 8 vo. conditional quantile of In w given by x. la 8 va. Quantile regression, O< 8 <1, is defined as the solution to the problem: (2) This is expressed as such: 133 Toe dependent variable is the logarithm of the labor income per hour. Toe independent variables being included are education (complete and/or incomplete elementary school, middle-school, high-school, and college) working experience and its squared, married, head ofhousehold, male, labor occupational groups, (informal employment, governmental, employer, self-employed) as well as controls of regions (north, center and south regions). In the following subsections, we will present the results for employees that have a working income in the age range of 18 to 65 years old. We will obtain results for the income function using the quantile regression, where common errors are calculated using the bootstrapping method. We will show results for income quantiles 10, 25, 50, 75, and 90 for the mentioned age groups. (3) Results Where ps(E) is the control function defined as po(s)-9,if e .i: O or pe(e) • (8 -l)E ifs ª 0 2004 Results This problem <loes not have an explicit form, but it could be solved using linear programming methods .. Standard errors could be obtained through the bootstrap methods. (see Koenker, 2005). Table 5.1 presents the results for the urban part of the ENET 2004, third quarter of the year, in a quantile regression of the real labor income per hour. Toe results must be read as a wage premia for having a particular skill (explainable variables) in a determined place of the labor of the worker (income quantile) Toe minimum absolute deviation (MAD) estimated in pis a particular case in this setting. This is obtained selecting 9 =0.5 (regression of the median). Toe first quantile is obtained selecting 8=0.25 and so forth similarly. When there is an increase in 0 from Oto l, a sign of the whole distribution of y, conditional to x. Data and Variables Toe data is obtained from the micro database of the ENEU and the urban part of the ENET. For this particular case, we will not be using the individual panel, but rather the total observations for the third quarter in 2004, 1996, and 1991. Toe categories without payment, unemployed and inactive are not included either, since they do not mak:e an income and thus they cannot be part of an income function. �134 / The Informal Sector in Mexico: Choracteristics and Dynamics Revista Perspectivas Sociales/ Social Perspectives primavera/spring 2()()7. Vo/.9, Num. / / Table 5.1 Regression Results for Income Quantiles 2004 Quantile Variable Incomplete Primary Complete Primary Incomplete Seconda,y Complete Seconda,y Incomplete Upper Seconda,y Complete Seconda,y College Mamed íñfonnal employment ... Govemment F.mployer '\ Self-employment Head North Center South Man R2 1i-..11-.. "' ......... (.0070435) -.2199963°• (.0071431) .1892912*º (.0086998) . 133675*º (.0199555) __4973430• (.0138877) .071326••· (.0086219) . 1002312•0 (.0138612) -.0013398 (.0142959) -.2077129*.. (.0157597) .0623962*º (.007271) .25 .0204382 (.01496*) .0945392•0 (.0152091) . 1289864•.. (.0181606) . 1479025*** (.0159408) .2299767*•• (.0170138) .3000898••· (.0168284) .6920182•0 (.0167268) .0788723h• (.005846) -.1790926*º (.0056355) .2499325*'* (.0069618) .277602... (.0130429) -.2452206* .. (.0088229) .0668863••· (.0059444) .103149*" (.0104258) .0066643 (.0103409) -.1934692••· (.0112846) .0796126*** (.0055834) 0.1385 0.1422 .1 .0511846** (.0265233) .1491348•h (.0256215) . 1723762°• (.0295809) .1858011 ... (.0256208) .2530756*** (.0283632) .3318542*•• (.0262393) .6341614°• (.026138) .055976'ith In Figure 5.1 wage premia or returns to the estimated educational levels in table 5.1 are shown4 . .0302701•• (.0117865) . 1063048*º (.0115348) .1633982*º (.0152688) .1830439••· (.0114809) .2910544••· (.0135515) .3766218*** (.0117658) .8869538*º (.0134109) .0785801 ... (.0051647) -.122985'7tº (.0048342) .277089'** (.0069683) .372789º* (.0124937) -.0382427*** (.0069408) .0604478••· (.0055631) .1139433*º (.0075823) .75 .0481907*•• (.0146283) .1234635... (.0136753) .2097974••· (.0172899) .2344791••· (.0139388) .3706626*"* (.0156332) .500191••· (.014922) 1.099131... (.0161795) .0834417*** (.0059051) -.0663344*.. (.0063538) .2687495*** (.0086934) .4995041... (.0136418) .129792*** (.0074375) .0680362*** (.0063712) . 1246903'*• (.0091802) .01os1o, .0!790l9h (.007225) -.162836••· (.0088197) .096692*** (.005026) (.0093 147) -.1231185•.. (.0103351) .0975305*** (.0057426) 0.1752 0.2164 .5 135 ., .071652*** (.0199309) .1676048º* (.0191575) .2817784•.. (.024478) .3080247•.. (.0192494) .5007157•.. (.0252299) .6786236*º (.020911) 1.315767... (.022214) .105596*.. (.0083876) .005781 (.0085511) .2334153*'* (.0101951) .6688364... (.0199851) .2839413•0 (.0117071) .076841*.. (.0090529) .1262436*" (.0117201) -.00621 , ... (.0121933) -.1263096•.. (.0136491) .0712728... Premio salarial por niveles educativos 2004 20 18 16 - + -primaria inco"l)leta 14 ~maria completa 12 _._secundaria incompleta 10 --H--seetJndaria completa 8 --ll-i>'epa · ~ • t a 6 -+-P<epa completa 4 ofesional 2 o 0.1 0.25 0.5 0.75 0.9 Cuan111es (nivel) de Ingreso Note: The rates are calculated comparing the previous educational leve!. From the chart, we can observe that returns to school tend to be higher when the income level of the employee is higher. Toe returns to the professional level increase the most as the leve) of income also increases. These are followed by the high school levels both complete and incomplete, which have similar levels to one another. (.0084942) 0.2188 N= 100044. Standard errors in parentheses. Bootstrapping method is used. Significance levels 1, 5, and 10% respectively. Base categories: formal wage jobs, Capital region: female Figure 5.1 For the occupation categories, graph 5.2 shows the wage premia for occupations based in calculations made in chart 5.1, where the comparison base is the category of formal salaried position. We can observe in general that the categories improve their performance as the employee raises hís/her income level. ~ Calculated by dividing the chang¡ in the mid point of the year previous to the study m education leve/ over the change ofcoefficient obtained in the regression ofschooling and the following downwards. �136 / The Informal Sector in Mexico: Choracteristics and Dynamics Revista Perspectivas Sociales / Social Perspectives primaveralspring 2007. Vol.9, Num. I / Figure 5.2 Premio salarial por ocupación 2004 __.,_Asalanado infoonal _._Gobierno _._Patrón - Autoen1)1eado 137 in the mid-range levels, it tends to decrease slightly in the upper-levels. For the regions variable, if an individual works in the northern region he has a positive wage premia over working in the capital city area (Mexico City and Metropolitan Area). This premia could range between 10 and 12 percent as the work.ing income increases. Toe mid-region does not have a significantly different premia from the country's capital in the upper-income levels. For the southem region, the wage premia is negative at all levels, and it even decreases as the income increases. This is, for a worker in the southem region compared to a worker with similar characteristics in the Capital, the wage premia is negative and even larger if the worker is in the lower income levels. 1996 Results -- 0.1 0.25 0.5 0 .75 0.9 Cuantiles (niveles) de Ingreso Source: Data based on table 5.1 Comparison category: formal employment. Individuals working in informal employment commonly receive a negative wage premia compared to individuals work.ing in formal employment. Although their premia is increased, it is not until the highest income levels where they receive a similar premia to the ones received in formal employment. Toe self-employed individual receives a negative wage premia (in comparison to people in formal employment) until they reach the mid-range income levels, after that, their premia are comparable in size. .. ,....... ......... , For govemment-employed individuals, the wage premia are positive throughout the income curve. They tend to increase in the mid-range levels and start decreasing slightly in the upper-levels. Toe employer offers the highest premia compared to formal employment in the market. These premia are always positive throughout the working income curve and in the upper-income levels; they are around 60 percent ofthe premia. Looking back at table 5.l, we found that the married variable has a positive sign and it increases based on the income levels. Toe head of household variable is also positive in al_la levels. Men in this category have a positive wage premia over women and even though it increases Table 5.2 shows the results for the regression by income levels using the third quarter of the year in ENEU 1996 for the real labor income per hour as a dependent variable. �138 I The Jnfonnal Sector in Mexicc: CharacJeristics and Dynamics Revista Penpeclivas Sociales / Social Perspectives primaveralspring 2007. Vol 9, Nran. J / Table 5.2 Regression Results for lncome Quantiles 1996 Quantile high-school) grow the fastest as the level of income of the worker also increases, in contrast to the lower educational levels. Variable Incomplete Primary Complete Primary Incomplete Secondary Complete Secoodary Incomplete Upper Secondary Complete Upper Secoodary College Manied ..._, Informal employmmt Govemmmt Fmployer Self-employmeot Head North Ceoter South Man Worl<iog Experieoce Workiog Experieoce1 Constaot Rl ......... .1 . 1122551••• (.0193429) .1935193... (.0190364) . 2293548••· (.024222) .3188051... (.0186295) . 3733448... (.0210376) . 5161528... (.0192422) .8227605... (.020362) .0686138•. . (.0059136) -.2657218°• (.0062376) . 2805529... (.0084216) .1697128*•• (.0169189) -.3331458... (.0098688) .0837167••· (.0071418) -.049037*•• (.0087918) -.1557438••· (.009551) -.2297399".. (.Olll987) .0280003•.. (.006752Z) . 0217186*.. (.0009291) -.0003885* .. (.0000187) 1.368827*•• (.0227383) 0.1571 .25 . 0779013••· (.0155775) . 1607395••· (.0151512) . 222497... (.0171238) .2923388... (.0158812) . 3803345*. . (.017253Z) .5263087*. . (.0154924) . 933857••· (.0161877) .0866623••· (.0043049) · .2315485••· (.0048499) .3168207••· (.0061455) . 2686584••· (.012273) · .1730801*. . (.0064) . 0818733*. . (.0053745) -.0029008 (.0074481) -.1184664••· (.007536) · .2031403••· (.0077039) . 0311422... (.0055246) .0232724... (.000622) ·.0003828" .. (.0000127) 1.562938••· (.0187221) 0.1911 .5 .0767728... (.0133274) .169556••· (.013042Z) .2560976... (.0145982) .330222••· (.0131502) .4432641 ... (.0152478) .618564• .. (.013649Z) 1.151073•.. (.0147181) .0974822••· (.0044044) -.2105107*•• (.00497:?} .3018258*.. (.0058275) .3576882*.. (.0118846) -.0570363••· (.0060364) .0724723•.. (.0054951) .0271821••· (.0072134) -.1066489••· (.0073342) -.1940708• •· (.0081512) .0300351••• (.0046499) .0254699... (.0006044) -.0003789* .. (.0000118) 1.760777*•• (.0157221) 0.2402 .75 .1013566••· (.0130383) .23062••· (.0127941) .3409061••· (.014364Z) .4376044 ... (.0128849) .6009575•.. (.0170523) .8121152••· (.014391:?} 1.435186• 0 (.0139282) .1046407••· (.0061323) -.164419••· (.0058265) .233858••· (.0066923) .5038049••· (.0162642) .0521786* .. (.0077985) .0718272* .. (.0064333) .0618053• .. (.0100186) ·- 1011695••· (.0098389) -.1730474••· {-0109776) . 0202127*•• (.0058695) .0283686••· (.0007813) -.0003771••· (.0000148) 1.901385• .. {-0192232) 0.2760 .9 .1579764... (.0172386) .3145682... (.0175479) .4583621... (.0219159) .584766... (.0198589) .8054488••· (.0246721) 1.066888... (.0207088) 1.69914••· Chart 5.3 Premio salarial por niveles educativos 1996 20 18 16 14 -+- primana '""""1)1eta 12 --primana con.,re1a ..,._ secundana ~ • t a 10 ~ -.1900191*•• (.0145897) .0156223.. (.0076389) .029725 ... (.0009204) -.0003376••· (.0000189) 2.019902••· (.0263856) 0.2646 N= 117699. Standard Errors in pareotheses. Bootstrapping method is used. Sigoificaoce levels 1, 5 aod 10% respeetively. Base category: no schooling; Formal salaried position; Capital region; female. Table 5.3 shows retums to education by levels and they tend to be increasing if the income (quantile) level increases. Toe professional category has the highest retums, although they are very similar in the last two income levels. Toe highest educational levels (professionals and -eo<!l)leta - -prepa ~ e t a (.0199884) .1319679• 0 (.0077828) -.0749106••· (.0090122) .1464727*•• (.0084852) .6580552*º (.0211757) .1931023••· (.0129109) .083737*•• (.0088078) .0810747*•• (.0125451) -.1286197*•• (.0125078) 139 - - -prepa con1)1eta 6 profes,onal 2 o 0.1 02 5 05 0.75 09 Cuan111es (nivel) de Ingreso Note: Toe rates are ealculated eomparing the previous level of edueation. Chart 5.4 shows the wage premia by labor categories in 1996 and they are based on calculations in table 5.2. With the exception of government employees, the other categories have wage premia that increase as the working income curve also increases. �140 / The Informal Sector in Mexico: Characteristics oml Dynamics Revista Perspectivas Sociales I Social Perspectwes primaveralspring 2007. Vo/.9, Num. / / Graph 5.4 Premio salarial por ocupación 1996 - . -Asalariado informal ---Gobierno ~ Patrón - 141 income. Men have a positive premia compared to women, although it tends to reduce slightly in the higher income levels. The northern region gives a negative wage premia compared to the Capital if the worker is in the lower income Ievels, changing into a positive premia once the income starts growing. If the worker is in the mid-region, he will have, ceteris panous, a negative premia compared to a similar worker in Mexico City. For the southem region, the premia is negative and significant at all levels of income, although it tends to decrease slightly in the upper income levels. Autoem eado Results 1991 Table 5.3 shows the results for the quantile regression using the third quarter in 1991. .......... 0.1 0.25 0.5 0.75 0.9 Cuantiles (niveles} de ingreso Source: Data based oo table 5.2. Comparisoo category: formal salaried position. For this year, we have in chart 5.4 that individuals in informal employment have a negative premia in the labor income curve. Even though this negative premia is limited to the highest levels of income, in general it is significantly lower compared to formal employment. The self-employed workers have a negative premia compared to the formal salaried individuals until the mid-range leve1, after that, the premia turns into a positive one. For the government-employed category, chart 5.4 shows a positive premia in the working income curve, which tends to increase slightly toward the mid-range income levels and decreases approximately half toward the upper-levels. The employer category has the highest wage premia compared to formal employment just as it turned out for 2004 previously shown. In the lower income levels, the largest wage premia is given by the government sector. For the married, the premia is positive and it increases slightly as the curve for working income advances. The heads of household also have a positive wage premia although it is the same throughout the curve of Table 5.3 Regression Results for Income Quantile 1991 Cuantil Variable Incomplete Primary Complete Primary .1 .0551687** (.0267286) .1237991 ••• (.0274932) Incomplete .1855255••• Secondary (.0298567) Complete .2182557*•• Secondary {.0287367) Incomplete Upper .247231**• Secondary (.0309723) Complete Upper .396267**• Secondary (.0299785) College .6020031• • • (.0312218} Manied .0823932... (.0090953) Infonnal -.1216903*.. employment (.0082643) Govemment .1333582*** .0111906 Employer .2889954**• (.0260289) Self-employment -. 