Non-Contributory Health Insurance and Household Labor Supply: Evidence from Mexico

Transcription

1 Non-Contrbutory Health Insurance and Household Labor Supply: Evdence from Mexco Gabrella Cont UCL and IFS Rta Gnja Uppsala Unversty Renata Narta Unversty of São Paulo July 14, 2016 PRELIMINARY AND INCOMPLETE. DO NOT CITE WITHOUT PERMISSION. Abstract Mexco ntroduced n 2002 a non-contrbutory health nsurance scheme drected to the half of the country s populaton whch was uncovered by Socal Securty protecton or employer provded health nsurance (the Seguro Popular, SP). SP s a transfer to the nformal sector workers and the nonemployed and a tax to the formal sector workers, whch may alter the ncentves for ndvduals to partcpate n the labor market or n whch sector to work (formal or nformal). The mplementaton of SP n a muncpalty s assocated wth an ncrease n the probablty of loss of Socal Securty protecton by low educated households wth chldren of 2-3%, and a decrease n the salares pad n the nformal sector. To understand the mechansms behnd the program mpact, we buld a household search model of wage determnaton whch ncorporates the valuatons of beng n the nformal sector or nonemployed relatve to the formal sector. The model s estmated usng the Mexcan Labor Force Survey on the perod before the ntroducton of SP. The model s able to replcate (1) the stocks of household types accordng to ther Socal Securty coverage and (2) the transtons n and out of employment and between formal and nformal jobs found n the data. We then use the estmated parameters to smulate counterfactual scenaros of employment and labor formalty under dfferent valuatons of the new health system mplemented n Mexco. JEL Codes: I13, J24,J3,J42,J6,O17 Keywords: Health Insurance, Socal Securty, Informalty, Mexco. Emal: gabrella.cont@ucl.ac.uk Emal: rta.gnja@nek.uu.se. Emal: rnarta@usp.br. We thank the comments of Joe Altonj, Orazo Attanaso, Matteo Bobba, Jan Eeckhout, Chrstopher Flnn, Santago Levy, Jean-Marc Robn and Costas Meghr and semnar partcpants at Unversté Catholque de Louvan, Insttute for Fscal Studes, SOLE Meetngs 2015, Stockholm School of Economcs, Unversty of Essex, 4th World Bank-Banco de España Conference. Ths paper s part of a project fnancally supported by the Swedsh Research Councl (Vetenskapsrådet), Grant No , and by the Brtsh Academy Newton Advanced Fellowshp, Grant No.AF Renata Narta thanks FAPESP grant No.2013/

2 1 Introducton A central topc n the global health agenda s unversal health care coverage. The World Health Organzaton (WHO) has defned unversal coverage as access of all people to comprehensve health servces at affordable cost and wthout fnancal hardshp through protecton aganst catastrophc health expendtures (WHO, 2010). The prmary goal of socal health nsurance schemes s to protect benefcares from the health and fnancal consequences of adverse health events. Many households lack suffcent fnancal resources to purchase essental health care, resultng n poor health condtons. Whle n ths sense there s scope for government nterventon n provdng nsurance, the mpacts of unversal health coverage on labor markets n developng countres are less clear. The Seguro Popular (SP) was ntroduced n 2002 n Mexco as a non-contrbutory health nsurance program and t was drected to half of the country s populaton, uncovered by socal protecton or employer provded health nsurance. That s, the nformal sector workers and the nonemployed. Pror to 2002, health nsurance n Mexco was tghtly lnked to employment. One of the few publc health nsurance schemes before SP was provded through the condtonal cash transfer Oportundades (now re-branded as Prospera, and called Progresa untl 2002), whch targets poor famles wth chldren, upon fulfllng some condtonaltes related wth school attendance by chldren and medcal examnatons. Oportundades has a component of publc health nsurance that ncludes free access to preventve health care, however famles wthout chldren would not qualfy for other publc health nsurance. To be elgble to the SP, an ndvdual needs to be uncover by employer provded health nsurance. Ths group consttuted half of the Mexcan populaton n Pror to SP, unnsured ndvduals could only access affordable health care through ther employer, thus the ntroducton of a non-contrbutory publc health nsurance scheme could have resulted n large effects on the labor market. In practce, the SP s a transfer(tax) to nformal(formal) sector workers and a transfer to the nonemployed. 1 On one hand, f the value placed on SP benefts s hgh, SP can lead to a negatve mpact on employment and/or formalty rates. On the other hand, wages n equlbrum mght compensate the ncrease n benefts n the nformal sector, and n ths case, the mpact on formalty rates and employment s ambguous. Thus, the labor supply and welfare mpacts of a non-contrbutory health nsurance program lke SP depend on how frms n each sector adjust wages gven benefts, on the allocaton of workers and frms across sectors and on how the newly free health servces are valued by famles. In ths paper, we analyze the effects of non-contrbutory health nsurance programs lke SP on labor market outcomes. We start by usng the staggered ntroducton of Seguro Popular across muncpaltes n Mexco n a dfferences-n-dfference strategy to show that SP s assocated to an ncrease n nformalty. In partcular, we use data from the Mexcan Labor Force Survey 1 Ths concern was voced n the Mexcan press (see, for example, fnanzas/59102.html 2

