Labor market search, informality and schooling investments

Bobba, Matteo; Flabbi, Luca; Levy, Santiago Working Paper Labor market search, informality and schooling investments IDB Working Paper Series, No. IDB-WP-863 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Bobba, Matteo; Flabbi, Luca; Levy, Santiago (2018) : Labor market search, informality and schooling investments, IDB Working Paper Series, No. IDB-WP-863, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0000983 This Version is available at: https://hdl.handle.net/10419/173908 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Labor Market Search, Informality and Schooling Investments Matteo Bobba Luca Flabbi Santiago Levy IDB WORKING PAPER SERIES Nº IDB-WP-863 January 2018 Department of Research and Chief Economist Inter-American Development Bank January 2018 Labor Market Search, Informality and Schooling Investments Matteo Bobba* Luca Flabbi** Santiago Levy*** * University of Toulouse ** University of North Carolina *** Inter-American Development Bank Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Bobba, Matteo. Labor market search, informality and schooling investments / Matteo Bobba, Luca Flabbi, Santiago Levy. p. cm. — (IDB Working Paper Series ; 863) Includes bibliographic references. 1. Labor market-Mexico-Econometric models. 2. Informal sector (Economics)- Mexico-Econometric models. 3. Wages-Effect of education on-Mexico. 4. EducationEconomic aspects-Mexico. I. Flabbi, Luca. II. Levy, Santiago. III. Inter-American Development Bank. Department of Research and Chief Economist. IV. Title. V. Series. IDB-WP-863 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. http://www.iadb.org 2018 Abstract This paper develops a search and matching model where firms and workers are allowed to form matches (jobs) that can be formal or informal. Workers optimally choose the level of schooling acquired before entering the labor market and whether to search for a job as unemployed or as self-employed. Firms optimally decide the formality status of the job and bargain with workers over wages. The resulting equilibrium size of the informal sector is an endogenous function of labor market parameters and institutions. The paper focuses on an increasingly important institution: a “dual” social protection system whereby contributory benefits in the formal sector coexist with non-contributory benefits in the informal sector. Preferences are estimated for the system—together with all the other structural parameters of the labor market—using labor force survey data from Mexico and the time-staggered entry across municipalities of a non-contributory social program. Policy experiments show that informality may be reduced by either increasing or decreasing the payroll tax rate in the formal sector. They also show that a universal social security benefit system would decrease informality, incentivize schooling, and increase productivity at a relative fiscal cost similar to that generated by the current system. JEL classifications: J24, J3, J64, O17 Keywords: Labor market frictions, Search and matching, Nash bargaining, Informality, Returns to schooling We thank participants at conferences, workshops and seminars at Barcelona GSE, Bordeaux, CEPR-IGC, EUDN, Geneva, IZA, Laval, NYU, SOLE, Toulouse, Turin, UCL-CeMMAP, and Wisconsin-Madison for very useful comments. Marco Pariguana, Jose Mauricio Salazar, and especially Matias Morales provided excellent research assistance. Financial support from the Agence Fran¸caise de D´eveloppement (AFD) and the Inter-American Development Bank (IDB) is gratefully acknowledged. 1 1 Introduction High levels of informality characterize many labor markets, typically in mediumand low-income countries.1Informality can be broadly defined as any deviation from labor contracts, such as avoiding payroll contributions and not conforming to labor law statutes. If regulations are rigid and imperfectly enforced, non-compliance allows both firms and workers greater labor market flexibility that may improve labor allocations across sectors and occupations. Firms’ direct cost of increased flexibility is the possibility of being discovered and punished. The main cost to workers – on top of the loss of some protections on the job – is the lack of social security coverage, such as the provision of health and pension benefits.2 Studying costs and benefits of informality – including consequences for productivity and welfare – requires an equilibrium model of the labor market taking into account how workers and firms jointly sort between formal and informal jobs. It also requires a wat to explain empirical evidence that does not conform to neither a segmented or a competitive view of the labor market. The evidence on labor market dynamics shows mobility of workers not only from formal to informal jobs, but also from informal to formal jobs. The evidence on wage distributions shows that on average wages are higher in formal jobs but also that many informal jobs pay more than formal ones, creating a large overlap between the formal and informal wage distributions. The frequent and significant flow of workers from formal to informal is in contrast with a segmented view where barriers restrict access to the formal sector. But it is also in contrast with a competitive view whereby the presence of the two types of jobs in equilibrium is justified by compensating differentials mapping into preferences and skills that are supposed to be stable over time. The overlap in the wage distributions does not seem consistent with either view. Both segmentation and compensating differentials should generate a much starker wage ranking between the formal and the informal sector.3 In this paper, we develop and estimate a search, matching, and bargaining framework that takes into account sorting and endogenous decisions over formality regimes and that replicates 1This issue is particularly acute in Latin America, where even large middle-income economies with welldeveloped labor market institutions feature more than half of the labor force in the informal sector [Perry et al., 2007; Levy and Schady, 2013]. But the phenomenon is also common in other parts of the world [La Porta and Shleifer, 2014]. 2There are other less direct but potentially more persistent costs related to informality. Informality may reduce firms’ access to capital markets, affecting their investment decisions in physical capital and technology. But it may also change workers’ labor market returns, affecting their investment decisions in human capital. Moreover, informality reduces government revenues, generating fiscal imbalances and hampering the government’s ability to provide public goods. 3Magnac [1991] presents these two competing views of the labor market. Meghir et al. [2015] provide an instructive discussion based on empirical evidence collected on Brazil. Maloney [1999] is a seminal contribution showing that transitions between the formal and the informal sector in Mexico are equally probable in both directions. Maloney [2004] is an update adding sociological and anthropological evidence to the economic evidence in supporting a non-segmented view. Recent evidence on Mexico is in Anton et al. [2012]. Perry et al. [2007] provide and review evidence on a large number of Latin American countries, confirming the empirical evidence summarized in the text. In Section 2 we confirm the same data patterns in our sample extracted from the Mexican labor force survey. 2 the main empirical features of labor markets with high informality. Search and matching frictions support the presence of different types of contracts in equilibrium. Optimal decisions rules based on reservation values and the presence of termination shocks generate transitions between labor market states in any direction. Finally, match-specific productivity and bargaining generate the overlapping wage distributions. We also develop an identification and estimation strategy that is able to recover all the structural parameters of the model using standard labor market data from Mexico. We next perform policy experiments where we can evaluate the impact of changes in the institutional parameters responsible for the emergence of informality in the first place. We further incorporate into our modeling and estimation strategy three features that take into account three relevant but overlooked issues. First, we introduce an endogenous schooling decision. Labor market distortions generate informality by affecting labor market behavior. But in a dynamic environment they may also affect crucial decisions taken before entering the market. Among them, the schooling decision is arguably one of the most relevant and one with the most persistent effects on workers’ outcomes. If the distortions leading to informality are found to be strong enough to significantly impact returns to schooling, then their consequences go beyond the misallocation of workers across jobs but also include the long-term composition of workers in terms of human capital.4 Second, we model in detail the structure of the social security system. In response to the lack of social security coverage for informal workers, several countries are increasingly providing social security benefits that are not based on payroll contributions. This has created a “dual” social security system where formal jobs come with benefits financed by payroll contributions and informal jobs come with benefits financed by resources collected outside the labor market.5 This duality creates subsidies and incentives that cannot be ignored if one wants to explain the observed levels of informality. On top of modeling the institutional details of the system, we go a step further by tackling an issue common to any system providing social security benefits: the benefits may not be valued by the consumer at full value. The consumer’s willingness to pay for the service may be very different from the contribution they are forced to pay to receive it or from the actual cost incurred to provide it. Recovering these preferences for the social security system is crucial for evaluating its impact on labor market outcomes. The problem is only exacerbated if the system is composed of two separate systems coexisting in the same labor market, as in Mexico and other countries with a dual system. Third, we propose a more nuanced definition of informality by allowing workers to perform an informal job as either an employee or a self-employed worker. “Necessity” self-employment – i.e., a form of self-employment requiring very limited skills and capital and imposing almost no barriers to entry – is so common and pervasive in labor markets with high informality that it 4So far, the literature on the long-term investment impact of informality has focused on the firms’ side. See for example, La Porta and Shleifer [2008]; de Paula and Scheinkman [2010, 2011]; Ulyssea [2015]. 