1030161••· (.0124856) Head .0441012••· {.0088653) North .0671651• • • (.0091485) .25 .0633283••· (.0146187) .1193638•• · (.0140155} .2028266*.. (.0166813) .2422491*.. (.0160813) .2853215*** (.0163914) .4617895•• · (.0 163283) .7455765*** (.0175419) .0818452• • · (.0073197) -.0924429••· (.0083655) .1563353*.. (.0089152) .4279039••· (.0211014) .0437407*•• (.0102258) .0724099••· (.0072246) .0933573••· (.0078559) .s .0556959• • · (.0202473) .1380245*.. (.0189419) .1983221... (.0224715) .2790569••· (.0204858) .3547888•• · (.0239595) .5512295••• (.0214404) .944523• •· (.0219422) .1120955••· (.0083919) -.0511854••· (.0085879) .1113794••· (.008792) .545495*.. (.0214468) .151444*.. (.0110676) .0683675••· (.0085853} .1194034**• (.0090717) .75 .0896135••· (.0202994) .1942556... (.0199053) .2702281• •· (.0230815) .3844345• • · (.0211764) .5083135••· (.02598) .7096411 • •• (.0221647} 1.196042•• · (.0223244) .1187584••· (.0098344) .0040867 (.0109363) .0324234*** (.0105753} .6308854*** (.019615) .2172464••· (.0131751) .0803049••· (.0100242) .1311516•• · (.011249) .9 .1325923••· (.0274726) .2917048••· (.0279492) .3757836*** (.0335228) .5214944••· (.0304094) .6979481••• (.0327972) .8842974••· (.0310439) 1.451293••· (.0302028) .1307988• •· (.014494) .1149119••· (.0136031) -.0052923 (.0145576) .7766481... (.0257003) .3178652••· (.0167693) .0829771... (.0139027) .0847959••· (.0174425) �142 / The Informal Sector in Mexico: Characteristics anti Dynamics -.0015701 (.0096832) -.1195821•.. (.0168307) .0928302... (.0078159) .0164481• 0 (.0011483) -.0003104• 0 (.0000216) 1.723015... (.031586) 0.0960 CeniaSouth Man Working Experieoce Working Experience' Constan! R, .0708555... (.0091711) -.1859325• 0 (.0147673) .085109• 0 (.0075597) .0230506• 0 (.0009473 -.0003'rº (.0000175) 2.035036... (.023684) 0.1575 . 0360782••· (.0088733) -.1420149... (.0141396) .0819983••· (.006441) .0183565••· (.0008857) -.0003189... (.0000163) 1.873853°• (.0195912) 0.1247 Revista Perspectivas Sociales/ Social Perspectives primavera/spring 2()()7. Vol.9, Num. I / .0640071 ... (.0112336) -.2544137'.. (.0200547) .085529••· (.0088772) .027509• 0 (.0011634) -.0003982• 0 (.0000223) 2.211142• 0 (.0260835) 0.1922 -.0018299 (.0184506) -.3056093• 0 (.0281755) .0928522••· (.0126079) .0317108... (.0016531) -.0004142*º (.0000332) 2.378112... (.03677) 0.2155 143 The wage premia for labor categories are shown in figure 5.6. With the exception of the government category, in ali other categories there are growing premia as the income curve advances . Figure 5.6 Premio salarial por ocupación 1991 90 80 N= 50709. Standard Errors in pareotheses. Bootstrapping method is used. • 0 ,•• , • Signficance levels at 1, 5 aod 10"/o respectively. Base category: no schooling; formal employmeot; Capital regioo; female. 70 60 50 - -Asalariado onformal Toe returns to education by levels shown in Figure 5.5 are growing with the quantile, although for the high school category it tends to decrease very slightly. Once more, the highest education levels tend to grow in their wage premia faster than in the lower education levels. For the lower income levels, the highest wage premia are in the high school category (complete), as for the highest income levels the professional levels offer a larger wage premia. <40 - -Gobierno 30 - - Patrón ~ 20 Autoem ado 10 o -10 -20 0.1 0.25 0.5 0.75 0.9 Cuantiles (niveles) de ingreso Source: Calculations are based on figure 5.3. Comparison category: formal Figure 5.5 occupied. Premio salarial por niveles educativos 1991 18 16 14 12 10 --+-,l)rimana ~ 8 ~manacon'C)leta .....secundana ,ncompieta 6 --M--5"'Undana completa 4 Informal employment shows a negative wage premia in the lower levels and it is not until quantile 75 that it evens formal employment. It is significantly different and positive in the higher income group. For the self-employed, the premia is negative for the lower income level only, turning into a positive for ali the other quantiles. This is quite different compared to the other years analyzed in this report, in which the midrange income group is where the premia turns positive. ~ l'ICOIT4)leta - 2 epacoo1)1eta ofesoonal o O1 025 0.5 075 09 c:u.,111es (nlveles) de Ingreso Note: The rates are calculated in base to the previous education level. For those working for the government, based on chart 5 .6, the premia is positive, although it is only slightly different to zero and not significant statistically speaking from the base of formal employment in groups in the upper income levels. Toe performance of government workers is the lowest and decreases faster than in the other years, 2004 and 1996. In the case of the employer category, the tendency for the previously analyzed �144 / The lnfomu,f Sector in Mexico: Choracteristics amiDynamics years persists, being positive and growing steadily. This is the highest yield from the labor categories. Just the same as previous years, the married, head of household and male variables all have positive premia. Toe northem region shows a positive premia in comparison to the workers from Mexico City, while the southern region reflects a negative premia. For the mid-region, the premia are negative at both ends of the curve although not very significant. Meanwhile, the premia are more signi:ficant and positive for the mid-range income levels. General Findings ...__ From the analysis shown in this section, we can infer that returns to education levels are higher when the worker is at a higher income leve!. After 1996, it seems like returns to lower education levels are reduced compared to the previous years, while the highest levels increase slightly. The effect seems connected to the increase in informal activities which causes a drop in the number of retums and an inefficient allocation of skills especially in the lower levels of education. (see Rodríguez-Oreggia, 2005). Employer and government activities present the highest wage premia, while informal employment (no social security) have negative wage premia ifcompared to formal employment. Just the same, self-employed individuals have a negative wage premia (compared to formal salaried positions) if they are in the lower income levels, which tends to change once in the upper income levels. As for informal employment, maybe the market <loes not offer a different job altemative, and the same for self-employment in the lower income levels. For the self-employed in the upper income levels, however, the market offers an altemative in this sector, where they receive a larger wage premia. Perhaps a focal point in terms of better working condition policies could be the integration of these groups (informal employment and self-employment in the lower income levels), which will be considered in the following section. Revista Perypectivas Sociales / Social Penpeclives primaveralspring 2007. Vo/.9, Num. f / 145 Conclusions This study aims to analyze the dynamics of the labor market taking as reference the informal sector of employment in different points in time. Especially, changes have been determined between labor sectors using transition matrices to determine labor mobility, as well as the socio-demographic and labor historical effects than have taken place in decision of being in a specific sector, and the proportionate returns in sector by labor income. Toe contribution of this study is clearly established in the understanding ofthe dynamics followed by the informal sector and compared to the formal one. Toe relevance ofthe analysis is understood in the sense that a great part of the market takes place in the informal sector, diverting resources and skills to activities with little impact on productivity and growth ofthe country. Additionally, workers in the informal sector are not protected in terrns of social benefits and their addition to the formal sector becomes relevant to policy making issues in the country. The understanding ofthe dynamics of the informality stated in this study is also relevant for poverty fight policies, since poor families get their income almost exclusively from the labor market, generally from the informal sector. From the transition matrices calculated, we can learn there is lirnited labor mobility between categories, which seems to happen mainly among the same informal categories (without social security). The less educated have a higher tendency to move to the informal sector if their income levels are low, and so it happens with self-employed. Additional to lower education, these groups are known for a higher proportion of children under twelve at home and located in groups of older age. The dynamics of these informal groups with a salary and self-employed is originated mainly from the fact of staying in the same informal category, but having strong movements coming from those without a salary or unemployed. This mobility is limited to informal categories within workers who are already located in these categories and <loes not allow an adjustment to labor offer, which leads to a vicious cycle of informality in these segments, taking it into a trap for individuals with jobs that offer lower premia for skills. �146 / 1ñe Informal Sector in Mexico: Characteristics and Dynamics Revista Perspectivas Sociales/ Social Perspectives primavera/spring 2007. Vo/.9, Num. / / Toe mobility mentioned above makes us think, first of all, that it is easier to enter the informal sector, and secondly, that there are barriers to enter the formal sector. Those barriers are mostly derived from regulations that mean a higher cost of employing in the formal sector (with social security),5 while the other part comes from a regulation framework tbat does not allow a broader creation of jobs (see Banco Mundial, 2004a). If integration policies from the informal to the formal sector are to be applied, there should be a coordination with the regulation framework and suggest improvements. but in getting the jobs in which temporary learning and formality are exchanged to lower wages. In this case, labor policies focused on the most vulnerable groups to stay in informal jobs (less educated, older), cannot be non coordinated from those implemented to create small business, where apparently there are incentives to stay in the informality, without training nor incentives to acquire more skills. 0n the other hand, the limited mobility shown in the previous analysis is higher among informal categories, also derived from the access to sorne resources tbat can be necessary for a higher mobility to formal jobs (with a salary or self-employed, where wage premia for sucb characteristics are higher), where the resources considered are education, training and access to capital. A basic requirement to have a higher mobility to formal sectors derives from equal opportunities to access education. Evidence in this paper indicates there is indeed a direct impact of education on mobility between labor categories, especially to the formal one. However, the distribution of education in general (not accounting for quality) is still uneven in the country, especially in higher grades of education. In fact, distribution has a regressive effect on higher levels, i.e. families with bigher incomes take the higher benefits (see López-Acevedo and Salinas, 2000 and 2000a). These limitations will certainly keep affecting tbe structure of limited mobility in the Mexican labor market for sorne generations if labor policies are not coordinated with education policies. Youngsters, as it was shown, seem to find a formaljob more easily tban adult workers, who usually find opportunities in the informal sector. lf we link this to tbe fact that the Mexican labor market develops mainly in small or micro enterprises, it seems lik:e the interest of companies is not given mainly by the creating of consistent skills for those companies, 5 For instance: Garro, Meléndez and Rodriguez-Oreggia (20O5)find that there is an increase offormal jobs related to diminishing contributions to social security in the 1997 reform, but such increase in too little compared to the size ofthe total labor force. 147 Finally, this study has sbown there are family nets that certainly have an impact on individuals obtaining a formal job, especially if another member of the family has already got a job of this kind. Funkhouser (1997) previously pointed that, for Guatemala, the decision of labor offer and hence labor mobility, are partly based on the family, vicinity and common networks. Socialization channels of individuals in our country occur through home and, as stated in this study, influence the limited labor mobility of workers. �148 / 7ñe Informal Sector in Mexico: Characteristics and Dynamics Bibliography Abadie, A.; Angrist, J. e Imbens, G. (2002). Instrumental variables estimates of the effects of subsidized training on the quantiles of trainee earnings. Econometrica, 70( 1), 91-117. Revista Perspectivas Sociales I Social Perspectives primavera/spring 1007. Vol. 9. Num. 11 149 Goldberg, P. y Pavcnik, N. (2003). Toe response ofthe informal sector to trade liberalization. NBER Working Paper 9443. NBER, Cambridge. Harris, J. y Todaro, M. (1970). Migration, unemployment and development: A two-sector analysis. American Economic Review, 1970, 60(1), 126-142. Auerbach, P.; Genoni, M. y Pagés, C. (2005). Social security coverage and the labor market in developing countries. Research Department Working Paper 537. Banco Interamericano de Desarrollo, Washington. Itzigsohn, J. (2000). Developing poverty. Pennsylvania, Pennsylvania University Press. Inter-American Development Bank (2004). Good jobs wanted: Labor markets in Latín America. IADB, Washington. Koenker, R. (2005). Quantile regression. Econometric Society Monograph Series. Cambridge University Press, 2005. Becker, G. S. (1962). Investment in Human Capital. Journal ofPolitical Economy, 10, 9-49. Koenker, R. y Basset, G. (1978). Regression quantiles. Econometrica, 46, 33-50. Calvó Armengol, A. y Ioannides, Y. (2005). Social networks in labor markets. New Palgrave Dictionary ofEconomics and the Law, 2nd edition. Palgrave-Macmillan. Krebs, T. y Maloney, W. F. (1999). Quitting and labor turnover: microeconomic evidence and macroeconomic consequences. World Bank Policy Research Working Paper 2068, World Bank, Washington. ECLAC (2005). Objetivos de Desarrollo del Milenio: Una Mirada desde América Latina y el Caribe. Santiago, ECLAC. Lewis, W. A. (1954). Economic Development with Unlimited Supplies ofLabour. The Manchester School, vol. 22, 139-191. De Soto, H. ( 1986). El otro sendero: La revolución informal, 8va edición. Bogotá, ILD. López-Acevedo, G. y Salinas, A. (2000). Toe distribution of Mexico's public spending on education. Policy Research Working Paper 2404. World Bank, Washington. Doeringer, P. y Piore, M. (1971 ). Interna/ Labor Markets and Manpower Analysis. Lexington, Lexington Books. Funkhouser, E. (1997). Demand-side and supply-side explanations for barriers to labor market mobility in developing countries: The case of Guatemala. Economic Development and Cultural Change, 45 (2), 341366. Garro, N.; Meléndez, J. y Rodríguez-Oreggia, E. (2005) Un modelo del mercado laboral mexicano con y sin seguridad social (JMSS). Documento de Trabajo 7, IIDSES, Universidad Iberoamericana. López-Acevedo, G. y Salinas, A. (2000). Marginal willigness topay for education and the determinants of emollment in Mexico. Policy Research Working Paper 2405. World Bank, Washington. Maloney, W. F.