3 between 2000 and 2012 and we start by showng that the mplementaton of SP n a muncpalty s assocated wth an ncrease n the probablty of loss of Socal Securty protecton by low educated households wth chldren of 2-3%. Ths ncrease s drven by the transton of from the formal to the nformal sector of low educated males. Second, the salares decrease n the nformal sector, wth the decrease concentrated on the lower end of the wthn muncpalty dstrbuton of salares. Ths decrease n salares due to possble change n compensatng wage dfferental assocated wth the ntroducton of SP s consstent wth the small mpact of SP n nformalty. Then, to understand the extent to whch the access to free-health servces s valued by households members when they make ther labor market decsons, we develop and estmate a structural model of wage determnaton whch ncorporates the value of the nformal sector and nonemployed relatve to the formal sector. The change n access to health care wth the ntroducton of SP s one of the components we am to capture n the wllngness to pay to be n the nformal sector and nonemployed relatve to the formal sector. In the model, workers search randomly on and off the job and they may receve offers from formal or nformal frms. The nonemployed and nformal sector workers are not enttled any employment protecton benefts, whereas the formal sector workers receve employer-provded health nsurance and other benefts secured by labor laws (for example, guaranteed mnmum wage, redundancy payment and retrement pensons). In the formal sector, frms also ncur n employer-provded health provson costs. We model the choces of the members of the couple (heads and spouses) to capture the man features of a socal protecton system lke the one n Mexco. In partcular, n our model each of the members of the couple decdes between three possbltes: workng n the formal or nformal sectors or not workng at all. In case one of the members decdes to work on the formal sector, then the other spouse wll automatcally be covered by Socal Securty. Chldren n the famly wll covered by Socal Securty f they are under age 16 (f the parents work n the prvate sector; or 18 f the parents work n the publc sector). If none of the members works n the formal sector, the household s unnsured and, as such elgble to SP after ts mplementaton n the muncpalty of resdence. The labor market model we propose nnovates n several aspects. Frst, t s the frst model to allow the members of a couple to choose between jobs n the formal and nformal sectors and nonemployment, whch the are relevant alternatves n developng countres. Second, by modellng smultaneously the choces of both spouses we allow for non-segmented labor market based on gender, where men and women compete for the same jobs. By estmatng jontly the job destructon rates and transton rates for men and women for dfferent educatonal groups, we are able to assess the bas of models that consder the choce of men and women separately. The model s estmated on the Mexcan Labor Force Survey on the perods before and after the ntroducton of SP. We use the quarter of the mplementaton of SP n the muncpalty of resdence of the household to defne the perods before and after the ntroducton of the program. 3

4 Due to possble heterogenety n valuaton of health nsurance, the model s estmated for 8 dfferent groups based on demographc and geographc characterstcs. In partcular, educaton of the head (whether the head has more or less than 6 years of educaton, whch corresponds to elementary educaton n Mexco), area of resdence accordng to level of poverty (states n the north and south of country) and famly composton (wth and wthout young chldren). The wllngness to pay to be n the nformal or nonemployment sector for members of these groups dffers even pror to the ntroducton of SP, snce the groups are heterogenous wth respect to the lkelhood of facng health shocks (accordng to age), lkelhood of beng nformal (nformalty rate s hgher among the least educated), avalablty of health servces n the area of resdence and also on the benefts from the generous coverage of servces by the SP (the coverage of the program was specally generous for condtons prevalent among poor chldren). The model s able to replcate (1) the stocks of household types accordng to ther Socal Securty coverage and (2) the transtons n and out of employment and between formal and nformal jobs found n the perod before the mplementaton of SP. We then use the estmated parameters to smulate counterfactual scenaros of employment and labor formalty n whch we change the valuaton of SP health system. Our results am to shed lght on why the emprcal lterature has found lmted mpacts of Seguro Popular on employment and nformalty, and the mechansms whch explan t. In the next secton, we present a summary of the lterature on the labor market effects of health nsurance schemes not attached to the employer. In Secton 3 we explan the exact detals of SP and context n t was ntroduced. Secton 4 descrbes the data. In Secton 5 we present reduced form estmates of the mpact of mplementng SP n a muncpalty on nformalty rates and on the dstrbuton of wages n the formal and nformal sectors. In Secton 6 we present our model and n Secton 7 we descrbe the estmaton procedure used. The estmates from the structural model are presented n Secton 8. Conclusons are n Secton 9. 2 Lterature Revew SP and nformalty n Mexco The evdence on the labor market effects of SP s mxed (see the revew by Bosch, Cobacho and Pages, 2012). The estmates range from no mpact on the nformalty rates (Gallardo-Garca, 2006; Barros, 2011; Campos and Knox, 2010, Agulera, 2011, Duval and Smth, 2011) to small ncreases n the share of nformal workers for those wth less than 9 years of schoolng, marred women wth chldren or older adults (Azuara and Marnescu, 2010, Aterdo et al 2010, Pérez-Estrada, 2011, Bosch and Cobacho, 2011). Aterdo et al, 2010, fnd that SP s a assocated wth a reducton on the flow out of unemployment and out of the labor force, but del Valle, 2014, fnds the women n famles wth dsable or dependent ndvdual reduce unemployment and nactvty to become nformal workers. There are few papers that analyze the effects of SP on wages, and the fndngs range from 4

5 no effects (Barros, 2009, and Azuara and Marnescu, 2010), or a negatve mpact on nformal wages (Aterdo et al, 2010, Pérez-Estrada, 2011). Fnally, regardng the effects of SP on broader measures of welfare, there s some ndrect evdence through lower wages n the nformal sector (Aterdo et al, 2010, Pérez-Estrada, 2011) and reducton n postneonatal and chld mortalty n poor muncpaltes (Cont and Gnja, 2016, and Cont, del Valle and Gnja, 2016), a decrease n mscarrages (Pfutze, 2013), but most studes fnd no effect on health outcomes (Knox 2008, Kng et al. 2009, Barros 2011). Health Insurance Reform n US and Labor Market Recent reforms n the US health nsurance system, whch relaxed the lnk between employment and the provson of health nsurance are assocated to a stream of papers studyng the effects of publc health nsurance on labor supply. Backer et. al (2014) use a recent expanson on the elgblty to Medcad n Oregon and fnd no effect on employment, but and ncrease n welfare dependence. Kolstad and Kowalsk (2013) use the 2006-Massachusetts Health Reform and fnd compensatng wage dfferentals due to employer provded health nsurance. Garthwate, Gross and Notowdgdo (2014) estmate large ncreases n the labor supply assocated to an abrupt reducton on the Medcad coverage n Tennessee. However, so far there s no work consderng the general equlbrum effects of non-contrbutory health nsurance on broader welfare measures and the mechansms through whch the lnk between employment contract and provson of health nsurance operate. The approach we use relates manly to the followng papers. Dey and Flnn, 2005, use a search-matchng-barganng framework to study the effect of employer-provded health nsurance on moblty rates. Dey and Flnn, 2008, extends the framework n the 2005 paper from a sngle agent model to ncorporates potental dependence of couples labor market decsons. Azawa and Fang (2013) s the frst labor search model whch ncorporates health shocks. The lterature on search wth formal and nformal sectors s recent but two papers are partcularly relevant for our study. Albrecht, Navarro and Vroman (2009) model formal and nformal sectors followng the Damond-Mortensen-Pssardes approach, and they assume workers can only move to the formal sector from unemployment. They then use the model to smulate mpact of tax polces n the formal sector. Meghr, Narta and Robn (2015) model formal and nformal sectors usng a Burdett-Mortensen approach, where workers and frms can choose ther sector endogenously. They estmate the model and then smulate the mpact of ncreasng the cost of nformalty. However, the Brazlan setup lacks a sharp polcy change, such as the ntroducton of noncontrbutory health nsurance, whch allows us to recover the workers value for the nformal/nonemployment status. Fnally, we also relate to Fnkelsten, Hendren and Luttmer (2015) who recover the welfare beneft to recpents per each dollar spent n Medcad. They fnd t vares between $0.2-$0.4. 5