5Non-contributory programs are generally financed with public resources that are independent of the revenues from the social security system within the formal sector. See Levy [2008] for a detailed description of Mexico and Melguizo et al. [2017] for a recent review for Latin American countries. 3 is frequently used as a proxy for informality itself. However, if it is true that almost all these self-employed workers are informal, it is not true that almost all the informal jobs are performed by the self-employed. A prominent portion of informal jobs is organized as a genuine subordinate working relationship with a well-defined employer. Since self-employed workers and employees have markedly different labor market dynamics, these differences must be explicitly taken into account in order to provide a complete characterization of the informal sector.6 We propose a model where individuals decide whether to acquire productivity-enhancing schooling prior to labor market entry. Conditional on schooling, they choose whether to search full-time for a job or to search only part of the time while working as self-employed. Agents are then randomly matched with firms that could offer formal or informal wage contracts. The number of meetings between workers and firms is governed by a matching function that depends on the endogenous proportions of searchers and vacancies in each schooling sub-market. Upon observing match-specific productivity, firms optimally post the formality status of the job and engage in bargaining with workers to negotiate wages. Searchers and informal workers receive a non-contributory social benefit, while formal workers receive a contributory social benefit, which partly depends on their wages. Workers are allowed to have preferences over these benefits. Firms face no costs in entering the market but they pay a flow cost to keep a vacancy open. Workers are heterogenous in their cost for acquiring additional schooling and in their ability to generate self-employment income. Jobs are heterogenous in their match-specific productivity. Accepted wages, the distribution of workers over labor market states, the level of benefits in the formal sector, the overall size of the informal sector, and schooling levels are all endogenously determined in the search equilibrium. The model is estimated on individual-level data from the Mexican labor force survey (ENOE).7 The empirical implications of the model together with some distributional assumptions are enough to identify most of the structural parameters from the micro data. The exception is the preference parameter for non-contributory benefits. In order to identify this parameter, we rely on an additional source of variation in the data: the time-staggered introduction of the Seguro Popular program across municipalities. The program increases access to health care benefits for individuals not covered by the contributory system.8Estimation results show reasonable values of the model parameters, including those harder to identify such as firms’ costs of being discovered and punished for hiring informally, workers’ preferences for the social security system, and the parameters 6Fields [1975] is a seminal contribution pointing out the specific role played by self-employment in labor markets with high informality. Margolis [2014] and World Bank [2012] provide an overview of the evidence on many developing countries. Bianchi and Bobba [2013] confirms that financial barriers are not an important obstacle to entering self-employment in Mexico. Both Meghir et al. [2015] and Bosch and Esteban-Pretel [2012] acknowledge that informal workers may be either self-employed or employees, but they aggregate them in one unique labor market state. Narita [2011] is a rare contribution making the distinction between self-employed and employees in characterizing workers’ informality. 7ENOE (Encuesta Nacional de Ocupaci´on y Empleo) is Mexico’s official labor force survey, which is similar to the US Current Population Survey. 8The same source of exogenous variation is used by Conti et al. [2017] in an equilibrium search model to identify the marginal willingness to pay for the benefit provided by the Seguro Popular program. 4 of the matching functions for each schooling sub-market. The estimated model is used to understand how the characteristics of the labor market and of the institutional setting lead to informality in the presence of optimal dynamic behavior. It is also used to perform policy experiments evaluating the impact of the dual social security system on informality levels, labor market outcomes, schooling acquisition, productivity and welfare. Results show that equilibrium wages and informality rates are very sensitive to the payroll tax rate in formal jobs and to the level of non-contributory social security benefits. Contrary to the usual belief found in previous literature and in the policy debate, the contribution rate is shown to have a non-monotone impact on the informality rate. Informality may be reduced by either increasing or decreasing the payroll tax rate from current levels. Consistently with previous literature, the noncontributory social security benefits has a monotone impact on informality, but the elasticity is not constant and it is high around current levels. The policy approximates recent reforms proposed in Latin America, which are aimed at increasing the generosity of the non-contributory benefits. Setting the non-contributory benefits to zero would be enough to completely eliminate informality among employees but would have a limited impact on informality among the self-employed. We finally propose a policy scheme that may represent a feasible and concrete alternative to the recent reforms proposed in Latin America. The scheme consists in harmonizing benefits between formal and informal workers, restoring the link between contributions and wages in the formal sector, and reducing the payroll tax rate of formal jobs. Results show that this policy would completely eliminate informality among employees with a High School degree and would provide incentives to acquire more schooling. In the post-policy equilibrium, the proportion of High School graduates increases to 70% from a baseline value of 40%. As a result of these composition effects, the value of production increases by 17 percentage points and overall workers’ welfare increases by 35 percentage points, while the relative cost of the policy remains very similar to benchmark levels. The results of these policy experiments on both schooling and aggregate productivity are very sensitive to equilibrium effects. For example, increasing the payroll tax rate from zero to more than 60% decreases schooling by only a few percentage points in partial equilibrium but cuts High School completion rates from 70% to 35% in general equilibrium. Endogenous meeting rates between firms and workers are the main channel behind this difference, which are influenced by the schooling-specific vacancy creations implemented by firms. The informality literature using equilibrium models of the labor market characterized by frictions is growing, but it is still thin. Meghir et al. [2015] is the contribution closest to ours. In it, the authors develop an equilibrium search model with wage posting in which firms endogenously locate in the formal or informal sector. They estimate the model parameters on Brazilian labor force data, showing that stricter enforcement reduces informal employment and increases welfare by improving the allocation of workers to higher-productivity firms. The other published contributions in this literature do not attempt model estimation. Bosch and Esteban-Pretel [2012] calibrate a two-sector model of the Brazilian labor market where firms have a choice of hiring workers formally or informally. Albrecht et al. [2009] is a theoretical contribution developing an 5 willingness to pay for the benefit.22 In the presence of partial enforcement, the trade-off between benefits and contributions determines the equilibrium level of informality. At the same level of contribution rate and benefit, the incentives to work formally may change considerably if the benefits are valued more or less than the contribution paid to receive them. As discussed before, B0is received by all the individuals except formal employees and is paid in a fixed amount equal for all recipients. Instead, B1is a function of wages and is only received by the formal employees. It is defined as: B1(w1(x;y, h)) ≡τt [w1(x;y, h)] + b1,(2) where τdenotes the share of the total contribution t[w1(x;y, h)] that is proportional to the worker’s wage and represents benefits such as a defined contribution retirement plan. The (1 −τ) share of the total contribution is instead redistributed equally among all formal employees. The equal amount received by each agent is denoted by b1, which is endogenous because it depends on the total amount contributed by formal employees, which is itself a function of how many agents work as formal employees in equilibrium and at what wages.23 The b1benefit is meant to capture another institutional feature that is present in the system: contributory benefits that are the