(1999). Does informality imply segmentation in urban labor market? Evidence from sectoral transitions in Mexico. The World Bank Economic Review, 13 (2), 275-302. McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. En P. Zarembka (Ed), Frontiers in Econometrics. Academic Press, New York. �150 -..... I 11,e Informal Sector in Mexico: Characteristics ami Dynamics Revista Perspectivas Sociales/ Social Perspectives primaveralspring 2007. Vol.9, Num. 1/ 151 Mincer, J. (1962). On-the-job training: Costs, Returns and Sorne hnplications. Joumal ofPolitical Economy, 70, 50-79. employment in LDCs: Toe labor turnover model. Quarterly Journal of Economics, 88(1), 194-227. Ozorio, A. Alves, L. y E. Graham. (1995). Poverty, deregulation and employment in the informal sector of Mexico. ESP Discussion Paper 54, World Bank, Washington. Tannuri-Pianto, M. y Pianto, D. (2004). Informal employment in Brazil: A choice at the top and segmenta/ion at the bottom - A quantile regression approach, Universidad de Brasilia, Departamento de Economía, Brasil, mimeo. Portes, A. (1994). Toe informal economy and its paradoxes. En N. J. Smelser y R. Wedberg (eds) The Handbook of Economic Sociology. Princeton, Princeton University Press. Tokman, V. E. (1992). Beyond regulations: The informal economy in Latin America. Boulder, Lynne Rienne. Portes,A., Castells, M. y Benton, L. (1989). The informal economy. John Hopkins University Press, Baltimore. World Bank. (1995). World Development Report: Workers in an integrating world. World Bank, Washington, 1995. Portes, A. y Sassen-Koob, S. (1987). Making it underground: Comparative material on the informal sector in Western market economies. American Journal ofSociology, 93, 30-61. World Bank (2004) La pobreza en México: Una evaluación de las condiciones, las tendencias y al estrategia del gobierno. World Bank, Washington. Rodríguez-Oreggia, E. (2005). ¿lnstitutions, geography and tbe regional evolution of returns to schooling in Mexico. Documento de Trabajo No 8, Instituto de Investigaciones sobre Desarrollo Sustentable y Equidad Social, Universidad Iberoamericana, México, 2005. World Bank (2004a) Doing business in 2004: Understanding regulation. World Bank, Washington. Rosen, S. (1986). Compensating Wage Differentials. Handbook ofLabor Economics, Vol. L O. Ashenfelter y R. Layard (eds), Elsevier Science Publisbers BV. Rosenzweig, M. R. (1988). Labor markets in low-income countries. Handbook ofDevelopment Economics, Vol 1, En H. Chenery y T. Srinivasan (eds). NewYork, North-Holland, 1988. Scbultz, T. W. (1961 ). lnvestment in Human Capital. American Economic Review, 51, 1-17. Shaphiro, C. y Stiglitz, J. ( 1984). Equilibrium unemployment as a worker discipline device. American Economic Review, 74, 433-444. Stiglitz, J. (1974). Altemative tbeories of wage determination and un- �152 / The Informal Sector in Mexico: Characteristics and Dynamics Appendix Transition Matrix 2003-2004. Age 46-55 Informal lnfonnal Transition Matrix 2003-2004. Ages 18-25 11.69 69.58 2. 09 0.19 1.62 0.29 4 .37 10.17 100 6.09 9.68 27.14 7.11 61.51 6.45 6.53 31.94 0.00 1.01 1.90 12.69 9.68 ?.28 1.02 29.03 34.17 4 .18 3.55 3.23 3.52 0.00 35.48 1.01 0 .76 2.03 6A5 29.66 100 100 100 100 4 .18 2.24 4 .36 0.42 0,08 0.44 S.02 2.03 4.25 0.84 3.11 3.98 IS.90 3.57 4 .05 26.36 72.82 41.16 100 100 100 lllfonn.i Emnlovmeol F_,al Eu,nlnvmUII PublicSedar EmnlnvaSelf-emnlovment Withoot F.mnlnvmeot S0.67 Formal Fmnlnvme:nt 18.57 23.95 4.52 5.32 24.27 9.38 20.0S 23.01 6.76 21.72 Poblic Employ..- Self-<m¡>IO)'llleol Withoul 0.63 0.36 S.92 3.23 Unanployed Inactive Total 3.50 17.13 •-en1 22.11 100 --en1 Uaemnlnved illadiv, Total Transition Matrix 2003-2004. Age 26-35 lllform.i EmnlavmUII Fonnal F,.mnJnu.mmt lnfonnal Formal Eu,nlnvmUIJ Emnlnvmcot 53.81 IS.97 9.49 Self•employmenl 2.88 2.08 11.ll 2.12 1.00 3.12 16. 34 1.97 13.78 100 4.23 0.38 2.31 3.46 100 90.91 1.68 0.67 46.64 3.87 6.72 5.15 0.17 1.26 1.92 48.60 O.SI I.IS 1.18 30.67 62.66 17.76 26.17 100 100 100 100 2L74 7.24 0.00 19.58 2.46 10.87 0 .24 1.01 19.57 84.02 32.89 100 100 100 Un<mployed Ina<tive Total 7.31 80. 38 O.SI 2.19 6.30 1. 79 0.93 0.00 1.87 15.22 1.18 13.23 8. 70 0.63 14.27 4.35 0.39 4.60 6.72 12.28 3.74 19.51 4.65 11.96 Informal 1.65 Unemploy,d 0.00 0.90 0.93 lnactive Total 13.SS Emnlnvmcot 43.56 Formal Emnlmnneot 12.44 Public Sede, 0.89 4.44 17.33 Withoot navmeot 0.44 l.33 19.56 100 Emolavmcot Fonnal 10.47 6S.4S 2.09 2.09 3.14 0.52 1.57 14.66 100 0.00 3.45 2.13 0.00 79.41 0.00 0.21 0 .00 0.74 45.69 4.l6 3.28 5.15 0,74 0.86 3.19 Withoot 1.47 6.90 9,79 1.64 0.00 0.86 0.64 0.00 12.50 9.48 20.00 44.26 100 100 100 100 Uoanok7Ved 5.88 5.88 laactiv, 2.71 8.88 0.27 7.37 0.00 0 .09 S.00 0.00 0.90 4.ll 0.27 0.00 0.36 0. 60 70_j9 87.96 100 100 100 Inactive Total Self--emplovmmt 1.98 10.42 ........ 2.26 3.85 2.56 100 100 14.27 10.24 0.94 39.89 4.62 3.61 1.21 29.26 54.03 IS.66 0.13 2.13 2.38 33.13 0 .94 1.06 1.06 0.00 31.33 100 100 100 100 25.00 6.27 16.18 26.4 7 3.43 24.91 6.62 1.71 12.35 1.47 0.12 2.63 S.88 13.11 0.00 2.01 2.08 11.76 1.71 1.87 22.79 76.35 26.86 100 100 100 8.40 2.9.1 0.58 51.38 Self-loym,ot lllÍonnal Un,mployed 1.60 2.64 1.20 15.43 1.57 1.35 Employ..- Transition Matrix 2003-2004. Age 56-65 PoblicSedc, Emnlnv.,. Withoot oavmenl l.29 86.SO 2.97 Em•-... Self-•I~ " " W"ithool navment Unemnmed loM:tive Employ... Inactivo Total Poblic Employ..- Self-employmeot Emnlnvmeot 4.72 7.98 7.40 4.82 Pa.blic Sedor Total n.90 Pobli< Emnlnvmem Fonnal Emnlmnn,ot Poblíc: Sector Em•Self~lavment Without navm,ot UnemoJOYed Wilhoot navm,ol 1.18 Formal Emnlovmail 10.83 Em•,,__cot Transition Matrices by Age Group, 2003-2004 lllfom,al 153 Revista Perspectivar Sociales I Social Perspectives primavera/spring 20()7. Vo/.9, Num. / / 2.66 13.61 Total 32.76 59.19 24.59 26.23 11.65 0.00 1.36 6.33 19.78 2.15 Sl.92 Transition Matrices by Age Groups, 1995-1996 Transition Matrix 1995-1996. Age 18-25 Emolavmem 49.84 Forn1al Emnlovm.., 20.27 12.72 71.99 - 6.67 24.32 24.57 21.13 23.61 10.04 19.66 Transition Matrix 2003-2004. Age 36-45 lllformal Formal lnfonnal Em•...... cot Emnlnvmcnt IS. IS 51.21 Emnlmrmait Fonnal Emn lmnnent l'ublicSedne 10.94 75.49 Em•""'" Self-emolovmcat Wilhoot 1.93 7.12 11.16 2.33 2.31 7.74 3.01 2.33 UaemnlMl'ed u .n loadive 4.63 12.53 11.02 1.2.l 16.66 - .., Total Public 2.41 2.34 lofonnal Employ... 1.22 1.46 Self-loym"" Withoul 14.21 . .1 0.67 3.81 0.10 90.94 1.86 o.n 51.08 26.32 1.55 l .90 0.00 7.81 4.07 60.38 17.44 1.34 44.n 3.94 1.25 17.22 3.94 0.89 5.14 L.73 0.19 15.15 0.79 7.01 15.92 2.55 2.43 Uoanployed 2.14 1.86 0 . )9 1.24 2.01 O_j8 13.39 1.25 1.63 load.ive Total 10.99 4.00 1.93 3.10 12.39 28.49 24AI 81.18 28.47 100 100 100 100 100 100 100 100 100 lofomll!I Emnlnvmem Formal Em.nlnvmco.t l'llblic Sector Eu,n lnv" Sclf......n1nummt Wilhout ~ ..1 Oaanninved Iaactive Total Poblíc: Employer Self-emplnyment lnactivt Total O.SS 6.6S Wilhout oavm,ot 3.29 Uoanployed 2.35 3.99 13.07 100 1.99 0.44 1.69 1.10 3.38 6.69 100 6.42 S.41 9.22 7.04 70.12 S.41 4 .78 1.97 0.00 21.62 3.75 0.00 2.47 18.92 34.4 7 6.20 1.23 S.41 1.SI 30.42 3.70 2.70 3.41 2.j4 9.38 16.22 12.29 30.70 100 100 100 100 24.28 5.15 6-96 2.46 0.00 0.14 23. 04 6.50 O.SI 4.01 2.42 4.76 2.90 3.2.l 4.26 13.14 3.84 4.27 26.73 70.89 37.01 100 100 100 �154 Transition Matrices by Age Groups, 1990-1991 Transition Matrix 2003-2004. Age 26-35 lllfnnnal Emnlovmml laformal Fonnal Emnlovma,l E,gn,__,a,l 51.99 16.86 l'Dblic Employ..- Self-anploymcnt Unemploytd lo.active Total 3.16 2.22 11.36 1.17 2.46 10.77 100 1.97 1.42 3.66 0.27 2.37 2.71 100 Without -mi Formal Emnlovmcnt Public Stdor EmnlnvaSclf...-nlnvmau W-1Joot 7.93 2.81 7.36 13.92 9.87 3.54 8.14 5.36 l.!17 87.29 1.94 2.17 1.32 0.73 46. 12 7.28 4.61 2.60 28.68 54.41 13.16 0.00 4.65 3.07 40. 13 0.63 L94 2.55 1.32 2.40 1.16 11.24 27.63 100 100 100 100 Unaanlavrd 20.62 6.26 13.60 15.98 2.99 23.10 13.40 1.25 14.84 3.09 0.38 3.73 9.79 5.50 12.53 1.03 1.80 2.24 12.37 1.41 2.13 23.71 80.41 27.82 100 100 100 - .., laadivt Total 79.66 - Transition Matrix 1990-1991. Age 18-25 Public EmployaSed« 2.62 6.12 Self-anploymmt Withoat Un<mploytd luad.iw Total 0.58 0.00 15.16 100 húonnal 40.23 fmn.,_,,,,tlll Formal [mn.,_,,,,tnt PDblicStdor Emn'""cr Self--nl-tnl Withoot 10.12 64.49 5.76 4.05 8.26 0.47 0.31 6.54 100 3.38 8.37 13.16 1.79 10.71 17.24 11.40 0.00 6l.60 6.90 5.04 7.14 4.14 35.96 7.24 179 6.20 24. 14 45.39 8.93 0.19 1.97 1.10 26.79 0.56 0.49 0.66 0.00 9.21 4.93 16.01 53.57 100 100 100 100 13.64 3.84 10.03 22.73 3.92 19.81 0.00 3.59 14.15 4.55 0.49 4.92 22.73 6.78 13.74 4.55 2.94 L93 13.64 0.33 0.46 18.18 78.10 34.96 100 100 100 ,_ ... Ooemnlmied lnadiw T'"al -tnt 12.54 lnfonnal fmnlnvmtnt Formal fmnlnvmmt Public Sector Em--... Self-•---· Withoot 10.64 1.31 4.71 10.09 4.12 74.28 1.87 5.10 3.03 2.06 2.00 87.29 2.35 1. 15 1.03 2.22 1.87 52.94 9.22 4.12 4.66 162 27.45 54.90 17.53 Unanploytd lnactivt T ..al 0.29 2.60 12.43 100 OM 1.11 4.66 100 Withoot 0.19 1.57 1.73 45.36 0.19 1.57 1.15 0.00 4.67 4.31 18.73 25.77 100 100 100 100 22.22 84.81 36.&5 100 100 100 "-mi Ua.cm .........cd 1aadive T'"al 22.22 3.87 9.56 5.56 0.61 11.57 1.39 0.53 13.38 4.17 0.61 6.77 27.78 6.83 18.18 U7 1..59 2.34 12.SO 1.14 1.35 Fcnnal E,nnlovmmt 12.24 67.35 1.36 1.36 2.72 0.68 0.68 13.61 100 Pal>licSednt 2.17 11,72 6.94 3.70 1.45 1.56 0.72 0.00 81.16 1.56 0.72 0.00 145 50.00 6.22 0.00 2.17 21.88 59.09 16.67 0.00 0.78 2.63 42.59 0.00 0.78 1.91 0,00 11.59 11.72 21.77 37.04 100 100 100 100 11.11 1.85 7.69 5.S6 0.48 5.56 0 .00 0.32 5.31 13.89 25.00 6.27 17.50 2.78 2.09 2.76 13.89 0.24 4.59 0.89 27.78 88.42 55.69 100 100 100 lafom,,I FJD..L-..mt E,nnt..,a- Sclf-nlr,vmeot ... Without - U■an--i-ec:t laad:ive T..al húonnol Fonnal Public E,nn....,_tnt Em......,,tnl 46.03 8.99 1.06 Employ..- Self-tmploym,nt _.., W~out Uncmploytd laadm T'"al 3.17 17.99 1.06 1.06 20.63 100 0.32 IDac:tivc Tola! 3.46 0.29 S.77 3.46 16 74 100 61.48 3.25 0.36 2.98 1.45 2.35 12.57 100 Emnlnvmeot Pnblic Stdor 11.31 Ea,nlov,r 28.51 Self-nl-tn1 Witbout ~mt 18.79 11.21 16.25 19.05 20.61 1034 57.95 4.76 1 .21 0.86 0 .00 14.29 3.03 0 .00 1.77 14 29 35.15 6.90 0.35 9.52 3.64 32 76 3,18 0.00 1.82 1.72 9.19 9.52 15.76 36.21 100 100 100 100 Uoemnlnved 15.52 8.42 16.01 30.17 12.40 29.00 11.21 3.52 7.13 0.86 0.12 0.40 S.17 2.02 4.44 0.86 3.17 3.40 6.90 2.83 2.86 29.31 67.53 36.75 100 100 100 Unanploytd lnadiw Total Transition Matrix 1990-1991. Age 26-35 Informal Emolovmml Fonnal fmn...,,,,ml 9.41 68.54 4.56 1.52 7.41 0.38 1.24 4.67 9.42 2.44 15.30 15.22 17.40 6.10 64.57 4.35 6.26 6.10 1.29 42.03 6.73 3.66 4. 19 22.46 38.52 13.41 0.32 0.72 1.16 2.3.17 0.64 1.45 1.16 1.22 9.0Z 4 .35 14.85 43.90 30.91 S.81 11.25 21.82 4.49 25.55 7.27 J.79 12.93 o.oo 9.09 4.30 10.00 0.00 2.5J 1.61 7.27 0.95 I.OS 2.3..64 77.70 34.47 Inronn,I Fonnal Emnlavmmt 38.00 F.mnlnvmtnt 28.00 1 .EmolD'Yer 1 Sclf-cmol-cnt 13.92 1 oavmm1 Without l'llblic Employe, Self-anploymeot Witboat 5.20 3.80 12.20 0.20 0.60 12.00 100 6.94 100 n--tlll 0.44 3.14 100 100 100 100 100 100 100 Transition Matrix 1990-1991. Age 36-45 Informal Jnfonnal FC111Ual Public Employ..- Sclf-cmploymcut Witbout Onemploytd lnadin Total E'mnlM,n,a¡J EmnL-.vmmt 22.74 Sedor 6.12 2..62 12.54 n-mt 0.58 0.00 15.16 100 64.49 5.76 4.05 8.26 0.47 O.JI 6.54 100 10.71 17.24 11.40 0.00 65.60 4. 14 35.96 7.24 1.79 6.20 24.14 45.39 8.93 0.19 1.97 1.10 26.79 0.$6 G.49 0.66 0.00 9.21 4.93 16.01 53.57 100 100 100 100 4.55 0.49 4.92 22.73 6.78 13.74 4.55 2.94 1.93 0.33 0.46 18.18 78.10 100 100 100 40.23 EDJnlnvn,ml Formal 10. 12 Emnlovma,t Pnblic Sector J.38 Emn.,,..,. 8.37 Self-nl-cot 13. 16 Without 1.79 navmeot Uuemo1avcd lnactivt T'"al Transition Matrix 2003-2004. Age 56-65 Unomploytd 15.55 Unao"lftVed lnac:tive "-mi Self....,ploymtnl 3U2 Toeal lllfnnnal Formal l'llblic Employer Self-tmploymenl F.m ..i...-..e:ot F.m..'---mt Sector 43.93 12.43 1.45 6.07 20.81 Employer Formal PublicSed« Transition Matrix 2003-2004. Age 46-55 l'llblic Stdor Wihoul n-•tnl 3.61 Toeal Fonnal Em-'--tnl 22.74 Formal fmnlmnntnl Transition Matrix 2003-2004. Age 36-45 Emn....,_a,t lofonnol F.,mnl,wn,eot [mn.....,nt 29.15 ' laf0'1Dol lnadÍ\--c lafonnal 155 Revisto Perspectivas Sociales I Social Perspectives primavera/spring 2007. Yol.9, Num. I / / The Informal Sector in Mexico: Charocteristics and Dynamics 13.64 J.84 10.03 22.73 3.92 19.81 6.90 5.04 7.14 0.00 J.59 14. 15 13.64 34.96 �156 / Revista Perspectivas Sociales I Social PerspecJives primaveralspring 2007. Yol.9, Num. l I Pág. I 57-175 The Informal Sector in Mexico: Characteristics ami Dynamics Transition Matrix 1990-1991. Age 46-55 Informal IDfonual Formal Em•'"""'ml Emnlnvmaat 38.S4 14.06 Public Employtr Sedar 6.25 5.21 S<lf-cmployma,t 14.58 - .., Wihout Uuanployed laad:ive Tcul 1.56 1.56 18.23 100 157 Access and Use ofHealth Care Services by Mothers and Children in the Texas-Mexico Border Region: Preliminary Findings from the 2006 Rio Grande Valley Health Survey Emnlovmmt 11.32 59 75 S.66 4.09 8.18 OJl 1.89 8.81 100 5.13 5.93 12.39 14.81 9.68 9.09 Wibout _.,,. 3.23 0.00 1.1,8 42.96 7.62 0.00 3.42 26.67 50.44 6.45 0.43 0.00 Z.05 25.81 0.85 0.00 0.29 0.00 11.11 Sclf-emolovm.ml 65.38 5.93 4.69 3.23 3.70 16.13 61.29 100 100 100 100 Uoaon.,..cd IDaaiw T'"al 11.11 22.22 2.48 14.50 0.00 1.76 10.10 11.11 2.58 8.53 1.14 5.48 33.33 4.03 14.09 0.00 1.24 1.44 0.00 0. 10 0.58 22.22 86.67 45.29 100 100 100 Formal Patricia B. Reagan* José A. Pagán** Emnlovmcat PublicScdor Em•- Transition Matrix 1990-1991. Age 56-65 lllform,1 lllform,1 Formal Emolovmcat 35.34 f.mn~-.uall 15.52 - ... Public Employtr Sedar S<lf-cmployma,I Wíboul Un... ployed Iald.ive Tc,aJ 1.72 3.45 16.38 0.00 2.59 25.00 100 EDJn1-nnent kmal 12.23 57.02 3.51 0.88 8.71 0.00 0.88 16.67 100 3.03 11.67 6.11 0.00 13.64 8.33 5.00 000 57.l8 1.67 1.67 0.00 1.52 36.67 7.78 0.00 3.03 21.67 51.11 1.52 1.67 1.67 23.08 0.00 0.00 0.00 0.00 19.70 18.33 26.67 46.15 100 100 100 100 0.00 2.06 6.84 25.00 1.80 0.00 0.90 4. 