6 3 Background We now descrbe the health system n Mexco, as well as the penson system (whch could affect the decsons to take offers n the formal or nformal sector). In Secton (4) we explan how polcy reforms n Mexco mpact our samplng choce. 3.1 The Mexcan Health System and the Seguro Popular The Health Care System before Seguro Popular The reform of the health care system n Mexco was a process whch had been maturng for years snce the decentralzaton of the health servces for the unnsured n 1982 and the modfcaton of the Consttuton n 1983 to defne the protecton of health as a ctzen s rght and not only as a labor beneft. Before SP, health care n Mexco was characterzed by a two-tered system. About half of the populaton was covered through a contrbutory system (stll n place today) guaranteed by the Socal Securty Insttutons: the Mexcan Socal Securty Insttute (Insttuto Mexcano del Seguro Socal, IMSS), coverng the prvate sector workers; the Insttute for Socal Securty and Servces for State Workers (Insttuto de Segurdad y Servcos Socales de los Trabajadores del Estado, ISSSTE), coverng the cvl servants; and Mexcan Petroleums (Petroleos Mexcanos, PEMEX), coverng the employees n the ol ndustres. Health coverage was provded by these nsttutons n publc hosptals; however, ndvduals could also pay for care n prvate hosptals, or buy prvate health nsurance. In 2000, IMSS covered 40%, and ISSSTE 7% of the populaton, respectvely (Frenk et al., 2006). Health care was also avalable to the poor through two programs. The frst one was the Expanson of Coverage Program (Programa de Amplacon de Copertura, PAC), whch started n 1996 and conssted of brgades vstng the more rural and margnalzed areas of the country. Besdes PAC, part of the unnsured populaton had access to basc health servces through the Program for Educaton, Health and Nutrton (Programa de Educacon, Salud y Almentacón, Progresa). Ths was launched n 1997 n rural areas as the man ant-poverty program n Mexco; t was renamed Oportundades n 2002 and expanded to urban areas. The program has some overlap wth SP, snce t ncludes a health component offered n medcal unts managed by the IMSS-Oportundades and Secretara de la Salud (Mnstry of Health). 2 The unnsured populaton not covered by PAC or Progresa could seek health care ether n 2 Frst, Progresa benefcares receve free of charge the Guaranteed Basc Health Package (Paquete Básco Garantzado de Salud), whch ncludes a set of age-specfc nterventons; second, the nutrton of both chldren and pregnant women s montored through monthly consultatons (and nutrtonal supplements are dstrbuted n case of malnutrton); thrd, nformaton on preventve health behavors s provded through communty workshops; fourth, emergency servces are secured by the Mnstry of Health, IMSS-Oportundades (the dedcated network of medcal unts for famles enrolled n the program) and other state nsttutons (only n relaton to pregnancy and chldbrth); lastly, benefcary famles protected by Socal Securty have also access to second- and thrd-level care n the unts admnstered by IMSS, whle those unprotected have only lmted access to second-level care. The legslaton of Oportundades was obtaned from Hstorcas (accessed May 10th 2015). 6

7 publc health unts run by the Mnstry of Health (Secretara de Salud, SSA) or n prvate ones. In both cases, payment was at the pont of use and patents had to buy ther own medcatons. Hence, n 2000, approxmately 50% of health expendtures was classfed as out-of-pocket expenses (Frenk et al., 2009), and 50% of the Mexcan populaton - about 50 mllon ndvduals - had no guaranteed health nsurance coverage. The publc per capta health expendture on the nsured was twce as much as that on the unnsured (see Frenk et al., 2006). The Implementaton of Seguro Popular SP was launched as a plot program n 2002 n 26 muncpaltes (n 5 states: Campeche, Tabasco, Jalsco, Aguascalentes, Colma) under the name Health for All (Salud para Todos), wth the am to extend t gradually to the rest of the country. Durng 2002, 15 addtonal states 3 mplemented the program, by agreeng wth the federal government to provde the health servces covered by SP. By the end of the plot phase, on 31 December 2003, sx addtonal states 4 had joned, for a total of 613,938 famles enrolled. The System of Socal Protecton n Health (Sstema de Proteccón Socal en Salud, SPSS) was offcally ntroduced on January 1st 2004 by the General Health Law (Ley General de Salud, LGS), wth the am to extend health coverage to the elgble populaton. The Federal Government also created the Natonal Commsson for the Socal Protecton n Health (Comson Naconal de Proteccon Socal en Salud, CNPSS) to manage the system. The rules of operaton of the program stated that the expanson should prortze states wth: (1) low socal securty coverage; (2) large number of unnsured n the frst sx decles of ncome; (3) ablty to ensure the provson of servces covered by the program; (4) potental demand for enrollment; (5) explct request of the state authortes; (6) exstence of suffcent budget for the program. 5 In 2004, three more states ntroduced the program (Nayart, Nuevo Leon and Querétaro). The last three states (Chhuahua, Dstrto Federal and Durango) joned SP n Elgblty and Enrolment Indvduals who are not benefcares of socal securty nsttutons, or who do not have otherwse access to health servces, are elgble to enroll n SP. The basc unt of protecton s the household. Enrollment n the program s voluntary, and s granted upon complance wth smple requrements. 6 Informaton about all ndvduals afflated n the system s lsted n an admnstratve regstry, called the Padrón. At the end of 2010, the Padrón ncluded 15,760,805 famles, for a total of 43,518,719 ndvduals. By Aprl 2012, 98% of the Mexcan populaton was covered by some health nsurance (Knaul et al., 2012) - a remarkable achevement aganst the 50% covered 3 Baja Calforna, Chapas, Coahula, Guanajuato, Guerrero, Hdalgo, Mexco, Morelos, Oaxaca, Quntana Roo, San Lus Potos, Snaloa, Sonora, Tamaulpas and Zacatecas. 4 Baja Calforna Sur, Mchoacán, Puebla, Tlaxcala, Veracruz and Yucatán. 5 Daro Ofcal, 4 de julo de 2003, Reglas de operacón e ndcadores de gestón y evaluacón del Programa Salud para Todos (Seguro Popular de Salud). 6 The requrements are: proof of resdence n the Mexcan terrtory; lack of health nsurance, ascertaned wth self-declaraton; and possesson of the ndvdual ID (Clave Unca de Regstro de Poblacón, CURP). 7