same for all formal employees, the most notable example being health benefits. The system has important distributional implications. Since the collection of contributions is proportional to wages and b1is equal for all formal employees, the system implies redistribution from high-wage earners to low-wage earners within the formal sector. Moreover, since b1is not schooling-specific and since workers with higher levels of schooling earn higher wages, it also implies redistribution from high-schooling workers to low-schooling workers. This feature introduces a crucial equilibrium link between the high schooling group and the low schooling group, which would otherwise be separated into two segmented labor markets. Firms search to fill vacancies, and they meet workers at a Poisson rate ζh. To keep a vacancy open, firms incur a flow cost νh. Once they meet a worker, the same behavior and sequence of events described above take place: a match-specific value xis observed, the formality status fis posted, a wage wf(x;y, h) is determined by bargaining, a decision about accepting or rejecting the match is taken. In making their decisions, firms take into account their flow payoffs, i.e., the instantaneous profits from a filled job. For given productivity, the profits are different if hiring formally or informally and are respectively defined as: x−w0(x;y, h)−chx(3) x−(1 + t)w1(x;y, h),(4) where xis the match-specific value generating revenues for the firm; w0and w1are the wages paid 22A similar setting and interpretation is used by Dey and Flinn [2005] to evaluate health insurance and by Flabbi and Moro [2012] to evaluate job flexibility. 23See Appendix A.3 for the formal derivation of b1in equilibrium. 12 to the workers in the informal and formal sector, respectively, and t– described in equation (2) – is the payroll tax rate. Equation (4) clarifies that it is withdrawn at the source by the firm. The parameter chis the way we model the cost to firms of hiring informally. As discussed in Section 2, there is a positive probability of being discovered hiring workers informally and having to pay a penalty. However, not all firms have the same probability of being caught: larger and more productive firms face a higher probability of being audited and fined. Given this institutional context and since our model does not allow to pin down firm size, we assume that the cost simply increases with productivity (in our notation, the match value x). Since we do not have direct observation of the monitoring process in our data, we impose a particularly parsimonious specification: the linear, one-parameter function chx.24 The last element that needs to be added to complete the description of the two-sided search environment is the specification of the matching process. We have anticipated that both firms and workers search at random in each schooling sub-market and meet each other at Poisson rates λh, γh, and ζh. Since the equilibrium proportions of workers searching for jobs and of firms searching to fill vacancies is endogenous, the meeting rates must also be endogenous. We capture this process by assuming a standard matching function.25 The number of jobs per worker by schooling level mhis assumed to depend on the measures of individuals actively searching for a job and on the vacancies, vh, according to the following parametrization: mh= (uh+ψhsh)ιh(vh)1−ιh,(5) where ψh∈(0,1] is a parameter denoting the lower search efficiency of the self-employed with respect to the unemployed. It may be interpreted as the time spent searching by each selfemployed worker or as the proportion of self-employed workers searching at each moment in time. We can now write all the contact rates as functions of the degree of tightness in the labor market, ωh≡vh uh+ψhsh: λh=mh uh uh uh+ψhsh=ω1−ιh h γh=mh sh ψhsh uh+ψhsh=ψhω1−ιh h ζh=mh vh=ω−ιh h. 3.2 Value Functions 3.2.1 Workers Before entering the labor market, workers face an individual-specific cost κ∼T(κ) of acquiring schooling level h= 1. Since the cost is assumed to be uncorrelated with future labor market performance, the only relevant state variable affecting the present discounted value of participating 24To the extent that size and productivity are positively correlated, this specification flexibly captures the notion that imperfect enforcement creates a size-dependent distortion in the economy [de Paula and Scheinkman, 2011; Ulyssea, 2015]. 25See Petrongolo and Pissarides [2001] for a survey. See Meghir et al. [2015] and Bosch and Esteban-Pretel [2012] for applications to Latin American countries. 13 in the labor market is the schooling level hacquired prior to labor market entry. To characterize this choice we just need to present the value function of completing a given schooling level before any labor market shock occurs and before any value of self-employment is revealed. We capture this choice with the function Z(h): Z(h) = Z0 Q(y, h)dR(y) (6) Q(y, h)≡max{S(y, h), U(h)},(7) where we introduce the functional Q(y, h) to simplify the conditioning on yin the rest of the paper. The present discounted value of participating in the labor market with a given schooling level his the value of searching in that market. However, an individual can choose if they want to search as unemployed U(h) or as self-employed S(y, h). If they choose the second, they enjoy income from the self-employment activity but they meet employers at a lower rate. The trade-off is clarified by looking at the value functions of these two searching states: (ρ+λh)U(h) = ξh+β0,hB0+λhX f∈{0,1}Zx max{Ef[wf(x), y, h], U(h)}dG(x|h) (8) (ρ+γh)S(y, h) = y+β0,hB0+γhX f∈{0,1}Zx max{Ef[wf(x), y, h], S(y, h)}dG(x|h).(9) The arrival rates of offers are λhand γh. The meeting can be with an employer offering a formal or an informal job. The formality status choice is a function of the match-specific productivity x, but it is posted by the firm: that is why, from the point of view of the worker, it appears in the option value as a simple sum. Conditioning on the formality status and the specific productivity draw, agents bargain over wages and decide to accept the job or not. The optimal decision is represented by the maximization between the current state (either U(h) or S(y, h)) and the new employee state (either E0[wf(x), y, h] or E1[wf(x), y, h]).26 While the option values of the two searching states have a very similar structure, there is an important difference between their flow values. Both states receive a constant flow value of non-contributory benefits β0,hB0, but the selfemployed also receive income ythat is allowed to vary among different self-employed searchers. Instead, all unemployed searchers have the same utility or disutility from searching ξh.27 The values of working as an employee with a formal or informal job contract are, respectively: (ρ+ηh)E0[w0(x;y, h), y, h] = w0(x;y, h) + β0,hB0+ηhQ(y, h) (10) (ρ+ηh)E1[w1(x;y, h), y, h] = w1(x;y, h) + β1,hB1[w1(x;y, h)] + ηhQ(y, h).(11) 26Notice that we force notation a bit by not differentiating between employees coming from unemployment and employees coming from self-employment. To be precise, we should eliminate the dependence of yfrom the value of employment of agents searching as unemployed just as the value of unemployment U(h) does not depend on y. 27This ex ante homogeneity is the usual assumption in search-matching-bargaining models, while the heterogeneity in the outside options for self-employed searchers is a feature akin to search models with on-the-job search. 14 The flow values received by employees is the sum of the wage and the value attached to social security benefits. The wage is a function of productivity, schooling level, formality status and, possibly, self-employment income. As will be shown in Section 3.3.3, wages depend on schooling and self-employment income because they both potentially affect the worker’s outside option when bargaining with the employer. The social security benefit is fixed for the informally employed, but it is increasing in wages and productivity for the formally employed. The only shock received by employees is a termination shock, received at the Poisson rate ηh. If employees receive the shock, they go back to their respective searching state: either U(h) or S(y, h). 3.2.2 Firms Firms post vacancies and search for workers to fill them. The value of a posted vacancy is: (ρ+ζh)V[h] = νh+ζh[uh uh+ψhshZx max{F1[x, y, h], F0[x, y, h], V [h]}dG(x|h),(12) +ψhsh uh+ψhshZyZx max{F1[x, y, h], F0[x, y, h], V [h]}dG(x|h)dR(y|h)]. The flow cost of keeping a vacancy open is denoted by νh. Employers meet potential employees at a rate ζh. Since potential employees may be unemployed or self-employed, the probability of meeting one or the other is a function of their proportion in the equilibrium measure of searchers. This is taken into account by the two fractions multiplying the integrals. If the employer meets an unemployed searcher, a match-specific productivity is extracted. Based on its value and the knowledge of the outside option of the potential employee (unemployment), the employer optimally decides to post the job offer as formal or informal. This is captured by the max operator over three possible options: F0[x, y, h], F1[x, y, h] and the status quo option V[h]. If the employer meets a self-employed searcher, the same process takes place, except that in this case the employer must also take into account that the potential employee’s outside option changes with self-employment income y. This is incorporated in expression (12) by integrating over the distribution of yvalues, R(y|h). Once the job is filled, either formally or informally, the corresponding value functions are: (ρ+ηh)F0[x, y, h] = x−w0(x;y, h)−chx+ηhV[h],(13) (ρ+ηh)F1[x, y, h] = x−(1 + t)w1(x;y, h) + ηhV[h].