13 0.00 1.29 3.91 75.00 0.00 1.67 1.58 o.oo 9.09 0.00 86.12 59.80 100 100 100 Em•'"""'cat PDblic S<<lo, Em•~cr s.lf-••-ai1 W~oul oavmenl UoemnlDVed Inactive T'"al JO.TI S.91 14.10 0.26 0.45 Abstract Background: Latino/a children have the highest rate of uninsurance among all ethnic/racial groups. There is sorne evidence that patterns of health care use between parents and their children are interrelated. Children are also more likely to have health insurance coverage if their parents are insured but public insurance programs do not cover all parents of participating children. The objective of this study was to assess whether there is a relation between patterns of health care utilization between Latina mothers and their children who reside in the South TexasMexico border region. Survey data on 495 Latina women with children from the 2006 Rio Grande Valley Health Survey were used to estimate bivariate probit models of the determinants of five health care access *Department ofEconomics, The Ohio State University 1945 N. High St. Co/umbus, OH 4322 1 Te/: 6/4-442-7385 reagan.3@osu.edu ** lnstitute far Population Health Policy and Department ofEconomics and Finance The University ofTexas-Pan American 1201 W. University Dr. Edinburg, TX 785412999 Tel: 956-318-5306 Fax: 956-318-5303 jpagan@utpa.edu Address correspondence to: Patricia B. Reagan, Ph.D., Professor, Department ofEconomics, The Ohio State University, 1945 N. High St., Columbus, OH 43221. Dr. Reagan is a/so Faculty Research Associate with the Center far Human Resource Research at OSU Dr. Pagán is a/so Director ofthe lnstitute far Population Health Po/icy al UTPA and Adjunct Senior Fellow of the Leonard Davis lnstitute of Health Economics at the University ofPennsylvania. Acknowledgements: Financia/ supportfor this study was provided by the National lnstitute of Child Health and Human Development through a seed grant from the lnitiative in Population Research al The Ohio State University. We thank Margare/ Plahuta and José E. Olivares f ar their excellent assistance in survey development. We also thank the participants of the 2006 Human Copita/ Symposium al the Universidad Autónoma de Nuevo León in Monterrey, México far their helpful comments and suggestions. ISSN 1405-1133 C 2007 Universidad Autónoma de Nuevo León, University ofTexas ofAustin, University ofTexas of Arlington, University ofTennessee, Universidad Juárez del Estado de Durango, Universidad de Colima. �158 / Access ami Use ofHealth Care Services by Mothers ami Children in the Texas-Mexico Border Region: Preliminory Findings from the 2006 Rio Grande Va/ley Health Survey and utilization indicators. Pattems of health care utilization between Latina mothers and their children were positively related for having a usual place of care, visiting a doctor, visiting an emergency room, and having delayed health care needed. Both desirable and undesirable parental and child health care access/utilization pattems for Latinos/as are interrelated. Interventions that promote good healtb care utilization behavior for Latina mothers seem to spillover to their children. Public health insurance programs that focus on covering uninsured children but leave tbeir parents uninsured may end up not taking full advantage of the positive spillover effects of health care access/utilization from mothers to their children. KeyWords Uninsured; Latino/a; Hispanic; Children; Health Care Resumen Los niños latinos tienen la más alta tasa de no aseguramiento médico entre todos los grupos raciales. Hay evidencia de la interrelación padres - hijos en los patrones de cuidados de la salud. Los niños tienen mayor probabilidad de estar cubiertos si sus padres lo están pero los programas públicos de seguridad no cubren a todos los padres de los niños participantes. El objetivo de este estudio fue evaluar la interrelación entre los patrones de utilización de cuidados de la salud entre madres latinas y sus hijos para residentes en la región de Texas fronteriza con México. Se utiliza la Encuesta de Salud del Valle de Río Grande 2006 con datos muestrales de 495 mujeres latinas con hijos y se estiman modelos Probit para cinco indicadores de acceso a los cuidados de salud. Los patrones de la utilización de los cuidados de la salud para madres latinas y sus hijos estuvieron positivamente relacionados para los indicadores de: lugar habitual de cuidado de la salud, visita al médico, visita a salas de emergencia y posposición en los cuidados. Los patrones de utilización de los cuidados de salud entre padres e hijos están interrelacionados tanto en formas deseadas como no deseadas. Las intervenciones que promueven una conducta de utilización de cuidados de la salud para madres latinas parecen extenderse a sus hijos por lo que los programas públicos de salud que se enfocan en los niños no cubiertos dejando de lado a los padres pierden este beneficio. Revista Perspectivas Sociales I Social Penpectives primavera/spring 2007. VoL9, Num. 1 I 159 Palabras clave Asegurados, Latino/a, Hispánico, Niños, Cuidados de la salud. Introduction Many communities in the U.S.-Mexico border region are characterized by high poverty rates and, not surprisingly, high uninsurance rates and low access to health care (Bastida, Brown and Pagán, 2007). About one of every four children and adults in Texas do not have health insurance coverage--the highest proportion in the U.S.- and access disparities are even higher in counties bordering Mexico. Slightly over one of every three children and adults in the Rio Grande Valley- which includes Cameron, Hidalgo, Starr and Willacy counties- are uninsured (U.S. Census Bureau, 2005). Lack of health insurance coverage has important negative consequences in terms of access to health care and health outcomes not only for individuals but also for entire families (Institute of Medicine, 2002; Institute ofMedicine, 2004). A recent report from the Institute of Medicine (Health Insurance is a Family Matter) concluded that children are more likely to have health insurance coverage if their parents are insured (lnstitute of Medicine, 2002). Many uninsured children are eligible for public programs such as Medicaid and the State Children's Health Insurance Program (SCHIP), yet, many uninsured children remain without health insurance coverage and there are severa} reasons for this. For example, many children become ineligible for sorne public health programs after a certain age and, thus, become uninsured as they get older. Employer-sponsored health insurance coverage is also important because the linkage between employrnent and health insurance coverage increases the chances that children become uninsured when parents transition from employment to unemployment, and vice versa. Recent research also has sbown that pattems of health care use between parents and their children are interrelated (Mink:ovitz et al., 2002). Parental age, ethnicity, race, education, socioeconomic status and the family unit structure are ali connected to a child's use ofhealth care services (Newacheck et al., 1998; Weinick, Weigers and Cohen, 1998). There is also sorne evidence that the self-reported health status of moth- �160 / Access and Use ofHealth Care Services by Mothers and Children in the Texos-Mexico Border Region: Preliminary Findings from the 2006 Rio Grande Va/ley Hea/th Survey ers and their children are correlated (Minkovitz et al., 2002). What is not known is whether sorne ofthese results extend to low income mothers and children, particularly ofLatino descent. Hispanics have the highest rate of uninsurance in the U.S. and a better understanding of the patterns of health care access and use between parents and children is an important health policy concern if we are interested in substantially reducing health disparities. In this study, we report preliminary results from the 2006 Rio Grande Valley Health Survey (RGVHS), a telephone survey of 908 women between the ages of 19 and 55 residing in Cameron, Hidalgo, Starr and Willacy counties. More specifically, we collected extensive socioeconomic, demographic, health status and health care utilization data on mothers and one randomly-selected child from each household. Toe preliminary results reported here pertain to 495 mother-child pairs for which we had complete data. Toe main objective of our study is to assess whether there is a relationship between patterns of health care utilization between borderdwelling Latina mothers and their children. This is an important policyrelevant issue because if good (and/or bad) health care utilization patterns of parents spillover to their children then it may malee sense to facilitate health insurance coverage to all parents regardless ofthe health insurance status of their children. Current SCHIP policy in Texas provides health coverage to some-but not all-parents of insured children participating in this public health insurance program (Task Force on Access to Health Care in Texas, 2006). Methods Data Description Toe 2006 Rio Grande Valley Health Survey includes socioeconomic, demographic, health status and health utilization data from 908 women between the ages of 19 and 55 that reside in Cameron, Hidalgo, Starr and Willacy counties in South Texas. Toe sample is representative of the estimated 290,811 women that resided in the RGV in 2006. Communities along this U.S.-Mexico border region have among the highestrates of poverty and uninsurance in the U.S. Toe most recent U.S. Census Revista Perspectivas Sociales I Social Perspeclives primavera/spring 2007. Vol.9. Num. / / 161 estimates show that 323,852 people (32.4 percent) were uninsured in the four-county RGV area in 2000, 120,269 of them were children. About 32 percent ofchildren/adults (34 percent of children) in this region were uninsured in 2000 (U.S. Census Bureau, 2005). Thirty-one percent of the population in this region lived below the poverty line in 2003 (42.5 percent of children ages 0-17) (U.S. Census Bureau, 2006). Telephone interviews were conducted in English or Spanish depending on the preference ofthe respondent from January to September 2006. The survey included sections with questions on demographics, immigration, health status, employment, health insurance, healtb care utilization and participation in social programs. Mothers also answered questions in these seven areas about one randomly-selected child between the ages of one to 13 years of age residing in the same household. Sampling weight adjustments were developed to account for households without a telephone and households with multiple residential telephone lines. Post-strati:fication adjustments were also conducted based on age, county of residence, education and income. The weighted sample is representative ofwomen 19-55 residing in the four RGV counties (290,811 in 2006, based on our own projections from the 2000 U.S. Census and the 2004 American Community Survey). Dependent Variables We constructed five mother-child pairs of dichotomous dependent variables based on the mothers' answers to the following questions: (1) "Is there a place that you usually go to when you are sick or need advice about your health?" and "Is there a place where you usually take him/her when he/she is sick or you need advice about his/her health?; (2) "During the past 12 months, how many times have you seen a medica! doctor about your own health? and "During the past 12 months, how many times has he/she seen any kind of medical doctor?"; (3) "During the past 12 months, did you visita hospital emergency room for your own health?" and "During the past 12 months, did he/she visit a hospital emergency room?"; (4) "During the past 12 months, did you visita dentist?" and "During the past 12 months, did he/she visita dentist?; and (5) "During the past 12 months, did you delay getting any other medica! care you �162 / Access and Use o/Health Care Services by Mothers and Children in the Texas-Mexico Border Region: Pre/imina,y Findings .from the 2006 Ria Grande Va/ley Health Survey felt you needed, such as seeing a doctor, a specialist or other health professional?" and "During the past 12 months, did you delay getting any other medical care for your son/daughter you felt he/she needed, such as seeing a doctor, a specialist or other health professional?". Toe motherchild questions on (2) above about the number ofvisits to a doctor were coded as one for one or more visits, and zero otherwise. We hypothesize that health care access/utilization patterns across these five dimensions--usual place ofcare, visiting the doctor, emergency room or dentist, and delaying medical care needed- are positively related between mother-child pairs, even after controlling for other factors that may be related to health care access/utilization. Revista Perspectivas Sociales / Social Perspeclives primavera/spring 2()()7. Vol.9, Num. I / 163 health care access/utilization varied for child-mother pairs across the independent variables included above (Greene, 2003). We used a bivariate probit specification because we are interested in evaluating the sign and statistical significance of the correlation coefficient of the residuals from the two jointly-estimated probit models. This coefficient tells us whether the health care access/utilization pattems of mothers and children are related even after controlling for all the factors posited to be related to access/utilization (i.e., perceived need for health care services, individual predisposing characteristics and factors that enable access to and utilization ofhealth care services). The bivariate probit models were estimated using Stata 9.2 (StataCorp, 2005). ResuJts Independent Variables ........ Toe determinants of health care access/utilization in a U.S.-Mexico border context for both mothers and children were selected based on the precept that access/utilization varies with the perceived need for health care services, individual predisposing characteristics and factors that enable access to and utilization of health care services (Andersen and Davidson, 2001; Andersen et al., 2002). Toe independent variables explaining the health care access/utilization patterns of children were years of age, gender, self-reported health status (fair/poor vs. good/very good/excellent), yearly household income (less than $10,000, between $10,000 and $30,000, between $30,000 and $50,000, and more than $50,000) and health insurance status (insured vs. uninsured). For mothers, the independent variables explaining health care access/utilization patterns were years of age, marital status (married vs. otherwise), immigrant status (immigrant vs. otherwise), self-reported health status (fair/poor vs. good/very good/excellent), education (less than high schooJ, high school and sorne college, and college graduate), yearly household income (less than$10,000, between $10,000 and $30,000, between $30,000 and $50,000, and more than $50,000) and health insurance status (insured vs. uninsured). Statistical Model We estimated five bivariate probit regression models to analyze bow Table 1 presents the weighted means and standard errors for all the childmother variables. Ninety-six percent ofchildren had a usual place where they obtained health care services but only 81.40 percent ofmothers had a usual place for medical care. Ninety-three percent of children had at least one visit to a doctor in the last 12 months befare the interview but only 80.44 percent of mothers had visited a doctor during the same time period. Only 12.00 percent of children had visited an emergency room in the last year compared to 16.08 percent ofmothers. Seventy-four percent of children had visited a dentist within the previous year compared to only 34.77 percent of mothers. Only 4.00 percent of children report to have had any delays in health care within the past year compared to 7.60 percent of mothers. �164 I Access ami Use ofHealth Care Services by Mothers and Chüdren in the Texas-Mexico Border Region: Preliminary Fimlings.from the 2006 Rio Grande Va/ley Health Survey TABLE 1: Sample Means 1 Mean Child-Usual Place Care 1 .9636 Motber-Usual Place Care 1 .8140 Child-Doctor Visit 1 .9340 Mother-Doctor Visit 1 .8044 Child-ER Visit I .1200 Mother-ER Visit 1 .1608 Child-Dentist Visit 1 .7371 Mother-Dentist Visit 1 .3477 Child-Delayed Care I .0400 Motber-Delayed Care .0760 Child-Age 5.2338 Mother-Age 33.2197 Child-Female .4570 Motber-Immigrant .7265 Child-Fair/Poor Health .1058 Mother-Fair/Poor Health .2769 Child-Insured .7943 Motber-Insured .2881 Married .6908 Less than HS .5433 HS & Sorne College .3552 College Graduate .1015 HH Income <$1 0K .4490 HH Income $10K-$30K .3945 HH Income $30K-$50K .0872 HH Income >$50K .0692 SE .0103 .0230 .0134 .0227 .0196 .0221 .0251 .0267 .0124 .0159 .1990 .4777 .0285 .0247 .0177 .0254 .0225 .0251 .0274 .0281 .0266 .0148 .0288 .0275 .0148 .0133 [95% .9433 .7688 .9076 .7597 .0834 .1174 .6878 .2952 .0157 .0446 4.8425 32.2801 .4010 .6778 .0710 .2270 .7501 .2387 .6370 .4880 .3028 .0725 .3925 .3404 .0580 .0430 CI] .9839 .8593 .9605 .8491 .1565 .2042 .7865 .4003 .0643 .1074 5.6251 34.1591 .5131 .7751 .1406 .3269 .8387 .3376 .7447 .5986 .4075 .1306 .5056 .4487 .1163 .0955 Toe mean age of children and tbeir mothers was 5 .23 years and 33 .22 years, respectively. Almost three quarters of the mothers surveyed (72.65%) were immigrants. About l 0.58 percent of the mothers reported that their children were in fair or poor health whereas 27.69 percent of mothers reported their own health to be fair/poor. Almost four-fifths of children (79.43 percent) were covered by any type of health insurance coverage compared to only 28.81 percent of mothers. This health insurance coverage rates are relatively low compared to what is typically reported for tbe U.S.