8 only 10 years earler. Fundng Between 1999 and 2007, the rato of the total publc expendture on health to GDP was relatvely stable at 2.6% (see Fgure A.1 n the Appendx). Ths was one of the lowest fgures among OECD countres: the correspondng fgures for Denmark (the country wth the hghest share), US and Brazl n 2004 were 8.2%, 6.9% and 3.4%, respectvely. Between 1999 and 2004, the rato of the total publc expendture on health to GDP for nsured (not elgble) and unnsured (elgble) was also stable at 1.8% and 0.9%, respectvely. However, after 2004, the rato for the unnsured (elgble) experenced a steady ncrease, from 1% to nearly 1.5% n 2009, whle that for the nsured (not elgble) remaned constant after a temporary drop between 2004 and Hence, the program seems to have been successful n accomplshng one of ts goals, that of redstrbutng resources from the nsured to the unnsured. SP s funded by revenues from general taxes, on the bass of a trpartte structure smlar to that adopted by the two major socal nsurance agences n Mexco, IMSS and ISSSTE: (1) a socal contrbuton (Cuota Socal) from the federal government; (2) soldarty contrbutons from both the federal government and the states (Aportacones Soldaras); 8 (3) and a famly contrbuton (Cuota Famlar). The cuota famlar s an annual fee ntroduced to replace the out-of-pocket payments prevously made at the pont of use. It s based on the average household ncome relatve to the natonal ncome dstrbuton, but n 2010, 96.1% of the enrolled famles were exempted from payng t, on the bass of ther low socoeconomc status: n practce, very few households contrbuted (Bonlla-Chacn and Agulera, 2013). Coverage of Health Servces Once a famly s enrolled n SP, she s assgned a health center (whch, n turn, s assocated to a general hosptal) and a famly doctor for prmary care. The famly has access to a package of health servces, whose number of nterventons covered ncreased yearly, from 78 n 2002 to 284 n 2012, and t was lsted n a Catalogue of Health Servces (snce 2006 called Catalogo Unversal de Servcos de Salud, CAUSES) revsed annually (see Knaul et al., 2012). A wde range of servces were ncluded, from preventon, famly plannng, prenatal, obstetrc and pernatal care, to ambulatory, emergency and hosptal care, ncludng surgery. The basc coverage was complemented n November 2004 wth the ntroducton of the Fund for Protecton aganst Catastrophc Expenses (Fondo de Proteccón contra Gastos Catastrófcos, FPGC). The FPGC s a reserve fund of unlmted budget wth the objectve to support the fnancng of care for hgh-cost dseases typcally assocated wth premature death such as breast and womb cancer, and chld leukema. A further expanson took place n 2006 wth the ntroducton of Health Insurance for a New Generaton (Seguro 7 Ths was due to a faled attempt to ncrease publc revenues to fund SP (Ngenda, 2005). 8 The federal soldarty contrbuton s computed based on the followng elements: () number of benefcary famles; () health needs, proxed by state s ndcators of nfant and adult mortalty; () addtonal contrbutons called the state effort (esfuerzo estatal); and (v) the performance of health servces. 8

9 Medco para una Nueva Generacón, SMNG), whch offers a specfc package of servces for chldren under fve. Delvery of Health Servces The non-contrbutory and the contrbutory systems have separate networks of hosptals and health centers, each to serve ts own afflates. The offcal mplementaton of SP n 2004 establshed that, n each state, the State Regmes of Socal Protecton n Health (Regmenes Estatales de Proteccón Socal en Salud, REPSS) should pool federal (transferred from the CNPSS) and state funds and purchase the health servces from publc and prvate provders through management agreements. These blateral agreements had to specfy the number of famles to be served n each year, 9 the qualty condtons, and the allocaton of resources and funds to provde care to the SP benefcares, subject to certan spendng lmts. Supply of Health Care One of the man objectves of the health reform was to ncrease nvestment n health care nfrastructure and to acheve a more equtable dstrbuton of health care resources, on the bass of a specfc master plan (Plan Maestro de Infraestructura). Indeed, the proporton of the Mnstry of Health budget devoted to nvestment n health nfrastructure ncreased from 3.8% n 2000 to 9.1% n 2006, wth the constructon of 2,284 outpatent clncs and 262 (communty, general and specalzed) hosptals between 2001 and 2006; 10 as a consequence, the number of muncpaltes covered by each hosptal declned from a 2000 average of 7 to a 2010 average of 5. Addtonally, under the LGS, no faclty provdng servces could partcpate n the nsurance scheme unless t was accredted - and accredtaton was gven only n presence of the requred resources to provde the covered nterventons (Frenk et al., 2009). As a result, the gap between ndvduals covered and not by Socal Securty was sgnfcantly reduced n terms of the avalablty of general and specalst doctors, nurses and beds (Knaul et al., 2012). Further redstrbuton was acheved by prortzng the resources n poor muncpaltes (see Cont and Gnja, 2015). 3.2 Other concurrent polcy changes: Contrbutory Pensons, Taxaton and Chld Care We now descrbe other polcy changes n Mexco between 2000 and 2012, whch could have had mpacts on the labor market choces of ndvduals ndependent of SP. The Penson System The current Mexcan system s characterzed by two parallel systems, where a contrbutory socal securty system wth a package of defned benefts for formal workers 9 Ths number was set so that 14.3% of the unnsured populaton (as estmated n 2004) was to be enrolled n the program between 2004 and In the publc sector as a whole, 1,054 outpatent clncs and 124 general hosptals were bult n the same perod (Frenk et al., 2009). 9