(14) The expressions are analogous to workers’ side expressions (10) and (11): flow values plus the option value given by the probability of the termination shock ηhtimes the value of the searching state. The flow values are simply the flow profits, but they parsimoniously incorporate all the complexity of the institutional system. This is why the mapping between productivity and wage paid by the firm depends on the formality status of the job. We represent this feature by indexing 15 the wages with the status indicator f, leading to w0in equation (13) and to w1in equation (14).28 3.3 Equilibrium 3.3.1 Schooling Before entering the labor market, workers have to decide whether to acquire the high schooling level h= 1 or remain at the default schooling level h= 0. Since acquiring additional schooling requires an investment κ∼T(κ), agents decide based on the following maximization: max h{Z(0), Z(1) −κ}, where Z(h) – defined in equation (6) – is the value of participating in the labor market given schooling level h. The cost κis assigned by nature and does not vary over time. Since Z(1) −κ is decreasing in κand Z(0) does not vary in κ, there exists a unique: κ∗:Z(0) = Z(1) −κ∗. The optimal decision rule is therefore a reservation value rule where only agents with κ<κ∗ acquire the schooling level h= 1, whereas agents with κ≥κ∗remain at the default schooling level h= 0. 3.3.2 Searching Status Once schooling is completed, agents take a draw from the self-employment income distribution R(y|h). Upon observing the draw, they decide whether or not to search for an employee job while also working as self-employed. Given the notation just introduced, the decision is equivalent to the following maximization: max{S(y, h), U(h)}. Since S(y, h) is monotone increasing in yand U(h) is constant in y, there exists a unique: y∗(h) : S(y∗(h), h) = U(h). The optimal decision rule is again a reservation value rule where only agents with y≥y∗(h) search for an employee job while also working as self-employed. 28Table B.1 in the Appendix summarizes the environment of the model and introduces the notation for the value functions. 16 3.3.3 Labor Market Dynamics Upon meeting a worker and observing the match-specific productivity x, the schooling level h, and the worker’s outside option Q(y, h), the firm chooses the formality status based on the following maximization: max f{F0[x, y, h], F1[x, y, h]}. Upon meeting a firm, the worker also observes the match-specific productivity xand the formality status proposal f. Worker and firm then engage in bargaining to determine the wage and to decide to accept the match or not. We assume the axiomatic Nash bargaining solution, which is equivalent to solving: max w|f{Ef[w, y, h]−Q(y, h)}αh{Ff[x, y, h]−V[h]}(1−αh). To define equilibrium conditions and optimal decision rules, it is useful to start from firms’ entry decisions. Since the arrival rate of offers to a given firm is decreasing in the number of firms entering the market, the value of posting a vacancy V[h] is monotone decreasing in vh. We assume free entry of firms in both markets. As a result, firms enter until the value of posting a vacancy reaches zero: V[h] = 0.(15) Imposing condition (15), the Nash bargaining framework leads to the following wage schedules: w1(x;y, h) = αh 1 + tx+(1 −αh) (1 + β1,hτt)[ρQ(y, h)−β1,hb1] (16) w0(x;y, h) = αh(1 −c)x+ (1 −αh)[ρQ(y, h)−β0,hB0].(17) The wage schedules have the usual structure generated by Nash bargaining in this context: they are a convex combination of the match-specific productivity values xand the values of the worker’s outside option ρQ(y, h) (recall that by (15) the firm’s outside option is zero). The higher the working bargaining coefficient αthe higher the weight on x. On top of this usual structure, the two wage schedules show the impact of the institutional parameters. Both the contribution rate tand the cost of hiring informally care partially transferred to the worker implying a negative relationships with wages at any x. The non-wage benefits of the employment relationship (β1,hb1 and β0,hB0) also decrease wages at any xsince the benefits are valued by the worker. Solving backward, we find the match-specific productivity value that makes the firm indifferent between posting a formal or an informal job: ˜x(y, h) : F0[˜x(y, h), y, h] = F1[˜x(y, h), y, h]. 17 Both F0and F1are linearly increasing in x, but F1is increasing faster.29 As a result, for any {y, h} there exists a unique ˜x(y, h).30 By equations (16)-(17) and the definitions of the value functions, we can compute its value and obtain: ˜x(y, h) = 1 ch [β0,hB0−φhβ1,hb1+ (φh−1)ρQ(y, h)] (18) where : φh≡1 + t 1 + β1,hτt;φh∈[1,1 + t]. Since the value of accepting the match is increasing in xfor both workers and firms, for any {y, h} there exist two unique productivity reservation values at which workers are indifferent between accepting the firm’s offer or continuing to search for a better match, and analogously firms are indifferent between filling the vacancy or not: x∗ 0(y, h) : F0[x∗ 0, y, h] = 0 ⇐⇒ E0[w0[x∗ 0(y, h)], y, h] = Q(y, h), x∗ 1(y, h) : F1[x∗ 1, y, h] = 0 ⇐⇒ E1[w1[x∗ 1(y, h)], y, h] = Q(y, h). The agreement result is assured by the axiomatic Nash bargaining solution. By the definition of the value functions and wage schedules (16) and (17), we obtain: x∗ 0(y, h) = 1 1−ch [ρQ(y, h)−β0,hB0],(19) x∗ 1(y, h) = φh[ρQ(y, h)−β1,hb1].(20) Equations (19)-(20) state that job formality status f∈ {0,1}has two opposite effects on the reservation productivity values at which the match is formed. It decreases the reservation value because employees receive additional benefits associated with the match (b1or B0), but it also increases the reservation value because the firm faces some costs (tor c) in order to activate one job contract or the other. As a result of these two effects, the equilibrium is characterized by different optimal decision rules depending on parameters and on {y, h}. Still, all the decision rules retain the reservation value property. We summarize this property in the following: Proposition 1 Equilibrium Characterization: optimal decision rules. There are only two possible decision rules, for any y, h: 29The proof is in Section A.1 of the Appendix. The intuition for this result is straightforward since the cost of signing an informal job contract is linearly increasing in c, while the cost of signing a formal job contract increases in xonly through the wage schedule (16). 30Notice that ˜x(y, h) only guarantees indifference on the firms’ side but not necessarily on the workers’ side. This is a direct implication of the “formality posting”assumption. 18 If ˜x(y, h)> x∗ 1(y, h): x < x∗ 0(y, h)⇐⇒ {Q(y, h); 0} x∗ 0(y, h)≤x < ˜x(y, h)⇐⇒ {E0[w0(x), y, h]; F0[x, y, h]} ˜x(y, h)≤x⇐⇒ {E1[w1(x), y, h]; F1[x, y, h]} If ˜x(y, h)≤x∗ 1(y, h): x < x∗ 1(y, h)⇐⇒ {Q(y, h); 0} x∗ 1(y, h)≤x⇐⇒ {E1[w1(x), y, h]; F1[x, y, h]} The proof of this result and the formal definition of the equilibrium are reported in Sections A.1 and A.2 of the Appendix.The intuition is clarified in Figure 2, where we depict the equilibrium case where ˜x(y, h)> x∗ 1(y, h). For low values of the match-specific productivity, firms prefer to keep the vacancy open. For intermediate values (x∗ 0(y, h)≤x < ˜x(y, h)), firms post informal job offers that are accepted by workers receiving a wage governed by (17). For larger values (˜x(y, h)≤x), firms post formal job offers that are accepted by workers receiving a wage governed by (16).31 3.4 Empirical Implications The equilibrium of the model is able to replicate and explain the main empirical evidence that characterizes labor markets with high informality, including those for Mexico as described in Section 2. The first set of stylized facts is the significant mass of workers in each labor market state and the significant amount of transitions between formal and informal status. In the model, individuals can accept jobs with different formality status as a result of different values of matchspecific productivity. Since they receive different draws of match-specific productivity in their labor market careers, some draws may lead to formal jobs and others to informal jobs generating the transitions that we observe in the data. That all labor market states are relevant and filled in equilibrium depends on parameters values. In Section 5, we show that an estimated version of the model on Mexican data delivers proportions that closely match the data. The second set of stylized facts refers to the wages distributions. Formal employees have on average higher wages than informal employees, but the two wage distributions overlap over a large portion of their support. Both results are delivered by the reservation match productivity value being higher for formal employment (Proposition 1) and by the two wage schedules being both monotonically increasing in the match productivity value but at different rates (equations (16) and (17)). Conditioning on the value of the outside option, the average productivity of the accepted matches between workers and firms is higher in formal jobs than in informal ones. This is a direct 31It is also possible that the reservation value ˜x(y, h) is negative or, equivalently, that ˜x(y, h)≤x∗ 1(y, h). This case is easy to see in Figure 2 by shifting up the Ff= 0 axis. When this is the case, there are no values of xthat induce the firm to post an informal job and only formal jobs will be realized in equilibrium. 19 result of the optimal decision rule: informal jobs are created when x∗ 0(y, h)≤x < ˜x(y, h), formal jobs when ˜x(y, h)≤x. Since wages are monotonically increasing in x, the productivity differential is typically enough to generate the observed ranking in accepted wages. The economic reason for the overlap between the two wage distributions is more elaborate but equally intuitive. Formal employees earn a lower net wage than informal employees with the same productivity because they receive higher non-wage benefits – i.e., β1,hB1[w1(x;y, h)] is larger than β0,hB0.32 Figure 3 shows, for a given outside option (y, h), the wage schedules for formal and informal employees as a function of the match value x.33 Define the reservation match values x0(y, h) and x00(y, h) as: x0(y, h) : w0(x0(y, h); y, h) = w1(˜x(y, h); y, h) (21) x00(y, h) : w1(x00(y, h); y, h) = w0(˜x(y, h); y, h).