-Mexico border region. Toe 2006 RGVHS only includes women between the ages of 19 and 55 and, consequently, the Revista Perspectivas SocitJ/es / Social Perspectives primavera/spring 2007. Vo/.9, Num. 1 I 165 health insurance coverage rates are likely to be mucb lower for this relatively younger population than for the overall population. Sixty-nine percent of women in the sample were married at the time of the interview. More tban half of the mothers (54.33 percent) bad less tban a high school education, 35.52 percent hada high school education or sorne college, and 10.15 percent were college graduates. Almost 45 percent hada household income of less than $10,000 per year, 39.45 percent hada household income between $10,000 and $30,000 per year, 8.72 percent earned between $30,000 and $50,000, and only 6.92 percent had a household income of more than $50,000 . Tables 2 to 6 report the coefficients for the five bivariate probit models of child-mother health care access/utilization. Before proceeding with the main results based on the correlation coefficient of the estimated residuals, it is worth noting a few of tbe statistically significant determinants of healtb care access/utilization for both mothers and children. Table 2 shows that girls were less likely to have a usual place of bealtb care than boys in our sample. Also, insured children were more likely to have a usual source of care. Children residing in households with income levels greater tban $50,000 had a higher probability of having a usual place for obtaining healtb care tban those earning less. Insured mothers were more lik.ely to bave a usual source of care tban uninsured mothers. �166 / Access and Use ofHealth Care Services by Mothers ami Children in the Texas-Mexíco Border Revista Perspectivas Sociales / Socíal Perspeclíves primaveralspring 2007. Vol. 9, Num. / / Region: Prelímínary Findíngs from the 2006 Río Grande Va/ley Health Survey TABLE 2: Bivariate Probit Model for Having a Usual Place of Care SE Coef. Child has Usual Place of Care 0.004 Age -0.370* Female -0.194 Fair or poor health 0.016 HH lncome $10K-$30K 0.273 HH lncome $30K-$50K 5.751 *** HH lncome >$50K 1.021 *** Insured 1.051*** Constant Mother has Usual Place of Care 0.010 Age 0.199 Married 0.179 I.mmigrant 0.008 Fair or Poor Health -0.100 HS & Sorne College 0.153 College Graduate 0.169 HH lncorne $10K-$30K 0.003 HH Incorne $30K-$50K 0.065 HH Incorne >$50K 0.560*** lnsured -0.114 Constant 0.569*** Rho \ \ 1 Chi2 Observations (0.026) (0.199) (0.278) (0.196) (0.328) (0.244) (0.179) (0.245) (0.009) (0.150) (0.165) (0.153) (0.145) (0.237) (0.147) (0.281) (0.333) (0.187) (0.385) (0.144) 15.602 495 Standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 TABLE 3: Bivariate Probit Model for Having Visited a Doctor Coef. Child-Doctor Visit Age - 0.029 Female 0.341 * Fair or Poor Health 0.101 lilI Income $1 OK-$30K -0.054 lilI Income $30K-$50K 0.060 lilI Income >$50K -0.106 Insured 1.017*** Constant 0.770*** Mother-Doctor Visit Age 0.002 Married 0.205 lmmigrant -0.017 Fair or Poor Health -0.016 HS & Sorne College -0.154 College Graduate - 0.053 lilI Income $10K-$30K 0.023 lilI Income $30K-$50K -0.412 lilI Income >$50K -0.005 Insured 0.552*** Constant 0.350 Rho 0.236* Chi2 Observations 3.525 465 Standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 SE (0.025) (0.194) (0.254) (0.202) (0.328) (0.345) (0.189) (0.250) (0.010) (0.154) (0.169) (0.154) (0.149) (0.226) (0.151) (0.259) (0.321) (0.178) (0.383) (0.126) 167 �168 \ / Access and Use ofHealth Care Services by Mothers and Children in the Texas-Mexico Border Region: Pre/iminary Findings from the 2006 Rio Grande Va/ley Health Survey TABLE 4: Bivariate Probit Model for Having an Emergency Room Visit SE Coef. Cbild-ER Visit (0.024) -0.051** Age (0.162) Female 0.035 (0.224) Fair or Poor Health 0.366 HH lncome $1 OK-$30K -0.236 (0.181) HH Income $30K-$50K 0.306 (0.279) (0.283) 0.272 HH lncome >$50K (0.202) Insured 0.197 (0.243) -1.161*** Constant Mother-ER Visit (0.012) Age - 0.016 Married 0.038 (0.186) (0.213) Immigrant 0.016 0.558*** (0.184) Fair or Poor Health (0.183) HS & Sorne College 0.105 (0.273) College Graduate - 0.222 (0.184) HH Income $10K-$30K - 0.429** HH Income $30K-$50K - 0.693* (0.375) HH Income >$50K - 0.316 (0.363) lnsured 0.336 (0.209) Constant - 0.674 (0.440) Rho 0.218* (0.121) Chi2 Observations 3.241 489 Standard errors in parentheses * p<0.1 O, ** p<0 .05, *** p<0.01 Revisto Perspectivos Sociales / Socio/ Perspectives primavera/spring 2007. Vol9, Num. J / TABLE 5: Bivariate Probit Model for Having Visited a Dentist Coef. Child-Dentist Visit Age Female Fair or Poor Health HH Income $10K-$30K HH Income $30K-$50K HH Income > $50K Insured Constant Mother-Dentist Visit Age Married Immigrant Fair or Poor Health HS & Sorne College College Graduate HH Income $10K-$30K HH Income $30K-$50K HH Income >$50K Insured Constant Rho chi2 e Observations SE 0.038** -0.060 -0.234 0.007 -0.133 0.256 0.714*** -0.134 (0.019) (0.134) (0.201) (0.145) (0.235) (0.294) (0.153) (0.204) 0.017* 0 .281* -0.057 -0.301** 0.173 0.326 0.416*** 0.211 0 .815*** 0.582*** -1.698*** 0.069 (0.009) (0.152) (0.158) (0.148) (0.147) (0.213) (0. 148) (0.242) (0.287) (0.158) (0.359) (0.092) 0.569 488 Standard errors in parentheses * p<0.10, ** p<0 .05, *** p<0.01 169 �170 / Access and Use of Health Care Services by Mothers and Children in the Texas-Mexico Border Region: Preliminary Findings from the 2006 Río Grande Va/ley Health Survey TABLE 6: Bivariate Probit Model for Having Delayed Care Coef. Child-Delayed Care Age Female Fair or Poor Health HH Income $10K-$30K HH Income $30K-$50K HH Income >$50K Insured Constant Mother-Delayed Care Age Married Immigrant Fair or Poor Health HS & Sorne College College Graduate HH Income $10K-$30K HH Income $30K-$50K HH Income >$50K lnsured Constant Rho ~ "-.__ ' \1 Chi2 Observations SE 0.000 -0.733** 0.366 -0.932** -0.631 -0.561 -0.656** -0.782** (0.032) (0.330) (0.367) (0.419) (0.441) (0.461) (0.290) (0.347) 0.002 0.059 -0.006 -0.093 - 0.119 -5.773*** 0.088 -0.256 0.585 -0.040 -1.509*** 0.424* (0.014) (0.269) (0.251) (0.251) (0.225) (0.511) (0.248) (0.520) (0.528) (0.296) (0.500) (0.217) 3.825 388 Standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 Revista Perspecüvas Socio/es I Social Perspeclives primavera/spring 2007. Vo/.9, Num. l l 171 Table 3 reports the results for having visited a doctor at least once within the previous year. Girls were more likely to have visited a doctor than boys. Health insurance was a statistically significant determinant of doctor visits for both children and their mothers. Table 4 reports the results for visiting an emergency room within the previous year. Toe child's years of age was negatively related to having visited an emergency room. Mothers in fair/poor health were more likely to have visited an emergency room than those in good, very good or excellent health. Household income was negatively related to visiting an emergency room within the last year. Table 5 shows that, for both the child and the mother, years of age were positively related to having visited a dentist within the Iast year. Insured children and mothers hada higher propensity ofhaving visited a dentist than uninsured children and mothers. Married mothers were more likely to have visited a dentist than non-married mothers. Mothers in fair/poor health were less likely to have visited a dentist than those in good, very good or excellent health. Household income was positively related to having visited a dentist. Table 6 shows the results for whether the child or the mother delayed needed health care within the last year. Girls were less likely to have delayed care than boys. Children with household income between $10,000 and $30,000 were less likely to report that they had delayed care than ali others. Insured children were less li.kely to have delayed needed health care than uninsured children. Mothers with a college degree were less likely to report that they had delayed care than those that did not have a college degree. Tables 2-6 also report the correlation coefficient for the residuals of the bivariate probit models. The residual correlation coefficients between the child and the mother probit equations were positive and statistically significant for having a usual place of care, visiting a doctor, visiting an emergency room, and having delayed health care needed. The coefficient was positive but statistically insignificant for having visited a dentist. �172 / Access ami Use ofHealth Ca1tl Services by Mothers anti Children in the Texas-Mexico Border Revista Perspectivas Socio/es I Socio/ Perspectives primavera/spring 2007. Yol.9. Num. J / 173 Region: Preliminary Findings from the 2006 Río Grant/e Ya/ley Health Survey Discussion Using survey data from the 2006 RGVHS we found that the residual correlation coe:fficients of child and mother bivariate probit equations of four health care access/utilization indicators were highly correlated. Regression-adjusted pattems of health care utilization between Latina mothers and their children were positively related for having a usual place of care, visiting a doctor, visiting an emergency room, and having delayed health care needed. Toe child-mother patterns were not statistically significant for having visited a dentist. The results presented above are consistent with the premise that both desirable and undesirable parental and child health care access/utilization pattems for Latinos/as are interrelated. That is, good health care utilization behavior from Latina mothers-such as having a usual source ofcare and visiting a doctor-are strongly related to the health care utilization behavior of their children. However, poor health care utilization and access pattems for Latina mothers- in the form of visits to the emergency room and delaying needed medica} care-are also strongly related to the health care utilization and access patterns of their children. Toe results presented here have important health policy implications because they suggest that interventions that promote good health care utilization behavior for Latina mothers spillover to their children. Thus, for example, public health insurance programs that focus on covering uninsured children but leave their parents uninsured may end up not taking full advantage of the positive spillover effects of health care access/utilization from mothers to their children. SCIIlP was created as part ofthe Balanced Budget Act of 1997 with the goal of providing health insurance coverage to uninsured children of families who did not qualify for Medicaid and at the same time could not to purchase prívate health insurance coverage. The federal goverrunent pays for almost three-fourths of the Texas SCHJP and, still, Texas has retumed money to the federal government every year since SCHIP began that could have been used to expand coverage to more families (Task Force on Access to Health Care in Texas, 2006). Toe results presented above provide further support to the idea that relaxing the eligibility rules for uninsured parents of children covered by public health insurance programs may make sense if we want to promote a better use of the health care system by Latina mothers and their children. �174 / Access and Use o/ Health Care Services by Mothers a,ul Children in the Texas-Mexico Border Region: Pre/iminary Findingsfrom the 2006 Río Grande Va/ley Health Survey References Andersen, R. and P. Davidson. (2001 ). lmproving access to care in America: Individual and contextual indicators. Cbanging the US. health care system: Key issues in health services, policy and management. R. R. Anderson, T. and G. Kominski. San Francisco, Jossey-Bass. Andersen, R. M., H. Yu, R. Wyn, P. L. Davidson, E. R. Brown and S. Teleki. (2002). "Access to medical care for low-income persons: how do communities make a difference?" Med Care Res Rev 59(4): 384-411. Bastida, E., H. S. Brown and J. A. Pagán. (2007). Health insurance coverage and health care utilization along the U.S._Mexico border: evidence from the Border Epidemiologic Study of Aging. The Health of Aging Hispanics: The Mexican-Origin Population. J. L. Angel and K. E. Whitfield. New York, NY, Springer. Greene, W. H. (2003). Econometric Analysis. Upper Saddle River, NJ, Prentice Hall. Institute ofMedicine. (2002). Care without coverage: Too little, too late. Washington, DC, National Academies Press. Institute of Medicine. (2002). Health Jnsurance is a Family Matter. Washington, DC, National Academies Press. Institute ofMedicine. (2004). lnsuring America shealth: Principies and recommendations. Washington, DC, National Academies Press. Minkovitz, C. S.; P. J. O'Campo; Y. H. Chen and H. A. Grason. (2002). "Associations between maternal and child health status and patterns of medical care use." Ambul Pediatr 2(2): 85-92. Newacheck, P. W.; J. J. Stoddard; D. C. Hughes and M. Pearl. (1998). "Health insurance and access to primary care for children." N Engl J Med 338(8): 513-9. StataCorp. (2005). Base Reference Manual. College Station, TX, Stata Press. Revista Perspectivas Sociales / Social Perspectives primavera/spring 2007. Vo/.9, Num. / / 175 Task Force on Access to Health Care in Texas. (2006). Code Red: The Critica/ Condition ofHea/th In Texas. Austin, TX, University of Texas System. U.S. Census Bureau. (2005). Small Area Health Insurance Estimates. U.S. Census Bureau. (2006). Small Area Income & Poverty Estimates. Weinick, R. M .; M. E. Weigers and J. W. Coben. (1998). "Children's health insurance, access tocare, and health status: new findings." Health Aff{Millwood) 17(2): 127-36. �Revista Perspectivas Sociales I Social Perspectives primavera/spring 2007. Vo/.9, Num. 11 Pág. 177-196 177 Diabetes and Employment Productivity: Does Diabetes Management Matter? H. Shelton Brown, ID t, José A Pagántt, Craig Hanis, t Adriana Pérez¡t* Abstract Diabetes has been shown to have a detrimental impact on employment and labor market productivity, which results in lost work days and higher mortality/disability. This study utilizes data from an ongoing diabetes-related survey of households in Brownsville, Texas, a largely Mexican American metropolitan area on the Texas-Mexico Border, in order to assess the impact of diabetes on work productivity. We focus on two questions. First, does the management of diabetes increase productivity in the short run? Diabetes management is measured by the interaction of having diabetes and glycosylated hemoglobin levels (Hbalc). Second, are women with diabetes less productive at higher levels of earnings? Methods used include ordinary least squares (OLS), quantile regression and Heckman regression. Concerning