10 n the prvate and publc sectors, whch coexsts wth a set of fragmented noncontrbutory servces and benefts offered through dverse socal protecton programs to the populaton lvng n poverty, wth low ncome, and n the nformal sector of the economy. The Mexcan Socal Securty faced a major reform n 1997, when the IMSS swtched the payas-you-go (PAYG) system to a fully funded system wth personal retrement accounts (PRAs). The penson beneft depends on the amount accumulated and captalzed n an ndvdual account (Agula, 2014). 11 The ISSSTE underwent a smlar reform n 2007, however, the change to a fully funded scheme was voluntary for workers who were already actve (Vllagómez and Ramírez, 2015). The Non-Contrbutory Penson System In 2001 the government of the Federal Dstrct mplemented the Nutrtonal Support, Medcal Attenton, and Free Medcnes Program for the Elderly (Programa de Apoyo Almentaro, Atencón Médca y Medcamentos Gratutos para Adultos Mayores), coverng elderly resdents older than 70 n the Dstrto Federal under a scheme targeted areas of hgh and very hgh level of margnalzaton (Vllagómez and Ramírez, 2015). The program became unversal n 2003, and n 2008 the beneft age was lowered to 68 years. In 2003 the government ntroduced the program Attenton to the Elderly n Rural Areas for adults older than 60 lvng n nutrtonal poverty and resdent n hghly and very hghly margnalzed rural communtes wth less than 2,500 nhabtants. The benefcares could not be partcpants of any other socal protecton program lke the Opportuntes. Ths was ntegrated nto the program 70 y más n The Oportundades created n 2006 a complement to benefcary famles wth adults older than 70. The benefts nclude cash transfers condtonal on benefcares complyng wth medcal exams and chldren attendng school. 12 The program 70 y más replaced the Attenton to the Elderly n Rural Areas program. Over the years the program was expanded untl t became the flagshp program of support for the elderly. Fnally, there are at least 13 state level program to support the elderly operatng n 2012, but accordng to Vllagómez and Ramírez, 2015, t s not possble to determne f ndvduals beneft from both state-admnstered programs and the federal programs such as Opportuntes and 70 y más. 11 The PAYG system s a well defned-beneft system and the benefts can be clamed through normal or early retrement. Mexco has no mandatory retrement age, but the normal retrement age s 65. The IMSS requres at least 10 years (500 weeks) of contrbutons to retre under PAYG rules. Socal securty benefts are computed as a proporton of the average wage n the 5 years before retrement, and benefts ncrease for each year of contrbuton beyond the requred 10 years. Under the PAYG rules, the mnmum payment guarantee, that s the mnmum socal securty benefts ndvduals can receve or socal securty beneft, s equal to the mnmum wage n Mexco Cty. To be enttled to ths beneft, the worker must contrbute for at least 1,250 weeks over hs work lfe. 12 From 2007 on, the beneft was provded to famles lvng n communtes of more than 2,500 nhabtants, whle those lvng n communtes up to 2,500 nhabtants and beng part of the Opportuntes program were supported through the program 70 y más. 10

11 3.3 Taxes Durng most of the perod n analyss there were no sgnfcant changes n the ncome or corporaton taxes n Mexco. The excepton was 2010, when a tax reform ncreased the margnal ncome tax rates for some workers but not others. In partcular, Mexco operated a dual ncome tax system for busness ncome where the taxpayer s lable to the hgher of ether the standard ncome tax (ISR) or a cashflow busness tax called the Impuesto Empresaral de Tasa Únca (IETU) from 2008 to The flat tax under IETU was not ncreased as part of the 2010 tax reforms, whlst the top rates of ISR were (see Abramovsky and Phlps, 2015). 3.4 Chld Care for Chldren of Mother n the Formal and Informal Sectors The government ntroduced n 2007 the program Estancas Infantles para Apoyar a Madres Trabajadoras, whch covers approxmately 90 percent of the cost of enrollng a chld under age four at a formal chld care center and s ntended to beneft women who are lookng for work, n school, or workng, that lve n famles wthout Socal Securty coverage. Ths program was expanded between 2007 and 2010 (see Calderon, 2014). 4 Data In ths paper we use data from two man sources. Padrón Ths s a consoldated regstry of all famles wth a vald enrolment n Seguro Popular by December 31st of each year snce 2002 (we have data untl 2010) and t s used by the Federal Government and by the States to decde the funds to be allocate to the program. The key treatment varable the date of mplementaton of SP n each muncpalty s constructed from ths data. The data contans detaled demographc and socoeconomc characterstcs of the enrolled famles, ncludng employment status, occupaton and assets. It also contans nformaton on the exact date of afflaton, resdence and he dentfers of the health center and general hosptal assgned to each famly at the tme of enrolment n the program. 13 The exact date of afflaton of famles s used to construct the date of mplementaton of the program n each muncpalty. We consder that a muncpalty has SP when the number of famles afflated to the program s at least 10 (our results are not senstve to ths defnton). Encuesta Naconal de Empleo (ENE) and Encuesta Naconal de Ocupacón y Empleo (ENEO) We use quarterly data from the Natonal Employment Surveys of Mexco. There are two perods of mplementaton (ENE for ) and 13 For the years 2002 and 2003 (n whch the program ran as a plot), only nformaton on the date of enrolment and on the state of resdence was recorded for each. However, t s possble to dentfy the exact date of mplementaton of SP n a gven muncpalty snce each famly has a unque dentfer. Thus, t s possble to lnk famles across years. 11