(22) then all the x∈[x0(y, h),˜x(y, h)] generate an informal employment relationship with accepted wages in the interval [w1(˜x(y, h); y, h), w0(˜x(y, h); y, h)]. At the same time, all the x∈[˜x(y, h), x00(y, h)] generate a formal employment relationship with accepted wages exactly in the same interval. As a result, accepted wages in formal and informal employment will overlap over the support [w1(˜x(y, h); y, h), w0(˜x(y, h); y, h)]. However, this support may be too tight to generate the large overlap we observe in the data. The larger overlap, potentially able to cover the entire support, is delivered by the heterogeneity in the value of the searchers’ outside options. All the agents searching as unemployed – i.e., such that y < y∗(h) – generate one unique overlap because their value in the searching state is identical. But all the agents searching as self-employed – i.e., such that y > y∗(h) – generate different overlaps because their value in the searching state is a function of y. The larger the y, the larger the reservation value ˜x(y, h), the more to the right the location of the overlap. Figure 4 shows these features on simulations based on estimated parameters. The left panel shows the overlap considering only workers transiting from unemployment to formal and informal employment. The overlap is present, but it is limited to a relative narrow portion of the support. The bottom panel considers only workers transiting from self-employment to formal and informal employment. As expected, the overlap is much larger, covering the entire support of the accepted wage distributions. Mixing over the two generates the wage ranking and the overlap observed in the data. 32Almeida and Carneiro [2012] emphasize the same argument in an application focusing on enforcement of labor regulations in Brazil. Their empirical results based on regional variations in inspections is consistent with our empirical implications: the jobs more susceptible to switching from formal to informal are those that are relatively lower paid. 33In the figure, the informal wage schedule is more sensitive to xand has higher intercept: this is not always the case, but it is the case for the combination of parameters that better matches the data. The slope of w0(x;y, h) is steeper when the cost cwith respect to the contribution rate tis small enough (formally, when c < t/(1 + t)). This condition is always satisfied at our parameter estimates, and its violation leads to a proportion of informal workers which is in general too low to fit the data. The intercept of w0(x;y, h) is higher when the valuation of the non-contributory benefit is small enough with respect to formal contributory benefits (formally, when 1 (1+β1,hτt)[ρQ(y, h)−β1,hb1]<[ρQ(y, h)−β0,hB0]).. . Again, this is what we find at our parameter estimates for most of the (y, h) combinations. It may be violated without major changes in the argument. 20 The last set of stylized facts refers to the differences between the unemployed and the selfemployed. The differences involve both transition rates out of the state and accepted employee wages once the transition has taken place. As observed above, the reservation productivity value and value while searching are the same among the unemployed but depend on yamong the selfemployed. This difference is enough to deliver different hazard rates and different accepted wages as shown by Proposition 1 and equations (16) and (17). 4Identification The data available for identification were presented in Section 2.2 and include both individual-level data and aggregate-level data. The individual-level data can be described by the following set: {w0(i;h); w1(i;h); y(i;h); tU(i;h); tS(i;h)}n i=1, where idenotes individual observations; w0,w1and yare, respectively, hourly wages for informal employees, formal employees and self-employed workers; and tUand tSare monthly durations in unemployment and self-employment, respectively. We observe the same set of variables in both schooling groups h∈ {0,1}. The aggregate-level data we use in the identification and estimation of the model are the schooling-specific vacancy rates vhand the economy-wide labor shares. Finally, we exploit the municipality-level rollout of the Seguro Popular program. In an institutional context that allows for the observation of {B0, τ, t}we need to identify the following set of parameters: {λh, γh, ηh, αh, ρ, ξh, β0,h, β1,h, ch, ψh, ιh, ζh, νh} and the following probability distribution functions: {G(x|h), R(y|h), T (κ)}.(23) Notice that the mobility parameters λh,γhand ζhare not really primitive parameters of the model since they can be obtained from the tightness and the matching function parameters ψhand ιh – see equation (5). However, it is convenient to keep them as separate parameters to ease the exposition of both the identification and the estimation strategy as well as to facilitate comparisons with previous work. We split the identification discussion into four parts. We first discuss the usual search, matching and bargaining parameters. We then focus on the preferences for social security benefits and the cost to firms of hiring informally. In the third part we consider the identification of the matching function and the other demand side parameters. We conclude by discussing the cost of schooling parameters. 21 As mentioned in the identification strategy, given a consistent estimator ˆ θ, the parameters can be consistently estimated by solving equations (28), (29) and (30) and by applying the definition of ζh. This is the procedure we follow in this second estimation step. We now focus on the choice of moments to be used in the quadratic form (31). We choose the moments in order to capture the stylized facts described in Section 2.2 and to describe in detail the data features we need from the identification strategy. We start by dividing the sample observations into four groups, where each group is defined by schooling level and by exposure to the Seguro Popular program (see Sections 2 and 4.2). For each of these four groups, we build moments derived from the proportions in each labor market state, from the durations in the searching states, from the wages at formal and informal jobs, and from the self-employment income. For durations, we compute means; for wages and incomes we compute means and standard deviations. To capture the overlap between the distributions of formal and informal wages, we follow the procedure proposed by Flabbi and Moro [2012] to address a similar problem. We compute quintiles over the distribution of accepted wage for formal workers. For each interval, we compute: i) the mean wage of informal employees, ii) the mean wage of informal employees and iii) the proportion of employees in informal jobs earning a wage in that interval. Finally, we compute the aggregate labor share necessary to identify and estimate α.42 5.2 Results The estimated parameter values are reported in Table 2. The values of parameters governing the rates of job arrival and termination {λh, γh, ηh}are comparable to previous estimates for similar models on high-income countries.43 There are differences between the two schooling groups, with lower arrival and termination rates for individuals who did not complete a high school degree.44 The differences in arrival rates between the unemployed and the self-employed are very large, explaining in part the observed persistency in the self-employment state and the high-turnover in the unemployment state. For example, an unemployed worker belonging to the Low Schooling group meets a firm on average every 3.5 months, while a self-employed worker on average does so at a frequency of one time more than every 2 years. Taking into account the endogenous acceptance probability, these rates translate in unemployed workers belonging to the Low Schooling group accepting a job after on average 5.1 months. Unemployed workers belonging to the High Schooling group receive more offers, but they are pickier, leading to a slightly higher average duration in unemployment (5.7 months). These durations result from transitions to either an informal job or to a formal job. The composition of these transitions is where the main difference between the 42The complete set of 108 sample moments (27 micro-moments defined in each of the four groups plus one aggregate moment) – along with the simulated moments at the estimated parameters and the corresponding weight used in the quadratic form – is reported in Appendix B, Tables B.3 and B.4. 43See for example the review in Eckstein and van den Berg [2007] and specifically models of individual search without on-the-job search such as Flinn and Heckman [1982], Flinn [2006] and Flabbi and Moro [2012]. 44A similar ranking by schooling levels is found in Flinn and Mullins [2015] (under the No renegotiation specification) and in Flabbi and Leonardi [2010], even if both papers use US data and define schooling levels differently. 