the first question, the management of diabetes \ \ 1 f University ofTexas Hea/th Science Center at Houston School ofPublic Health Division of Management, Policy, ami Community Health Schoo/ of Public Health Building Brownsville, Texas 78520 Tel. 956-882-5164; Fax 956-882-5152 E-mail: henry. s.brown@uth.tmc.edu ff University o/Texas-Pan American 1201 W University Drive Edinburg, Texas 78541 Tel. 956-318-5306; Fax 956-318-5303 E-mai/:jpagan@utpa.edu t University ofTexas Hea/th Science Center at Houston School ofPublic Hea/th Division of Epidemiology 1200 Herman Pressler Dr., RAS E401 Houston, Texas 77030 Tel. 713-500-9807 E-mail: Craig.L.Hanis@uth.tmc.edu 11 University ofTexas Health Science Center at Houston Schoo/ ofPublic Health Division ofBiostatistics School ofPub/ic Hea/th Building Brownsville. Texas 78520 Tel. 956-882-5160; Fax 956-882-5152 E-mail: adriana.m.perez@uth.tmc.edu •Financia/ supportfor this $·tudy was provided by the Hispanic Health Research Center at the Brownsville Regional Campus ofthe University o/Texas School ofPub/ic Health (N[H CMHD P20 MD000170-04). ISSN 1405-1133 O 2007 Universidad Autónoma de Nuevo León, University ofTexas ofAustin, University ofTexas of Arlington, University ofTennessee, Universidad Juárez del Estado de Durango, Universidad de Colima �178 / Diabetes anti Employment Productivity: Daes Diabetes Managemen/ Matter? Revista Penpectivas Sacia/es / Sacia/ Perspeclives primavera/spring 2007. Vo/.9, Num. 1/ does not appear to have a discernible impact on labor market outcomes in the short-run. However, diabetes does negatively affect both male and female productivity. Women with diabetes are less productive at higher wage levels. portant part of these costs are labor productivity losses (Bastida & Pagán 2002, Brown m, Pagán & Bastida 2005, Kahn 1998, Ng, Jacobs & Johnson 2001 , Lavigne, Phelps, Mushlin & Lednar 2003). Further, the US Census Bureau estimates that from 2002 to 2020 the number of individuals diagnosed with diabetes will increase by 44 percent to 17.4 million (American DiabetesAssociation 2003). With the prevalence and incidence of diabetes increasing, accurate estimates of the labor market cost of diabetes are important in order to develop appropriate health policy responses. Keywords Quantile Regression, Heckman Model, Hbalc, Labor. Resumen Se ha demostrado que la diabetes tiene un impacto negativo en el empleo y en la productividad del mercado de trabajo, resultando en días perdidos de trabajo, en mayores días de incapacidad y mayor mortalidad. Este estudio utiliza información de una muestra de hogares relacionada a la diabetes que está en progreso en Brownsville, Texas, la cual es un área metropolitana en gran parte México-Americana en la frontera de Texas con México, y que busca evaluar el impacto de la diabetes en la productividad del trabajo. Nos enfocamos en dos preguntas. Primero, ¿el manejo de la diabetes incrementa la productividad en el corto plazo? El manejo de la diabetes se mide como la interacción entre tener diabetes y los niveles de hemoglobinaglicosilada (Hbalc). Segundo, ¿son menos productivas las mujeres con diabetes a altos niveles de ingresos laborales? Los métodos utilizados incluyen Mínimos Cuadrados Ordinarios (MCO), regresiones cuantílicas y la regresión de Heckman. En relación a la primera pregunta, el manejo de la diabetes no parece tener un impacto discernible en el corto plazo sobre los resultados en el mercado de trabajo. Sin embargo, la diabetes afecta negativamente la productividad tanto de la mujer como del hombre. Las mujeres con diabetes son menos productivas a niveles salariales más altos. Palabras clave Regresión cuantilica, Modelo de Heckman, Hbalc, Trabajo. Introduction Diabetes is a disease wbich has economic implications. Toe American Diabetes Association estimates that diabetes costs the U.S. econoJDY $132 billion per year (American Diabetes Association 2003). An iJD- 179 Given the high economic costs of diabetes (American Diabetes Association 2003), public health officials have been arguing that diabetes prevention is important. Prevention could mean one oftwo things: Toe prevention of the onset of diabetes, or the prevention of diabetes-related problems through the management of glycosylated hemoglobin levels (Hbalc) for people already diagnosed with diabetes. If the productivity costs of diabetes when diabetes is managed are low, scarce prevention dollars could be concentrated on the much smaller subpopulation already diagnosed with diabetes. On the other hand, if the costs associated with diabetes are substantial, whether managed or not, then prevention dollars must be spread over the much larger general population. Of course . . ' 1t ts likely that dollars should be spent on both preventing the onset of diabetes as well as its management after onset. However, there is currently no information to inform policy-makers on how to apportion scarce prevention dollars between diabetes onset and diabetes management. Toe labor market component of the overall cost of diabetes is important. However, while overall diabetes-related costs are rising, it is not clear that per capita labor costs associated with diabetes are increasing. !echnological changes over the last three decades have led to changes m the labor market and in the medical field. First, there are increases in the number ofjobs that are less physically demanding and therefore accessible to persons with diabetes (Kahn 1998). Second, new drugs, glycosylated hemoglobin levels (Hbalc) monitoring devices and food science advances, such as artificial sweeteners, are also making diabetes ,, " . management eas1er and less costly than before. Brown, Pagán and Bastida show that the effects ofdiabetes on work �180 / Diabetes and Emp/oyment Produclivily: Does Diabetes Management Matter? propensity are more important for men than for women (Brown m et al. 2005). Anecdotally, many believe tbat women are more inclined to manage their diabetes through pbysician visits, diligent use of pharmaceuticals, use of monitoring devices and health behavior modification. This would explain why they have less diabetes-related labor market problems in comparison to roen. With the laboratory-measured Hba1e data in our study, we can test this hypothesis. This study adds two important elements to the growing literature on diabetes and labor market outcomes. First, we examine whether poor diabetes management is the cause of adverse labor market outcomes rather than diabetes per se. In our data, Hbalc levels are measured in a laboratory for all participants, whether or not they have been diagnosed with diabetes by a physician. Thus, we know the extent to whicb persons with diabetes have managed their Hbalc. Second, we examine whether diabetes affects labor productivity, whether managed or not, across wage levels using quantile regression. ...... \ \ We used microdata from tbe Diabetes Impact Project, an ongoing survey from a predominantly Mexican American area of South Texas. Our data has important advantages. Toe Diabetes lmpact Pro ject surveys Mexican Americans, a population which has a high prevalence and incidence of diabetes. Toe percentage of this population diagnosed with this health condition is expected to rise from 1.4 million in 2002 to 2.9 miIlion in 2020, a l 07 percent increase (American Diabetes Association 2003). By contrast, the total US population diagnosed with diabetes is expected to increase by 44 percent during the same time period. Our data also has laboratory-measured Hbalc levels, necessary for the assessing the level ofblood sugar management for persons with diabetes. Diabetes Self-Management and Health Capital One of our main bypotheses is that unmanaged diabetes is correlated with labor productivity. Michael Grossman argued that "healtb capital" is a measure ofhealth stock, reflective ofpast health behavior (Grossman 1972). If a personjogs or changes their diet, for instance, their health stock does not instantly change. However, jogging and dietary changes maintained over a long period increase his/her bealth stock. Further, if Revista Perspectivas Sociales I Social Perspeclives primavera/spring 2007. Yo/. 9, Num. ¡ ¡ 181 a person tempo~arily stops jogging or occasionally eats unhealthy food after a long penod of adberence, his/her health stock will not be greatly affe~ted. ?f c~urse, the Grossman model maintains that healthy behavior reqwres time mputs (Grossman 1972). He~oglobin is a protein which is found in red blood cells. Like a canary m a coal mine, hemoglobin in red blood cells revea} the Ievel of blood glucose ov~r a two to three month period. Toe greater the percentage ~f sugar that 1s the blood supply, the higher the percentage ofhemoglob_m has been glycosylated. Note that the percentage ofhemoglobin ~atls glycosylated is invariant to day-to-day blood sugar Ievels (Woerle, Punenta, Meyer, Gosmanov, Szoke, Szombathy Mitrakou & Gericb 2004). Thus, it is an excellent measure of health ;tock in the Grossman sense (Grossman 1972). The American Diabetes Association defines type 2 diabetes as manª?ed based o~ glycosylated hemoglobin levels (Hbalc) (http://www. diabetes.org/diabetes-research/summaries/woerle-ogtt.jsp). Hba 1c Ievels ?f seven percent or less indicates that the person with diabetes is managmg bis or her diabetes. Among persons without diabetes, lower Hbalc levels should not be con:elated with_labo~ market success. On the other hand, among persons diagnosed w1th diabetes, labor productivity for persons with Iow Hbalc leve!s should be higher than for persons with high Hbalc levels. ~ altematJve hypothesis is that time and effort expended on managing ~abetes may come at the expense of current, short-run labor productiv~ty. Thus, it may be that e:fforts to manage diabetes lower productivity ID the short-term. Our strat~gy to t~st the hypothesized association oflow wages with unmanaged diabetes mvolves creating an interaction term between diabetes an~ Hbalc levels. For persons without diabetes, the interaction term value 1s zero. For person with diabetes, the interaction term will be their ~ale. !herefore, those who bave managed their diabetes will have an :t~c_non term_value ofseven or lower, those who have not managed etr diabetes will have an interaction term value of greater than seven. We expect to find a negatively relationship between this interaction term and log wage. �182 / Diabetes and Employment Productivity: Does Diabetes Management Matter? Data and Methods Toe data for this study comes from an ongoing survey in Brownsville, Texas, a metropolitan area with a total population of 139,722 in 2002 that is located in the US-Mex.ico border region (U.S. Census Bureau 2006). Brownsville is the most southernmost city in Texas and 91.3% of its population is ofHispanic origin (U.S. Census Bureau 2006). Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vo/.9, Num. / I 183 sample selection bias, we also estimate the log wage equation using the method of Heckman. Following Greene ( 1990), let z = 1 when the wage is known, z = Ootherwise. Toen, a probit model z; = e,~+ Diab; X Hbalca1 + U¡, U~ N[O, l], z = 1 if z• > O, (1) (2) z = o if z.. ~ o, Brownsville, as many other Texas border cities, is characterized by high poverty levels and low educational attainment. For example, 36.5% of people in Cameron County, where Brownsville is located, live below the poverty level (which ranks it next to last in the country). Only 60% of the population 25 years and over have completed high school (which ranks it last in the country). Approximately 46.5% ofchildren under 18 years are below the poverty level, which ranks it next to last in the country (http: //www.census.gov/acs/www/Products/Ranking/2003/RO 1T050 .htm). Even though median household incomes in Brownsville are among the very lowest in the country, there is variation. Therefore, the Diabetes Impact Study selected a representative sample based on income for the Mex.ican-American population of Brownsville. To do so, we selected census blocks from census tracts with median household incomes in the first and the third quartiles. 1 Toe 2000 United States Census of Population and Housing was used to select the probability sampling frame. A multi-stage cluster sample of participants is being collected in these two Brownsville area locations and it is expected to be complete by the beginning of 2008. Within all randomly selected clusters (census blocks), all the households in the census block are contacted. A participant, between 35 to 64 years old, from each household is randomly selected using a ten digit permutation algorithm. Heckman Model Brown m, Pagán & Bastida (2005) show that especially for MexicanAmerican men, diabetes a:ffects work propensity. For persons not working the wage is missing due to selection. Therefore, in order to account for 1Note that households selectedwithin the census tracts may have differing income /eve/s. Prob[z = 1] = ~(C:6 + Dial>; x Hbalcn1), Prob[z = º'= ~«.s + 1- Diab, X Hbalcn1) estimates the probability that wage is observed. In (1), ci is the vector of exogenous variables related to observing wage and iS is the corresponding vector of coefficients. Diab¡ x Hba le; is the interaction term and 'Y1 is the associated coefficient. Note that diabetes may be related to whether the person earns a wage or not because he or she may select out of the market. Toe Heckman model is a two-stage model, where equation (1) is the first stage. Toen, the log wage equation is estimated in the second equation as, Log Wage = ~jJ + Diab; x Hbalc,-"f + e;, observed only if z = 1, (u,e) (3) ~biw.riate normal[O, O, l ,u. ,p] where Pis the correlation between e; and u ¡ and q e is the variance of the disturbance (1). Vector X; is the exogenous variables related to wage and fJ is the corresponding vector of coefficients. It is shown in Greene ( 1990) and elsewhere that E(Log wagelz. = IJ= z/J + Diab; X Hbal.cy + po-,Á(C:ó + Diab; X Hbalca1), ( 4) where ,\ = 'P(C:-6 +Diabn1)/cl>(uf,-6+ Diaba1) is the inverse Milis ratio. ef> is the marginal probability from the normal distribution and ~ cumulative �184 / Revista Perspectivas &x:iales I &x:ial Perspectíves primaveralspring 2007. Vol.9. Num. JI Diabetes aruJ Emp/oymenl Productivity: Does Diabetes Management Matter? probability from the normal distribution. Table 1: Descriptive statistics, by gender Quantile Regression Variable n Log hourly wage 121 Employed (l=yes; O=no) 205 Age 205 Age squared 205 Bom in Mexico (1 =yes; O=no) 205 Years of schooling 205 Years residing in Brownsville (US) 205 Married (1 =yes; 0=no) 205 Log of other household income 205 DiabeticxHbal c 190 Diabetic (1 =yes; O=no) 205 ''New" Diabetic (l=yes; O=no) 205 Mean 2.119 .722 47.195 2,318 .659 10.307 21.868 .600 2.871 1.367 .166 .044 Variable Log hourly wage Employed (l=yes; 0=no) Age Age squared Bom in Mexico (l=yes; O=no) Years of schooling Years residing in Brownsville (US) Married (1 =yes; O=no) Log of other household income DiabeticxHbal e Diabetic (1 =yes; 0=no) ''New" Diabetic (l=yes; 0=no) Women Mean S.D. 1.947 .479 .482 .500 47.256 8.934 2,313 868 .721 .449 9.028 4.218 21.811 15.851 .517 .500 5.164 4.562 1.336 3.647 .136 .343 .041 .200 We estimate log wage quantile regressions for the following equation, Log wage = rJJ + Diab; X Hbalcy + e;, (5) where Diab¡ is the dummy variable defined above and Diab; x Hbalc¡ is the interaction term (Koenker & Hallock 2001). Vector X; is the exogenous