12 (ENOE for ). The ENE survey was mplemented quarterly between the second quarter of 2000 and the second quarter of 2004, for 4 localty szes n all the 32 Mexcan states and for one each cty n each state. The ENOE started n 2005 and households are followed for fve quarters. The data s a rotatng panel at the ndvdual and household level and t covers more than 11 mllon ndvduals from the second quarter of 2000 to fourth quarter of 2012 between 18 and 65 years old. From ths data set we observe the Socal Securty status of a specfc ndvdual across quarters, as well as hs/hers labor ncome when employed. All monetary values are deflated to the frst quarter of 2011 usng the CPI of Banco de Mexco. Sample restrctons The ENE covers just over 640 muncpaltes every quarter, whereas the ENOE covers about To keep a consstent sample of muncpaltes throughout the perod n analyss, we focus on the sample of muncpaltes surveyed every quarter snce Thus, we restrct our attenton to muncpaltes only n ENE and ENOE. That s, 640 muncpaltes. Then, we mpose the addtonal restrcton that a muncpalty must be present n the data at least for 2 years (8 quarters), whch reduces the sample to 628 muncpaltes. We restrct the sample of workers to be n marred households where the head s between 20 (where the chance of returnng to full-tme educaton s very low among the low educated) and 59 years old, who are stll not elgble for any non-contrbutory penson program for poor elderly. 14 Our restrcton to marred households ndvduals dscards 22% (243,229) households. Fnally, we drop 7% of households where the head of household s a female (60,005 households) and 2% (21,604 households) the sample where there s mssng nformaton about the gender of the spouse. Our fnal sample ncludes 748,181 households. An ndvdual s an nformal worker f he/she does not have access to health servces provded by hs/her job through one of the Socal Securty nsttutons n the country (IMSS, ISSSTE or PEMEX) or through the spouse s job. Note that we do not make a dstncton between self-employed and nformal employees, snce the defnton of nformalty depends on the Socal Securty coverage. As we show below, about 50% of all famles n Mexco n 2001 dd not have Socal Securty coverage. The mnmum wage s bndng and should be the mnmum amount pad to all formal employees. Of all heads and spouses ages n the data only 1% workers under a formal contract earn less than the mnmum wage, and we drop these ndvduals from the sample. We follow ndvduals for three months between ther frst and second surveys. We dentfy job-to-job transtons, unemployment-to-job, or job-to-unemployment transtons durng ths perod. In our model, we allow for job-to-job transtons wth formal and nformal sector, but we gnore them n our emprcal applcaton, snce wthn sector transtons are poorly measured 14 In Mexco 65 s the retrement age, but the partcpaton rate among nformal workers s very hgh among ndvduals between 65 and 70 years (47% and 6% of males n ths age range report to be nformal and formal workers, respectvely). 12

13 n the Labor Force Survey. 15 We use transtons between the frst and second ntervews snce about half of the observatons households-quarter whose head s 20 to 59 years old are observed only n the frst and second ntervews only. For each ndvdual n our sample (e, heads and ther spouses), we observe the employment status n the frst and second ntervews. From the second ntervew, we construct the transton ndcators and we observe the wage among ndvduals (e, heads and ther spouses) whch transtoned between non-employment and formal or nformal work. We present results for the sample of hgh and low educated famles, where we defne a famly to be n the low educaton group f the head has at most 6 years of completed educaton. Ths corresponds to elementary educaton n Mexco and n 2001, just before the mplementaton of SP, 40% famles n our data were n ths group. Descrptves We now present some basc facts regardng the labor market n Mexco. To document these basc facts we use quarterly data from the Mexcan Labor Force Survey ENE- ENOE for 2000 and In all results presented below we restrct the sample to couples where the head s years old. We consder that n each moment an ndvdual can be (1) unemployed or out-of-the-laborforce, (2) work n the formal sector or (3) work n the nformal sector. Table 1 basc statstcs of the data. In the table we nclude the employment status and wages for both heads and ther spouses n 2001 (just before the ntroducton of SP) and n 2007 (when the program reached all muncpaltes n our sample). The statstcs are presented separately for two groups of educaton: hgh educaton households (where the head has more than 6 years of educaton) and low educaton (f the head has at most 6 years of educaton). The table shows that pror to the ntroducton of SP about 36% of households n the hgh educaton group dd not have Socal Securty coverage, and ths fgure reached 62.5% among the low educaton group. The proporton of households wthout Socal Securty coverage remaned constant n the hgh educaton group, but ncreased by 3p.p. among low educated famles. Interestngly, the ncrease n nformalty among low educated famles s assocated wth an ncrease n the proporton of households where both members are nformal and a decrease n the share of households where both members are formal workers. The man changes n the quarterly transtons of heads of household between 2001 and 2007 pont to an ncrease the share of nformal households: the transtons from non-employment to a formal job decrease by nearly 3p.p., whereas the transtons from non-employment to an nformal job ncrease 11.5p.p., and when the spouse looses a formal job, the heads s also more lke to enter the labor market through the nformal sector (the transtons from non-employment to an nformal job, when the spouse looses a formal job ncrease by 1.6p.p.). Fnally, among spouses there s an ncrease n entry n labor market 15 Between 2000 and 2004 the Labor Force does not regster whether and ndvduals swtch jobs between any two surveys, thus the measure of job-to-job transtons across sectors whch we use n our emprcal work s a lower bound of all job-to-job transtons, snce wthn sector transtons are not observed n the data. 13