28 two schooling groups rests. Both groups have a higher probability to accept a formal employee job than an informal one but the extent of the difference is much larger in the High Schooling group. The probability of accepting a formal job compared to an informal one is eight times higher in the High Schooling group but only two times higher in the Low Schooling group.45 Important differences between the two schooling groups are also observed in the estimated values of the parameters of the match-specific productivity distribution {µx,h, σx,h}and the selfemployed earning distribution {µy,h, σy,h}. As reported in the bottom panel, average productivity in the High Schooling group is about 6.3% higher than in the Low Schooling group when working as an employee and about 18.2% higher when working as self-employed. These differences in productivity – along with the differences in the mobility parameters and in the valuation of the benefits – generates differences in labor market performance that affect the returns to investing in additional schooling. We focus on this feature in Section 5.3. The Nash bargaining coefficient αis estimated at 0.48, giving a slightly weaker bargaining position to the worker, but quite close to the value of 0.5 that characterizes symmetric bargaining. This value is higher but comparable to those estimated on US data using a similar identification strategy.46 The estimated values of the preference parameters {β1,h, β0,h}show that both formal and informal benefits are valued less than the monetary value used to provide them. However, the valuation of the non-contributory benefit is close to full monetary value (about 91 cents to the peso for both schooling groups) while the valuation of the contributory benefit is much less than full value (about 67 cents to the peso for the High Schooling group, and about 56 for the Low Schooling group). Since this implies that formal employees have a willingness to pay for the benefit significantly lower than the contribution paid to receive it, the payroll contribution rate introduces a net loss that reduces the incentive to work formally. Based on parameters estimates and equilibrium matches, the worker at the average accepted formal wage in the Low Schooling group pays about 7 pesos in payroll contributions while receiving a monetary benefit of about 8. But that amount of benefit is valued less than 5 pesos by the worker, leading to a loss of about 12% of the average wage. In the High School group, the average loss is similar – about 11% – because the higher loss due to redistribution is compensated by the higher valuation of the benefit. We speculate about two possible explanations of why the valuation of the benefits is close to full value for informal workers and much lower for formal workers. The first relates to the contributory nature of the benefit. Formal employees contribute a proportion of their wages to 45These results are different from the estimates obtained by Meghir et al. [2015] on Brazil where they estimate on the basis of an equilibrium search model that it takes on average three years to transit to a formal job from unemployment but only a few months to transit to an informal one. The main reason for the difference – on top of some country-specific factors – is the definition of informality. Meghir et al. [2015] do not differentiate between self-employed and informal employees in their definition of the informal sector or in their modeling and estimation strategy. 46Flinn [2006] estimates αin the range of 0.39-0.43 on a sample of young low skilled US workers. Flinn and Mullins [2015] estimate it at 0.25 on a sample of very young US workers (25 to 34 years of age) but with a broader range of skills. 29 obtain the benefit while informal employees, self-employed workers, and unemployed workers do not. As a result, the attitude toward the service provided and its valuation may be different, even if the quality is comparable. The second possible interpretation relates to the composition of the benefit. The extra-wage benefits in the formal sector, B1(w1(x;y, h)), bundle together two types of benefits: a retirement benefit and a health benefit – see equation (2). Only the second benefit (b1) is comparable to the non-contributory benefit received by informal and unemployed workers, B0, but the valuation parameter is estimated over the bundle of both types of benefits. If retirement benefits are valued less than health benefits, then β1,h should be lower than β0,h. The estimates of the cost of hiring informally parameter chare 10.9% of job productivity in the Low Schooling group and 8.5% of job productivity in the High Schooling group. The parameter captures all the costs associated with hiring informally, including the probability and penalty of getting caught. While the estimated cost is economically important, at a relatively low productivity level – for given self-employed income, y– it is still lower than the cost of hiring formally, justifying the significant but not dominant presence of informal employees observed in our labor market. For example, the cost of hiring informally as measured by chxat the mean productivity of the realized informal matches is between 2.4 and 2.2 pesos per hour; the cost of formality as measured by tw1(x;y.h) at the mean productivity of the realized formal matches is between 7 and 9.1 pesos per hour.47 The flow value of being an unemployed searcher ξhis estimated to be negative in both schooling groups. A negative value was expected in order to generate enough wage dispersion in accepted wages.48 However, the value is not as unrealistically low as in previous work due to the presence of a searching state that generates labor income – the self-employed state – and due to the provision of non-contributory social security benefits. The last set of parameters in Table 2 refers to the matching function, the demand side and the schooling decision. The first matching function parameter ψhrepresents the lower search efficiency of self-employed workers with respect to unemployed workers. If we interpret it as the time spent searching by each self-employed worker and we assume that an unemployed worker searches fulltime, the estimated values imply between four and five hours a week devoted to job search by the average self-employed worker. The second matching function parameter ιhrepresents the elasticity of the number of jobs with respect to the measure of searchers – see equation (5). Our estimated values are lower than those estimated using macro data on high-income countries [Petrongolo and Pissarides, 2001] and on Mexico [Arroyo Miranda et al., 2014] but not far from the 0.5 frequently used in calibration. Our estimated matching function also implies an arrival rate of workers to firm ζhof about 1.8 in the Low Schooling group and of about 2.1 in the High Schooling group. The other demand side parameter is the flow value of keeping the vacancy open νh. We estimate that keeping a vacancy open and unfilled is 23% more costly when the job must be filled by a 47Notice that these are only direct costs, i.e., they do not take into account that through bargaining firms are able to partially transfer them to the workers, as seen in the equilibrium wage schedules (16) and (17). 48See Hornstein et al. [2011] for an extensive treatment of the issue. 30 more educated worker, a ranking we find reasonable. Finally, δis the parameter of the cost of schooling distribution. As reported at the bottom of Table 2, it implies that the average cost of completing High School with respect to stopping at Junior High is about 137.6. This value should be compared with the overall values of participating in the market Z(h). These values are reported at the bottom of Table 3 and they are between 300 and 400. As a result, the average cost of acquiring additional schooling is about one third of the value of participating in the labor market as a High Schooling type. This relatively high cost is consistent with less than 40% of the population completing additional schooling. 5.3 Returns to Schooling Table 3 reports various outcomes useful to interpret the returns to schooling in our context. The first panel reports accepted wages and it is similar to common measures obtained by comparing conditional means or by estimating wage equations. Completing High School, with respect to completing at most Junior High School, increases average accepted wages by 22.2% when working as informal employee, by 28.9% when working as formal employee, and by 29.4% when working as self-employed. However – as previously pointed out by Eckstein and Wolpin [1995] and following literature – accepted earnings are not an appropriate measure of returns since they are selected by the decision of accepting or rejecting a given job match. A more appropriate measure can be obtained by looking at offered wages.49 We report mean offered wages in the second panel of Table 3. Here, the ranking of the returns is different, with the return in formal employee jobs (11.8%) being lower than the one in informal employee jobs (15.3%) or in self-employment (18.2%). The result shows that the extent of the higher selectivity of accepting a formal job is different between the two schooling groups. Finally, we can provide a measure of returns that summarizes not only wages and self-employment incomes but also labor market frictions, the probability of job termination, the selection over labor market states, and the valuation of non-wage benefits. As seen in Section 3.3, this is the relevant measure when deciding to complete additional schooling or not. It is the measure that summarizes the value of participating in the labor market as a High Schooling worker or as a Low Schooling worker, and we have denoted it in the paper with Z(h). The overall estimated return to completing High School with respect to Junior High is therefore 20.7%, as reported in the bottom panel of Table 3. 49Notice that offered wages by formality status may be defined in different ways in our context. We have decided to exclude any endogenous truncation in computing these measures: they are simply the mean offered wage obtained by integrating the wage schedules (16) and (17) over the primitive productivity distribution G(x|h). That is the reason why the average wage offer in informal jobs may be higher than the average accepted wage in informal jobs: the first measure uses the entire support of xwhile the second measure has support with upper bound at ˜x(y, h). 