variables related to wage and /3 is the corresponding vector of coefficients, as defined in (3). Other versions of (5) that will be estinlated include the diabetes dummy variable without the interaction term. Quantile regression is useful for two reasons. First, diabetes and managed diabetes may affect productivity differently at different levels of the wage distribution. Second, our data is highly compressed at the low end of the wage distribution and wages may be constrained from falling further due to diabetes. \ We estimate the simultaneous quantile regression with the 'sqreg' function in Stata Version 8 (Stata 2003). Confidence intervals are constructed at each quantile via boot-strapping. The quantiles selected are from .05 to .95, in steps of 0.05. Thus, 19 simultaneous quantile regressions were estimated. Results Table 1 reports the descriptive statistics of the variables used in the OLS, Heckman, and quantile regressions by gender. Note that women are less likely to be employed and have n 177 434 434 434 434 434 434 433 433 408 434 434 185 Men S.D. Min Max .523 1 4 .449 o 1 9.564 21 64 919 4414,096 .475 o 1 4.302 o 20 17.399 o 63 .491 o 1 4.338 o 13 3.577 o 16 .373 o 1 .205 o 1 Min Max 1 4 o 1 35 64 1225 4,096 o 1 o 20 o 64 o 1 o 12 o 19 o 1 o 1 slightly less schooling than meo, and are more likely to be immigrants. Men are also paid more than women. Approximately four percent of women and men not previously diagnosed with diabetes had fasting glucose levels above 126 mg/ dl. This is the variable labeled "New Diabetic." Although a second reading would be necessary to officially diagnose diabetes, the vast �186 / Diabetes and Emp/oyment Productívity: Does Diabetes Management Matter? Revista Perspectivas Sociales / Social PeT$JH!Clives primavera/spring 2007. Vo/.9, Num. J / ma jority ofthem likely have diabetes and have therefore been diagnosed for the first time by the medical personnel in our survey. is related to working and productivity in a non-linear fashion. Being bom in Mexico has a negative effect on wages. Toe variable Table 2: Self-management ofDiabetes, by gender Table 3: OLS model oflog wage, Men n Var Hbalc Bloodsugar 54 59 Hbalc Blood sugar 332.5 26 34 S.D. Women 10.1 3.5 200.98 82.04 Men 2.7 9.98 73.7 189.02 Mean Min Max 4.7 18.9 74.3 434 5.8 16.2 84.l Of people with diabetes, women are no more likely to have their Hba 1e managed in comparison tomen. However, neither gender appears to be managing their diabetes effectively. Based on our previous research (Brown ID et al. 2005), we also estimated the correlation between the interaction term and our instruments (family history of diabetes) using a tobit model (not reported). However unlike the correlation between diabetes status and family his' tory of diabetes, the correlation between the interaction term and family history of diabetes was weak.2 This is likely because although genetics explains much of the variation of whether a person develops diabetes, diabetes-related genetics does not explain whether a person with diabetes is able to have it managed. Dueto the weak correlation between our instruments and our interaction term, we will not use instrumental variable techniques to examine diabetes management. Variable Coefficient Std. Err. Coefficient Log hourly wage Age - 0.105 ** 0.051 -0.107 ** Agesquared 0.001 ** 0.001 0.001 ** Boro in Mexico ( l =yes; O=no) -0.279 ** 0.137 - 0.260 ** Years of schooling 0.031 ** 0.013 0.033 ** Years residing in Brownsville {US)-0.004 0.004 -0.003 Married (1 =yes; O=no) 0.181 * 0.101 0.1 70 * DiabeticxHbalc -0.019 0.018 Diabetic (1 =yes; 0=no) -0.278 * Constant 4.169 *** 1.176 4.207 *** Significance levels: *: 10% **: 5% ***: 1% 187 Std. Err. 0.049 0.001 0.126 0.012 0.004 0.096 0.151 1.152 'years of schooling' is positively related to wage. Being married is positively related to wage at the 10% level. Finally, the variable 'years in Brownsville' is negatively related to log wage. For most people, the number ofyears residing in Brownsville is synonymous with the number of years in the United States. The negative sign likely re:flects the fact that the border area is a temporary place of residence for many Mexican Americans. Many who develop language and other work sk:ills move away from the border to places in the U.S. OLS and Heckman Models Table 3 reports the estimates for the OLS regression formen, assuming that diabetes is exogenous. Let us first consider the model in the leftmost columns, where diabetes management is considered. Age and age squared 2 The results are available upon request. Diabetes management (the interaction term, DiabeticxHbalc) coefficient is negatively related to log wage, but the association is not statistically significant at this level ofpower. However, this suggests that diabetes has a detrimental e:ffect on productivity, even when managed. Let us now consider the model in the rightmost columns, where diabetes per se is considered. Diabetes is negatively related to log wages. Toe �188 / Diabetes an,J Employment Producffvity: Daes Diabetes Management Matter? wage premium is quite high-having diabetes lowers wages by 28%. Table 4: Heckman model oflog wage, Meo In order to take sample selection into account, let us now turn our attention to the Heckman model estimates in Table 4. Age and age squared are now not significantly related to log wage. In both the selection equation and the wage equation, the diabetes Hbal e interaction term is significantly and negatively related to working and negatively related to log wage. Further, the magnitude of the effect ofthe diabetes Hbalc interaction term is approximately twice the effect when sample selection is not taken into account. For each unit increase of Hbalc for persons with diabetes, log wage declines by four percentage points. Variable Even though diabetes management is negatively associated with productivity, diabetes status has a stronger association. After accounting for sample selection, the negative effect of diabetes on productivity increases by approximately 50%, from 28% to 42%. Note that diabetes is negatively related to earning a positive wage. Being married is positively related to earning a positive wage; Log of other household income is negatively related to eaming a positive wage. \ Table 4 also includes a likelihood ratio test of the covariance between ei and u;, defined as pin equation (3). In each specification of the model, the error terms are correlated at the 10% level. This indicates a bivariate normal distribution of the sample selection and log wage equations. Therefore, the Heckman estimates are warranted. Table 5 reports the estimates for the OLS regression for women, assuming that diabetes is exogenous. Only the variables 'years of schooling' 'years in Brownsville' are positively related to wage. Being married is positively related to wage at the 10% level. Diabetes management (the interaction tenn, DiabeticxHbalc) coefficient is negatively related to wage, but not statistically significant. This suggests that diabetes has a detrimental effect on productivity, even when managed. Toe results are similar to those for men. Diabetes per se 189 Revista Perspectivas Sociales/ Social Perspectives primaveralspring 2007. Vol.9, Num. J/ Coefficient Std. Err. 1 Coefficient Log hourly wage -0.069 0.056 - 0.081 0.001 0.001 0.001 Std Err. Age 0.055 Agesquared 0.001 Born in Mexico (l=yes; O=no) -0.269 * 0.146 -0.282 ** 0.135 Years of schooling 0.027 ** 0.014 0.028 ** 0.013 Years residing in Brownsville (US) -0.003 0.004 - 0.003 0.004 Married (!=yes; O=no) 0.291 *** 0.108 0.289 *** 0.104 DiabeticxHbalc - 0.040 ** 0.018 Diabetic (1 =yes; O=no) *** - 0.422 0.158 Constant 3.193 ** 1.316 3.506 *** 1.278 Wage Observed (1 =yes; O=no) Age 0.174 0.118 0.146 0.117 Agesquared -0.002 * 0.001 - 0.002 0.001 Born in Mexico (l=yes; O=no) -0.197 0.279 -0.290 0.266 Years of schooling 0.014 0.026 0.007 0.025 Years residing in Brownsville (US) -0.004 0.008 -0.006 0.008 Married (l=yes; O=no) 0.751 *** 0.206 0.805 *** 0.201 Logof other household income -0.075 *** 0.023 - 0.079 *** 0.022 DiabeticxHba le -0.074 ** 0.030 Diabetic (l=yes; O=no) -0.638 ** 0.280 Constant -3.251 2.749 -2.455 2.725 Ho: P =O (X2 (1) = 3.08, Prob = 0.079 X2 (1) =3.27, Prob = 0.071 Significance levels: *: 10% **: 5% ***: 1% �190 / Diabetes ami Employment Productivity: Does Diabetes Management Maller? Revista Perspectivas Sociales I Social Perspectives primaveralspring 2007. Vol.9. Num. J / Table 5: OLS model oflog wage, Women Table 6: Heckman model oflog wage, Women l Coe:fficient Std. Err. Coefficient Std. Err. Log hourly wage 0.030 0.050 Age 0.011 0.051 0.000 0.001 Age squared 0.000 0.001 0.083 Boro in Mexico (l==yes; O=no) - 0.071 0.085 - 0.064 0.024 *** 0.009 Years of schooling 0.020 ** 0.009 0.009 *** 0.003 Years residing in Brownsville (US) 0.009 *** 0.003 0.067 0.068 Married (l==yes; O=no) 0.091 0.069 DiabeticxHbalc - 0.014 0.009 - 0.160 0.099 Diabetic (1 ==yes; O=no) 0.828 1.165 Constant 1.284 1.181 Significance levels: *: 10% **: 5% ***: 1% Variable is negatively related to log wages, but not at any statistically significant level. Toe wage premium is much lower than for women. Table 6 shows the Heckman results for women. The likelihood ratio test of whether p , which is the correlation between e¡ and u¡ in (3), is significantly different from zero does not revea! a difference. Therefore, the results for the wage equation are again equivalent to OLS estimates with no selection. Quantile Regression In order to consider the effect of diabetes management on productivity, we estimate log wage quantile regressions. Figures 1, 2, and 3 display the results. For each (A) figure (on the left), the line represents the diabetes coe:fficient estímate and its 95% confidence interval for each quantile of log wage. These are from equation (5). The dotted line is the zero coe:fficient. For each (B) figure (on the right), the line represents the diabetes Hbalc interaction coefficient estímate and its 95% confidence interval for each quantile of log wage. These are from equation (5). Note that the other regressor are the same as above (Constant, Age, Age squared, Boro in Mexico (1 ==yes; O=no), Years of schooling, Years residing in Brownsville (US), Married (l=yes; O==no)). 191 Variable Coefficient Std. Err. Coefficient Std. Err. Log hourly wage Age 0.004 0.050 0.025 0.050 Age squared 0.000 0.001 0.000 0.001 Boro in Mexico {l=yes; O=no) -0.078 0.084 -0.068 0.081 Years of schooling 0.019 ** 0.009 0.023 *** 0.008 Years residing in Brownsville (US) 0.009 *** 0.003 0.009 *** 0.003 Married (1 =yes; O=no) 0.100 0.068 0.075 0.068 DiabeticxHbalc - 0.014 0.009 Diabetic (1 =yes; O=no) - 0.163 * 0.097 Constant 1.511 1.194 0.996 1.182 Wage Observed (l=yes; O=no) Age 0.143 0.090 0.151 * 0.088 Age squared - 0.002 * 0.001 -0.002 * 0.001 Boro in Mexico (l=yes; O=no) 0.051 0.172 0.042 0.165 Years of schooling 0.024 O.O 18 0.026 0.017 Years residing in Brownsville (US) 0.002 0.005 0.001 0.005 Married (l=yes; O=no) 0.185 0.141 0.130 0.137 Log ofother household income -0.109 *** 0.015 -0.108 *** 0.015 DiabeticxHbalc 0.012 0.019 Diabetic {l=yes; O=no) 0.205 0.193 Constant -3 .104 2.127 -3.320 2.081 X2 (1) =0.52 Prob= 0.4704 (X2 (1) = 0.27 Prob = 0.6064 Significance Ievels: *: 10% **: 5% ***: l % Let us first consider the effects of diabetes on men and women combined, as shown in column A of Figure 1. The results show that the detrimental effects of diabetes are higher at the higher wage quantiles. This could indicate that 'left wall' effects at the low end of the wage because they are already at the lowest possible wage for their profession. As shown in column B of Figure 1, when the interaction term is included rather than the diabetes dummy, most ofthe detrimental effects �192 / Diabetes ami Employmenl Productivity: Does Diabetes Management Matter? Revista Perspectivas Sociales/Social Perspectives primaveralspring 2007. l'o/.9, Num. J / on labor productivity disappear. low end of the wage distribution may be less productive than persons distribution. In reality, persons with diabetes (managed or not) at the without diabetes, but this <loes not translate to wages Figure 2: Women: (A) Diabetes (B) Diabetes xHbalc This indicates that persons with diabetes who have managed their diabetes are no more productive than persons with diabetes who have not managed their diabetes. Note that when the diabetes dummy and the interaction term are both included, neither affects diabetes at any portion of the log wage distribution. i ó ! ~ : !"! 1 ~ ~ 193 o .. . - .. ; Figure 3: Men: (A) Diabetes (B) Diabetes xHbalc - Figure 1: Men & Women: (A) Diabetes (B) DiabetesxHbalc ~ 2 & § § i,ó I; h'I g 'I i 'I .. A - - ~ - a.s g~ .. ~ ; .. . .. ~ ,. ;¡ ~ 2 °'"""" Let us now consider women, as shown in Figure 2. Column A of Figure 2 illustrates 1 the estimates of diabetes on labor productivity. 1bis indicates that the detrimental effects of diabetes are higher in the higher wage quantiles. However, when Hbal e management is taken into account, most of the detrimental effects of diabetes on labor productivity disappear. Note that when the diabetes dummy and the interaction term are both included, neither affects diabetes at any portion of the log wage distribution. Formen, as shown in Figure 3, diabetes has a large effect on productivity, whether or not we include the diabetes dummy or the interaction term. However, the effects do not differ by the quantile of the wage. 4 Discussion and Conclusion Our results reported above are inconsistent with previous research suggesting that diabetes mainly affects the labor market of men, but not women (Bastida & Pagán 2002, Brown III et al. 2005). Ourresults using quantile regression revea] that at the high end ofthe wage distribution, the �194 / Diabetes anti Emp/oyment Productivity: Does Diabetes Managemenl Matter? Revista Penpeclivas Sociales I Social Penpectives primaveralspring 2007. Jlo/.9, Num. J/ productivity of females with diabetes is lower than for females without diabetes. This is possibly due to tbe fact that women with diabetes earn wages near the left-wall of the wage distribution. References Our results for Hispanics also are higher quantitatively than previous research. Formales, the wage productivity premium for avoiding diabetes is approximately 42%. For tbose witb diabetes, there is an approximately four percentage point wage premium for every unit ofHbalc increase. Our results suggest that in order to avoid productivity losses for males associated with diabetes, scarce prevention resources should be spent on the prevention ofonset of diabetes rather than tbe management ofHbal e for those already diagnosed with diabetes. 195 American Diabetes Association (2003). 'Economic costs of diabetes in the U.S. in 2002 ', Diabetes Care 26(3), 917-932. Bastida, E. & Pagán, J. A (2002), 'The impact of diabetes on adult employment and eamings ofMexican-Americans: Findings from a community based study', Hea/th Economics 11 (5), 403-13. Brown III, H. S., Pagán, J. A. & Bastida, E. (2005), 'Toe impact of diabetes on employment: Genetic IVs in a bivariate probit', Health Economics 14(5), 537-544. Unfortunately, this means resources must be spread among populations most predisposed to diabetes, sucb as Hispanics, rather than concentrated on the smaller group with diabetes. This is not to say that management of diabetes will not improve quality of life or prevent medica! costs in the future. Productivity in the future will likely be higher for those who currently are managing their diabetes. Greene, W. H. (1990). Econometric Analysis, Macmillan Publishing, NewYork. Our results on self-management of diabetes are valid in the short-run only. The Grossman model suggests that bealth capital leads to longrun productivity gains. However, we only compare short-run effects of diabetes management with cross-section data. Kahn, M. (1998). 