14 through the nformal sector, wth transtons from non-employment to nformalty ncreasng by 3.3p.p., and there s a decrease n destructon of nformal jobs (the transtons from nformalty to non-employment decrease by 4.3p.p.). Table 2 shows that the salares of both heads and spouses are lower n the nformal sector than n the formal sector, regardless of the educatonal group. These dfferences reflect unobserved productvty dfferences between the ndvduals who select nto the nformal sector wthn educatonal groups. Among men (heads) there s an ncrease n the wage gap across sectors between 2001 and 2007, whch s assocated wth a decrease n wages n the nformal sector relatve to the formal sector wages. 5 Reduced Form Estmates Before movng to the behavoral model, we explot the varaton n the tmng of mplementaton of SP at the muncpalty level. We analyze the mpact of Seguro Popular on the proporton of nformal famles and on the dstrbuton of wages for males and females wthn a muncpalty. Fgure A.3 n Appendx dsplays the year of mplementaton of SP n each muncpalty n Mexco, between 2002 and Ths graph shows that there s consderable varaton, both across muncpaltes and over tme, regardng the tmng of the adopton of SP n dfferent muncpaltes n the country. Thus, we start wth a basc specfcaton whch s a dfferencen-dfferences model, where we compare changes n outcomes for muncpaltes that ntroduced SP at dfferent years between 2002 and We estmate the followng model at muncpaltyquarter level: y mst = βsp mst + γx mst + µ ms + π t + ϕ s t + ε mst (1) where y mst s on of the two outcomes we study: share of nformal households n a muncpaltyquarter and log wages (of males and females). Because the SP can be assocated to dfferental changes across the dstrbuton of wages, besdes the mean wages we also consder the wages at the 10th, 25th, 50th and 75th percentles wthn a muncpalty. m ndexes the muncpalty, s the state, t ndexes the quarter. SP mst s an ndcator varable equal to one f muncpalty m n year t has mplemented SP. The muncpalty of resdence s measured n quarter t. Because muncpaltes adopted SP at dfferent tmes, we compare those wth or wthout SP access at the tme of survey by vrtue of the muncpalty of resdence. Thus, we can allow for unrestrcted muncpalty effects µ ms, whch control for unobserved determnants of y mst that are constant at muncpalty level and whch affect the outcome ndependently of SP; unrestrcted quarter effects π t and state-year lnear trends ϕ s t to account for state specfc trends whch affect outcomes ndependently of SP (such as federal-state budget agreements n place and ndependent of SP). The parameter of nterest s β, the effect of exposure to SP, whch s dentfed from varaton across muncpaltes and quarters. We control for the followng 14

15 demographc characterstcs of muncpaltes: the share of households n each muncpaltyquarter by educaton group (ncomplete prmary, complete prmary, complete lower secondary educaton or complete upper secondary and hgher educaton), and by age of the head (20-29; 30-39; and 50-59), share of households by localty sze, and quarter fxed effects. We also allow for lnear trends n characterstcs of the muncpalty resdence of the household taken n In partcular, for trends n the followng varables: quadratc n the ndex of deprvaton, log of total populaton, share of unnsured ndvduals, share of occuped ndvduals workng on the prmary, secondary and tertary sectors, and the number of hosptals and health centers n 2001, total number of doctors and nurses n hosptals per 1,000 unnsured ndvduals (taken n 2001). ε mst are dosyncratc shocks. The standard errors are clustered at muncpalty level to account for autocorrelaton n the outcome (Bertrand, Duflo and Mullanathan, 2004). We allow the mpact of SP nformalty to vary by the presence of chldren under 15 n the household for three reasons. Frst, the package of servces covered by SP ncludes a number of condtons prevalent among poor chldren. Second, the extenson of coverage of Socal Securty to chldren depends on the sector where the parent that earns coverage works. If the parent works n the prvate sector the coverage s extended to chldren under 16 (under 25 f they are studyng) or f under 18, n case the parent works n the publc sector. We do not consder the possblty of parents workng on prvate and publc formal sectors, thus we use the most strngent defnton, whch also concdes wth the age at whch chldren termnate mandatory educaton n Mexco. Fnally, we consder the presence of chldren under 15 n the household snce the Labor Force Survey does not contan the data of brth and thus, the chld may be close to turn 16 at the survey date at whch she would loose elgblty to Socal Securty coverage f not enrolled n school. The estmates n table 3 show that the mplementaton of SP n a muncpalty s assocated wth an ncrease n the probablty of loss of Socal Securty protecton by low educated households wth chldren of 2p.p (column 1, Panel A.2). Ths ncrease n nformalty s drven by a shft from formal to nformal sector jobs by low educated males (columns 3 and 4 n Panel A.2). Panel B shows that SP s not assocated to changes n the nformalty rates of hgh educated famles. Interestngly, SP s assocated to an entry n the labor market through the nformal sector by women wth chldren n hgh educated famles (see columns 5 and 7 n Panel B.2). We now turn to the mpacts on salares on tables 4 and 5. Columns 1-5 of table 4 shows that SP s assocated to a decrease n the mean wage of low educated men, wth the decrease beng more pronounced n the lower end of the dstrbuton (e, at percentle 10th there s a stronger decrease than at percentle 75th, wth the mpacts on wages rangng from 42% to 11%). There s also a small decrease n the wage of nformal hgh educated workers (see Panel B of table 4). Columns 6-10 of table 4 show no mpacts of SP on the wages n the formal sector. SP s not assocated to changes n the wages of low educated women workng n the formal or nformal sectors (see Panel A of table 5), but there s a declne n the salares of hgh educated 15