31 5.4 Model Fit Tables B.3 and B.4 in the Appendix report the complete set of moments targeted by the MSM estimator. They are computed separately for individuals belonging to municipalities exposed to the Seguro Popular program and for those that are not. They are also unconditional to the labor market state to guarantee a smoother and well-defined quadratic form during the optimization procedure. There are no major mismatches in the moments targeted by the procedure but some data features are captured better than others. To ease the discussion and the interpretation, Table 4 reports moments aggregated over the differential exposure to the Seguro Popular program but conditional on the labor market state. The distribution over the four labor market states is well matched by the simulated data on both schooling groups with the exception of the unemployment state. The unemployment rate we generate at estimated values is lower than the one observed in the data, in particular for the Low Schooling group. While this is a concern, unemployment is the least relevant of all labor market states since both the rates and the durations are relatively low. The second set of moments in Table 4 reports means and standard deviations of the accepted wages and of the self-employment incomes. The match is quite good on all the means, with differences in the 10% range. Standard deviations on informal employees, instead, are lower than in the sample in both schooling groups. This is a direct consequences of the constraints imposed by the theoretical model: informal wages have a bounded support, with lower bound at w0(x? 0(y, h); y, h) and upper bound at w0(˜x(y, h); y, h). While this constraint does not necessarily prevents us from reaching a good fit, it is not too surprising that this particular data feature is the one we have the most difficulty in replicating. A peculiar and relevant feature of Mexico’s and other labor markets with high informality is the overlapping of the formal and informal accepted wages distributions. We are able to replicate the overlap quite well, both in terms of the proportions of informal employees in each quantile and in the mean accepted wages by quantiles. This result is obtained through two channels. The first is the endogenous mapping between match-specific productivity and wages implied by bargaining. The second channel is the flexibility introduced by allowing the self-employment state to be a searching state, with heterogenous productivity levels that are pinned down by the (observed) income generated while in self-employment. 6Policy Experiments Our model incorporates the structure of the social security system implemented by several countries in response to the lack of coverage for informal workers. The resulting dual system is characterized by contributory benefits – governed by a payroll tax rate, a benefit level increasing in wages and a redistributive component – and non-contributory benefits. The estimates of the preferences over these benefits indicate that contributory benefits generate a net loss, reducing the incentives to work formally, while non-contributory benefits amount to a net subsidy, increasing 32 the incentives to work informally. The impact is mediated by the redistributive component of the contributory system, which favors formality on the range of productivities most likely to be at the margin between formality and informality. To evaluate the impact of this complex system of incentives and disincentives, we use the estimated model to generate counterfactual experiments that varies the policy parameters of the social security system. Thanks to the structure of the model, we can evaluate the impacts in an equilibrium context, looking at labor market outcomes, informality levels, schooling rates and overall productivity and welfare. We report results on two crucial policy parameters: the payroll tax rate in formal jobs tand the per capita level of non-contributory social benefits B0. First, we focus on each of the two policy parameters, changing their values over a large neighborhood of the benchmark values. We present the impact on informality rates in Section 6.1 and the impact on schooling and productivity in Section 6.2. Then, we focus on a policy combining both parameters and setting them at levels that are both realistic in terms of policy implementation and able to generate better incentives in terms of labor market outcomes. We call this policy universal social security benefit and we discuss it in Section 6.3. The policy experiments procedure works as follows. For each value of the policy parameter, we find and compute the new equilibrium holding fix the other institutional parameters and the estimated parameters. Then, we simulate the labor market careers for 20,000 individuals (10,000 individuals in each schooling group) in these counterfactual labor markets. Finally, we compute the relevant statistics on the simulated data. 6.1 Labor Market Informality Figure 6 reports simulation results on the equilibrium rate of informality. In the top panels, we compute the rate only as proportion of informal employees, while in the bottom panels we also include the self-employed workers. Panels (a) and (c) of Figure 6 report the impact of changes in the payroll tax rate tfor formal employees. The informality rate increases quickly as the rate moves from zero to positive, the effect predicted by most of the existing literature. However, above a threshold quite close to the rate charged in the Mexican system (the vertical line), an additional increase in tleads to a decrease in the informality rate. The effect is particularly pronounced when focusing only on informal employees (top panel). The reason for the non-monotone impact of the payroll tax rate on informality is the redistributive feature of the social security system. As described in Section 3, a proportion τof the contributory benefit is increasing in the contribution (the retirement benefit) but the rest is redistributed equally among all the formal employees (the health benefit b1). As a result, an increase in timpacts informality through two main channels. The first channel makes formal jobs less attractive across the board because the firm has to pay a higher contribution for each level of productivity. The second channel makes formal jobs more attractive at relatively low productivity levels because the workers filling these jobs receive a relatively larger portion of 33 their benefit through b1. Since these workers and jobs are exactly those at the margin between formality and informality, the second effect may dominate when a high contribution rate makes the transfer through b1generous enough. Panels (b) and (d) of Figure 6 report the impact of changes in the non-contributory social security benefit B0. As expected, the effect is monotone in this case but the elasticities indicate interesting dynamics. Setting the non-contributory benefit to zero would be enough – in this labor market – to completely eliminate informal employees. This is a policy-relevant result since many countries in the region are attempting to eliminate informality by increasing enforcement while at the same time adding resources to non-contributory benefits.50 The sensitivity of the informality rate to changes in the benefit increases non-linearly with the benefit’s level and it becomes large exactly in the area where the benchmark Mexican values are.51 The impact on the overall informality rate as measured by the sum of illegal employees and the self-employed is still monotone but less sensitive to the policy. In particular, a benefit set at zero will not eliminate self-employment. The reason is the relative productivity of different workers in the employee sector and in the self-employment sector. Since individuals are heterogenous in their ability to generate self-employment income, particularly productive individuals in the sector will always choose to spend at least some time in self-employment. This result is informative about the distinction that we introduce in the paper between informal workers working as an employee or as self-employed. Since these two groups of workers respond to different incentives, the policy instruments suitable for targeting one group or the other should take this into account. 6.2 Schooling and Productivity Figure 7 reports simulation results of the impact of the same policy changes on schooling levels and the overall value of production. Schooling levels are measured as the proportion of individuals acquiring the high schooling level (High School completed). The value of production include the productivity of all the realized matches in equilibrium (the x’s above x?(0, h)) and of all the selfemployed active in equilibrium (the y’s above y?(h)). We normalize the value of production in the benchmark case to 1, and we aggregate over the two schooling levels. Each panel presents two lines. The dashed line reports the general equilibrium case, i.e., the full model where the demand side is taken into account by allowing firms to post more or fewer jobs in each of the two schooling markets as a result of the policy changes. In this case, the contact rates are endogenous and governed by the matching function (5). The solid line reports the partial equilibrium case, which is frequently studied in search models estimated on individual-level data, i.e., a model in which the contact rates are policy-invariant and set at the benchmark values ˆ λhand ˆγhreported in 50Improving enforcement of the labor regulations is the policy studied by Meghir et al. [2015]. Since their model does not incorporate the details of the dual social security system, their setting cannot generate a policy experiment such as the one conducted in this section. 51The two vertical lines correspond to the benefit level of individuals in municipalities with and without Seguro Popular. 