'Health and labor market performance: Toe case of diabetes', Journal ofLabor Economics 16(4), 878-899. Grossman, M. (1972). The Demandfor Health: A Theoretical and Empirical lnvestigation, Coumbia University Press (National Bureau for Economic Research). Koenker, R. & Hallock, K . F. (2001 ). 'Quantile regression', Journal of Economic Perspectives 15(4), 143- 156. Lavigne, J. E.; Phelps, C. E., Mushlin, A. & Lednar, W. M. (2003); 'Reductions in individual work productivity associated with Type 2 diabetes mellitus', Pharma-coeconomics 21(15), 1123- 1134. Ng, Y. C.; Jacobs, P.; & Johnson, J. A. (2001). 'Productivity losses associated with diabetes in the U.S. ', Diabetes Care 24(2), 257-261. Stata (2003). Stata Reference Manual [computer program]. Version 8, 1-4 edn, StataPress. U.S. Census Bureau (2006), American Factfinder, Technical report, U.S. Census Bureau. generated by Shelton Brown. �196 / Diabetes and Emp/oyment Productivity: Does Diabetes Managemenl Matter? Woerle, H. J.; Pimenta, W. P.; Meyer, C., Gosmanov; N. R., Szoke; E., Szombathy, T., Mitrakou, A. y Gerich, J. E. (2004). 'Diagnostic and Therapeutic Implications of Relationships Between Fasting, 2-Hour Postchallenge Plasma Glucose and Hemoglobin A le Values', Arch lntern Med 164(15), 1627- 1632. GUIDELINES FOR CONTRIBUTORSNORMAS DE PRESENTACIÓN DE ARTÍCULOS �Revista Perspectivas Sociales / Social Perspectives primavero/spring 2()()7. Vol.9, Num. J I 199 GUIDELINES FOR CONTRIBUTORS OF 'SOCIAL PERSPECTIVAS/PERSPECTIVAS SOCIALES' Social Perspectives/Perspectivas Sociales, a bilingual, bi-nationaljournal, is seeking manuscripts to be published in 2008. Toe journal is a joint project ofthe Facultad de Trabajo Social and the Facultad de Economía of the Universidad Autónoma de Nuevo Leon, Mexico, the School ofSocial Work at the University of Texas at Austin and The University of Texas atArlington, the School ofSocial Work at the University ofTennessee and the Facultad de Trabajo Social de la Universidad Juárez del Estado de Durango. We are seeking papers that focus on issues connected to the U.S.-Mexico border and the persons moving in both directions across that border, social work practice issues that are common to individuals, families and communities in both nations, social policy issues that are common to both nations, social work education relevant, and research on social conditions. Papers describing innovative practices, empirical research, policy and program developments are welcomed. Interdisciplinary and intemational papers are encouraged. Contributions should adjust to the following rules: 1. Contributions submitted must be original and should not be under consideration in any other journal 2. Contributions should be submitted electronically to one of the following e-mail addresses: veronikasieglin@yahoo.de; lorikay@ mail.utexas.edu; or to mramor@facts.uanl.mx. Toe journal prefers Microsoft Word for Windows. 3. Author(s)' information (author(s)' name, academic degree, affiliation including telephone, postal address and e-mail address) should be typed on a separate sheet. 4. All papers deemed appropriate for the joumal are sent out anonymously to two referees of the Scientific Board of the journal thatconsists ofan intemational panel. Contributions will be published only if they are accepted by the referees. 5. Contributions will be published in English or Spanish. Papers should not be shorter than 1Opages and longer than 30 pages. They should be typed 1.5 spaced, Times New Roman. Avoid fancy typefaces. �Revista PerspecJivas Sociales / Social Perspectives primavera/spring 2007. lk>l.9, Num. / / 200 6. 7. 8. 9. Use cursive type font only for foreign words. Toe typical manuscript is about 20 pages including references, and abstract (300-350 words) in English and Spanish. Include a brief abstract (300-350 words) summarizing the findings and fi.ve key words. Textual quotations should use quotation marks instead of cursive letters. Please indent any citations in the body of the text that are longer than four lines as a block quotation; give them a deeper indent than the rest of your text. References must be presented at the end of the paper in a separate References section as followed: author (family name, name), (year), title, place, editorial. Examples: a) Books Edelman, P.; H.J. Holzer, and P. Offner (Eds.). (2006). Reconnecting disadvantaged young men, Washington DC, Urban Institute Press. Russo, R.J. (1980). "State problems and the need for research-based planning in the drug fi.eld", In C. Akins, and G. Beschner (Eds.), Ethnography: A research tool for policymakers in the drug and alcohol fields, Rockville, Maryland, U.S. Department of Health and Human Services, pp. 40-65. \ b) Journal Articles: Murray, D.M.; R.V. Luepker, C.A. Johnson, andM.B. Mittelmark(1984). "Toe prevention of cigarette smoking in children: A comparison of four strategies", Joumal ofApplied Psychology, 14(3), pp. 274-288. c) Inforrnation from web-sides: McBride, D. C.; C.J VanderWaal, Y. M. Terry, and H. VanBuren (1999). Breaking the Cycle ofdrug use amongjuvenile offenders [On-line]. Retrieved October 24, 2002, from http://www.ncjrs.org/pdffiles 1/179273. pdf 10. Write sentences and paragraphs clearly and succinctly with a 201 minimum of jargon. Writing should demonstrate theoretical soundness and scientific accuracy. 11. Authors will be notified after the reviewers return their comments to the editors. Toe results of the reviewers may be in any of these terms: publishable as it is, not publishable, or publishable with commendations and/or modifications. 12. Toe articles published in Perspectivas Sociales/Social Perspectives may be distributed in any press or electronic format that the editorial committee considers pertinent. �202 NORMAS DE PRESENTACIÓN DE COLABORACIONES PARA 'PERSPECTIVAS SOCIALES/ SOCIAL PERSPECTIVES' Perspectivas Sociales/Social Perspectives es una revista bilingüe y binacional que invita a trabajadores sociales y científicos sociales a someter manuscritos para ser editados durante el 2008. La revista constituye un proyecto conjunto de la Facultad de Trabajo Social y de la Facultad de Economía de la Universidad Autónoma de Nuevo León, México, Toe School ofSocial Work at the University ofTexas atAustin y Toe University of Texas at Arlington, the School of Social Work of the University ofTennessee y de la Facultad de Trabajo Social de la Universidad Juárez del Estado de Durango. Buscamos artículos que se enfocan en temas relacionadas con la frontera México-Estados Unidos y las personas que se desplazan en ambas direcciones; tópicos de importancia p~ra la práctica del trabajo social en torno a individuo, familia y comumdad; las políticas sociales, la formación del trabajo social; e investigacio~es científico-sociales acerca de las condiciones sociales. Se da una especial bienvenida a trabajos que analizan prácticas innovadoras, presentan resultados de estudios empíricos y que revisan críticamente políticas Y programas de desarrollo social. Se alienta asimismo a trabajos interdisciplinarios e internacionales. \ \ t Los artículos deben ajustarse a las siguientes normas de presentación de originales: 1. Los documentos deberán ser versiones definitivas e inéditas. 2. Los trabajos se enviarán por correo electrónico en formato Microsoft® Word a alguna de las siguientes direcciones electrónicas veronikasieglin@yahoo.de ó lorikay@mail.utexas.edu . 3. Deberá enviarse, en un documento anexo llamado "Datos del autor", la siguiente información: nombre completo, grado universitario máximo, institución donde labora, cargo actual que desempeña, número telefónico, dirección postal, dirección electrónica En el caso de coautorías deberán indicarse los datos de todos los colaboradores. 4. Las colaboraciones serán evaluadas por la dirección de la revista para verificar que se ajusten a las presentes normas. De ser así, serán enviadas a dos dictaminadores miembros del Comité Científico de Revista P=¡,ectivas Sociales / Social Perspectives primaveralspring 2007. Vol. 9, Num. 1 / 5. 6. 7. 8. 9. 203 · la revista, cuyo arbitraje favorable es requisito indispensable para la publicación del trabajo. Los artículos se publican en inglés o español con un resumen en ambos idiomas. Los manuscritos deben tener como extensión mínima 1Opáginas y máximo 30, en fuente Times New Roman, interlineado de 1.5, sin macros ni viñetas de adorno, sin hacer énfasis con fuentes tipográficas, y utilizando cursivas sólo para voces extrajeras. El manuscrito típico tiene alrededor de 20 páginas incluido el resumen (300-350 palabras), y la bibliografia. Los artículos iniciarán con un resumen (300-3 50 palabras) e incluirán cinco palabras clave. Las citas textuales se consignarán entre comillas, no mediante cursivas. Cuando se trate de citas breves, se mantendrán dentro del párrafo en que se produzca la referencia; si la cita rebasa las cuatro líneas, se colocará a bando, con márgenes más amplios, a un espacio y sin entrecomillado. La bibliografía irá al final del artículo en este orden: autor (apellidos, nombre), año (entre paréntesis), obra (en cursiva), lugar de edición, editorial. Ejemplos. a) Libros Bauman, Zygmunt (2002). La ambivalencia de la modernidad y otras conversaciones; Barcelona, Paidós. Adelantado, José, José Antonio Noguera y Xavier Rambla (2000). "El marco de análisis: las relaciones complejas entre estructura social y políticas sociales", en José Adelantado (coord.), Cambios en el Estado de Bienestar, Barcelona, Editorial Icaria, pp. 23-60. b) Revistas: Boltvinik, Julio (octubre 2001 ). "Opciones metodológicas para medir la pobreza en México", en Revista Comercio Exterior, vol. 51, núm. 10, México DF, pp. 869-878. e) Sitios de Internet: �204 Cámara Nacional de la Industria Tequilera (2004). Jnforme de la Cámara Nacional de la Industria Tequilera sobre su comportamiento durante el año de 2005, México. Disponible en: http:J/www.camaratequilera.com. mx/ (Recuperado el 19/10/06). 1O. Respetando el estilo de cada escritor, sugerimos redactar los textos a través de construcciones sintácticas sencillas, párrafos preferentemente breves y articulación entre profundidad teórica, rigor científico y claridad expositiva. 11. Una vez emitidas las evaluaciones de los árbitros consultados, se comunicará al autor los resultados del dictamen en cualquiera de los términos siguientes: se publica, no se publica o se publica con las recomendaciones o modificaciones que se consideraron pertinentes. 12. Los artículos publicados en Social Perspectives/Perspectivas Sociales serán difundidos y distribuidos por todos los medios impresos y/o electrónicos que el Comité Editorial de la revista juzgue convenientes. FORMATO DE SUSCRIPCIÓN SUBSCRIPTION FORM �REVISTA PERSPECTIVAS SOCIALES/SOCIAL PERSPECTIVES FORMULARIO DE SUSCRIPCIÓN/SUSCRIPTION (anual) PRECIO* República Mexicana D Institucional D Individual $150.00 $100.00 Estados Unidos D Institucional D Individual $25 U.S Dollars $20 U.S. Dollars DIRECCION/ADDRESS Nombre/Name .___ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _____, Calle/Street Ciudad/City .___ _ _ _ _ _ ___, \ Estado/State l._______, País/Country .___ _ _ _ _ _ ___. C.P./ZIP Teléfono/Phone \¡ Correo electrónico/E-mail • Adjunto mi cheque certificado o giro postal a nombre de la Facultad de Trabajo Social de la Universidad ~ - - - - - - ~ Autónoma de Nuevo León por la cantidad de$ L--- - - - - - - - ' • 1 send check payable to Facultad de Trabajo Social, Universidad Autónoma de Nuevo León for $ ~------~ * Para envío al interior de la República, se deberá agregar 160.00 pesos por gastos de envío (cubre /os dos números). En caso de envíos fuera de la República se necesitará calcular /os gastos según el país. ��� Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Perspectivas Sociales = social perspectives Description An account of the resource Revista de Ciencias Sociales de la Facultad de Trabajo Social, publicada en la década de los noventa. Editada por Emma Adame W., Luis Leñero O, Ma. Del Carmen Elú, Víctor Zúñiga y Rodolfo Garza. Contiene artículos de investigación científica sobre educación, trabajo social, ciencias sociales, psicología, etcétera. Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Título Uniforme Perspectivas Sociales = social perspectives Año de publicación El año cuando se publico 2007 Volumen Volumen de la revista 9 Número Número de la revista 1 Mes de publicación Mes cuando se publicó Primavera Día Día del mes de la publicación 1 Periodicidad La periodicidad de la publicación (diaria, semanal, mensual, anual) Semestral Relación OPAC https://www.codice.uanl.mx/RegistroBibliografico/InformacionBibliografica?from=BusquedaAvanzada&bibId=1753308&biblioteca=0&fb=20000&fm=6&isbn= Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Perspectivas Sociales : Social Perspectives, 2007, Vol 9, No 1, Primavera Creator An entity primarily responsible for making the resource Cárdenas G. de Silva, Rosa María Subject The topic of the resource Ciencias Sociales Investigación Trabajo social Educación Psicología social Description An account of the resource Revista de Ciencias Sociales de la Facultad de Trabajo Social, publicada en la década de los noventa. Editada por Emma Adame W., Luis Leñero O, Ma. Del Carmen Elú, Víctor Zúñiga y Rodolfo Garza. Contiene artículos de investigación científica sobre educación, trabajo social, ciencias sociales, psicología, etcétera. Publisher An entity responsible for making the resource available Universidad de Nuevo León, Facultad de Trabajo Social ; Austin Texas, School of social wook, University of Texas at Austin Contributor An entity responsible for making contributions to the resource Sieglin, Veronika, Editora Holleran, Lori, Editor Date A point or period of time associated with an event in the lifecycle of the resource 01/04/2007 Type The nature or genre of the resource Revista Format The file format, physical medium, or dimensions of the resource tex/pdf Identifier An unambiguous reference to the resource within a given context 2016264 Source A related resource from which the described resource is derived Fondo Universitario Language A language of the resource spa/eng Spatial Coverage Spatial characteristics of the resource. Monterrey, N.L., (México) Access Rights Information about who can access the resource or an indication of its security status. Access Rights may include information regarding access or restrictions based on privacy, security, or other policies. Universidad Autónoma de Nuevo León Rights Holder A person or organization owning or managing rights over the resource. El diseño y los contenidos de La hemeroteca Digital UANL están protegidos por la Ley de derechos de autor, Cap. III. De dominio público. Art. 152. Las obras del dominio público pueden ser libremente utilizadas por cualquier persona, con la sola restricción de respetar los derechos morales de los respectivos autores Alimentos Diabetes Madres México Niños Oportunidades educativas Pobres Reformas fiscales Sector informal Servicios de salud Sesgo de género