16 nformal female workers (columns 1-5 of Panel B n table 5), who are more lkely to enter the labor market through nformal jobs. 6 Jont Labor Search Model 6.1 The Basc Setup We now present a labor market model, whch we use to compare our causal estmates to the predctons from the model. The model parameters are estmated n the pre-reform perod. Tme n the model s contnuous and households seek to maxmze ther expected lfetme ncome. We consder shocks and decsons taken by spouses 1 (the head) and 2 (the spouse) n a household. Spouses can be: nonemployed (n), formal (f), or nformal (). The household enjoys nstant utlty gven by: u = w 1 (j) + w 2 (j) + a(1 H ) + γh, f both spouse 1 and 2 work u = w 1 (j) + b 2 + a(1 H ) + γh, f only spouse 1 works u = b 1 + w 2 (j) + a(1 H ) + γh, f only spouse 2 works u = b 1 + b 2 + γ, f nether works wth j = formal or nformal; and H an ndcator functon for nformal household (when both j are not formal). w 1 s the labor ncome of spouse 1, w 2 the labor ncome of spouse 2, b 1 the non-labor ncome of spouse 1, and b 2 the non-labor ncome of spouse 2. In the formal sector, w s after tax wage (but before socal securty contrbutons); w n the nformal sector s the gross wage. The parameter a captures all amentes n the formal sector relatve to those n the nformal sector, except the value of health nsurance by Seguro Popular. We assume that γ s the value of health nsurance by Seguro Popular, whch s offered when no spouse s n the formal sector. We explan below how these two parameters are dentfed. In ths verson we assume that ndvduals are rsk neutral. In future work we wll allow for rsk averson. Spouse 1 and spouse 2 who face mutually exclusve shocks n the labor market. There s one excepton to ths: when spouse 1 has a job and spouse 2 s nonemployed, a shock that destroys the job of spouse 1 may nstantly create an opportunty for spouse 2 to move the nformal sector. We consder that when spouse 1 gets a formal job opportunty, spouse 2 may go nto nonemployment. The same holds for spouse Household s Value Functons Let W jk be the value functon for a household where the head (spouse 1) s n status j = f, and the spouse (spouse 2) s n status k = f,. Further, for j = f, we use W jj (w 1, w 2 ) = W jn (w 1 ) 16

17 f w 2 = 0, W jj (w 1, w 2 ) = W nj (w 2 ) f w 1 = 0, and equal to W jj (w 1, w 2 ) = W nn f w 1 = w 2 = Only one member works n the formal sector rw fn (w 1 ) = w 1 + b 2 + a + δ s 1 f (1 ps 2 ) (W nn W fn (w 1 )) + δ s 1 f ps 2 max {W n (x) W fn (w 1 ), 0} df s 2 ff max {W fn (x) W fn (w 1 ), 0} df s 1 f max {W n (x) W fn (w 1 ), 0} df s 1 f nf n max {W ff (w 1, x) W fn (w 1 ), W nf (x) W fn (w 1 ), 0} df s 2 f max {W f (w 1, x) W fn (w 1 ), 0} df s 2 (x) where p s 2 s the probablty that spouse 2 moves from nonemployment to nformal gven that spouse 1 moves from a formal job to nonemployment. Ths s a formal household, wth Socal Securty coverage. When the head looses the formal job, then wth probablty p s 2 the spouse gets an offer from the nformal sector. In ths case the household may re-evaluate ts plan by consderng the flow of gans f the spouse takes the nformal offer (W n (x)) aganst the opton of not takng t. Wth probablty 1 p s 2 the spouse does not get the offer from the nformal sector, n whch case there s no decson to be made by the household. New offers from the formal sector to the head arrve at rate ff, and the household decdes whether the head wll take the offer or not. Emprcally, we do not consder ths hypothess, snce transtons wthn the same sector are not perfectly observed n our man data sets (e, there s no nformaton about whether the ndvdual swtched jobs wth sector between the frst and second ntervews). New offers from the nformal sector to the head arrve at rate f, and the household decdes whether the head wll take the offer or not. Job offers from the formal sector to the spouse arrve at rate nf, and the household now faces 3 possbltes: (1) the spouse may take the new formal offer, (2) snce the household wll Socal Securty coverage f the spouse becomes a formal worker, then the head may qut hs formal job, fnally, (3) the household may do nothng (e, the formal offer by the spouse s not accepted). Fnally, job offers from the nformal sector to the spouse arrve at rate n, n whch case the household evaluates ts current stuaton where the head s formal worker and the spouse non-employed aganst the stuaton where the spouse enters the nformal sector. 17

18 The value functon W nf (w 2 ) s smlar to the above equaton and ts also a formal household. There s only an exchange n the status between spouses 1 and 2: rw nf (w 2 ) = w 2 + b 1 + a + δ s 2 f (1 ps 1 ) (W nn W nf (w 2 )) + δ s 2 f ps 1 max {W n (x) W nf (w 2 ), 0} df s 1 ff max {W nf (x) W nf (w 2 ), 0} df s 2 f max {W n (x) W nf (w 2 ), 0} df s 2 f nf n max {W ff (x, w 2 ) W nf (w 2 ), W fn (x) W nf (w 2 ), 0} df s 1 f max {W f (x, w 2 ) W nf (w 2 ), 0} df s 1 (x) and p s 1 s the probablty that spouse 1 moves from nonemployment to nformal gven that spouse 2 moves from a formal job to nonemployment. n the nformal sector where q s 2 rw n (w 1 ) = w 1 + b 2 + γ + δ s 1 (1 q s 2 ) (W nn W n (w 1 )) + δ s 1 q s 2 max {W n (x) W n (w 1 ), 0} df s 2 max {W n (x) W n (w 1 ), 0} df s 1 f nf n max {W fn (x) W n (w 1 ), 0} df s 1 f max {W f (w 1, x) W n (w 1 ), W nf (x) W n (w 1 ), 0} df s 2 f max {W (w 1, x) W n (w 1 ), 0} df s 2 (x) s the probablty that spouse 2 moves from nonemployment to nformal gven that spouse 1 moves from an nformal job to nonemployment. Ths s an nformal household, wthout Socal Securty coverage. When the head looses the nformal job, then wth probablty q s 2 the spouse gets an offer from the nformal sector. In ths case the household may re-evaluate ts plan by consderng the flow of gans f the spouse takes the nformal offer (W n (x)) aganst the opton of not takng t. Wth probablty 1 q s 2 the spouse does not get the offer from the nformal sector, n whch case there s no decson to be made by the household. New offers from the nformal sector to the head arrve at rate, and the household decdes whether the head wll take the offer or not. As mentoned above, we do 18