34 Table 2. This comparison is informative about the quantitive importance of a general equilibrium approach in evaluating policies.52 As expected, an increase in the payroll tax rate of formal employees decreases overall schooling monotonically (Panel A). Since higher schooling levels means higher wages and productivity and since the payroll tax rate is proportional to wages, but the contributory benefit less so (again, due to the b1component), an increase in the payroll tax rate affects proportionally more the High Schooling group. As a result, equilibrium returns to schooling decrease and individuals acquire less education. This is a potentially very important but frequently neglected dimension of social security systems in a context of high informality. Indeed, the effect in general equilibrium is very large: increasing the payroll tax rate from zero to about 60% leads to cutting high schooling rates in half, from about 70% to about 35%. In partial equilibrium, instead, the impact is relatively small: the same change leads to cutting high schooling rates only by a few percentage points. The source of this stark difference lies in the reaction of the demand side. A rate increase affects the firms’ decision to post formal or informal jobs, magnifying the impact of the policy on the equilibrium returns to schooling. The positive externality of this effect is also reflected in the overall value of production. Lowering tcan potentially increase production by reducing distortion but also by inducing more individuals to acquire additional schooling and therefore becoming more productive. In fact, lowering the payroll tax rate to zero would increase the value of production by about 17% in general equilibrium but only by about 6% in partial equilibrium. However, the sensitivity to a further increase of the payroll tax rate from the benchmark level (the vertical line in Panels A and C) would be quite low both in general and partial equilibrium since the curves are relatively flat in that region of the tsupport.53 Changes in non-contributory benefits B0generate a similar results (Panels B and D). Schooling levels and productivity are monotone decreasing in the benefit’s amount, and the general equilibrium impacts are much larger than the partial equilibrium ones. The reasons are also similar: since the benefits are the same for all, they favor relatively more the low schooling individuals. Since firms decide posting by schooling, the difference is magnified in general equilibrium. The large impact on schooling levels in equilibrium is worth emphasizing: the fraction of workers completing secondary schooling almost doubles with respect to its benchmark value – from 39.5% to roughly 70% – when the benefit is completely eliminated. The impact of non-contributory benefits on long-run investment is frequently ignored in the policy debate, and it is shown here – conditionally on our model and estimates – to have potentially large effects. The result highlights the trade-offs faced when expanding social security benefits to informal workers through B0. On one hand, workers have better coverage; on the other, formal employment and output fall, and so 52Examples of this partial equilibrium approach are: Flinn and Heckman [1982]; Eckstein and Wolpin [1995]; Dey and Flinn [2005]; Flabbi and Leonardi [2010]. 53This channel is clarified by looking at Figure B.1 in the Appendix. The figure reports the arrival rates of employee job offers to the unemployed (top panel) and to the self-employed (bottom panel). In partial equilibrium, they are fixed to the benchmark values while in general equilibrium they change as a result of the searchers/vacancies ratio. The arrival rates of offers to the High Schooling group is much higher for smaller values of tand B0. This significantly improves the labor market opportunities of the more educated. 35 does the fraction of workers acquiring the High Schooling level. 6.3 Universal Social Security Benefits As shown, each of the policy parameters has significant effects on labor market outcomes, schooling investment decisions, and aggregate productivity. A combination of both policy instruments is promising in improving outcomes from current levels and in balancing incentives and equity objectives. Instead of attempting an optimal policy design, we simulate policies that may solve one of the main distortions of the system: the “dual” nature of the social security benefits where contributory benefits in the formal sector coexist with non-contributory benefits in the informal sector and redistribution occurs only within the formal sector. The objective is to mimic a universal social security system where benefits are not linked to the formality status of the job. The first policy consists in simply increasing the level of per capita non-contributory benefits to the equilibrium value of the lump-sum component of the social security benefits in the formal sector (B0=b1= 4.04 pesos per hour). Again, this is a policy that is currently implemented in Mexico and many other countries (see Section 2). The second policy achieves the same objective but introduces changes in the contributions paid in the formal sector in order to reduce the distortions in the choice of the formality regime. To achieve this, we eliminate the redistributive component of the contributory social security system and we provide some non-contributory benefit to formal employees. Specifically: i) all the contributions in the formal sector are retirement benefits (τ= 1), ii) all the health benefits are non-contributory for all workers in any formality regime and are set to the benchmark formal employee level (B0=b1= 4.04 pesos per hour), and iii) the payroll tax rate for formal jobs is reduced accordingly (t= 0.33 ×0.45 = 0.15). Table 5 reports the results of the policy experiments on labor market outcomes for the lowschooling group (Panel A), for the high-schooling group (Panel B) and for aggregate benefits, schooling, productivity and costs (Panel C). The first column reports statistics on the benchmark economy, the second column on the experiment that simply equates the benefit, and the third column on the policy that restores the full proportionality between the benefits and the contributions for formal employees. The first policy creates an incentive to work informally: the relative fractions of informal employees with respect to formal employees increases substantially in both schooling groups. The difference between the match-specific productivity cutoff value for accepting an informal job and the cutoff value for accepting a formal job widens, creating significant gaps in the average accepted wages in equilibrium between formal and informal employees. The value of participating in each schooling-specific labor market goes up since the system is receiving an increase in non-contributory benefits. Since the benefits are not paid by the agents, the fiscal cost is high. Since favoring informality reduces the returns to schooling, overall schooling completed is reduced and overall production decreases. The second policy provides advantages for both formal and informal workers. In equilibrium, the advantage is higher for the formals since their benefits’ levels do not change while their payroll 36 tax rate is lowered. The incentives are enough to completely eliminate informal employment in the High Schooling level sector but have almost no impact on the informality level in the Low Schooling sector. However, the overall informality rate in the economy decreases substantially because the share of agents with High School completed increases by 30 percentage points, reaching almost 70%. The asymmetry of the effects by schooling is due to the complete removal of the redistributive component (τ= 1) and to firms’ reaction (higher vacancy posting in the High Schooling sector). What is notable is that the positive externality on schooling levels induces an increase in the value of production large enough to keep the relative fiscal cost of the benefit system almost constant. The relative fiscal cost only increases from 2.8% of overall value of production in the benchmark to 3.1% post-policy. As a result of the joint increase in benefits, in schooling, and in the value of production, overall workers’ welfare (Z) increases substantially, reaching a value 35% higher than in the benchmark model.54 7 Conclusion Informality is a defining feature of many labor markets. In Latin America, over 50% of the labor force work is employed informally. Studying costs and benefits of informality requires an equilibrium model of the labor market that takes into account how workers and firms endogenously sort between the formal and the informal sector. If the model wants to generate credible estimates and relevant counterfactual policy scenarios, it also needs to replicate the empirical regularities and the salient institutional features observed in these markets. This paper developed and estimated a search and matching model where firms and workers endogenously decide to form matches (jobs) that can be formal or informal. The model replicates the main features of labor market dynamics in Mexico by allowing endogenous formality posting and endogenous wage determination through bargaining. The sources of heterogeneity determining wages and formality status are the match-specific productivity and the income of self-employed workers. Meeting rates are also endogenous, as they are governed by the equilibrium proportions of (unemployed and self-employed) searchers and vacancies. In this environment, we introduce three relevant but neglected features. First, recognizing that the labor market distortions that generate informality may affect not only short-run labor market outcomes but also long-run investment decisions, we allowed for endogenous schooling decisions. Second, we modeled a crucial and increasingly important institutional feature: the presence of a dual social security system where non-contributory benefits targeting informal workers coexists with a standard contributory system reserved for formal workers. Finally, we introduced the crucial distinction between informal employees and informal self-employed workers. We identified and estimated the model parameters using a combination of individual-level data, 54Z(h) defines the value of participating in the labor market with a given schooling level h(see equation (6)). In Table 5, we report the average value over the two schooling group, where the average is weighted by the equilibrium proportion of agents that have completed the High or Low level of schooling. 37 Table 2: Estimates of the Model Parameters Low Schooling: h= 0 High Schooling: h= 1 Coeff. Std. Error Coeff. Std. Error Parameters: λh0.2890 0.0162 0.3597 0.0083 γh0.0279 0.0005 0.0354 0.0009 ηh0.0071 0.0004 0.0102 0.0008 µx,h 2.8114 0.0095 2.6116 0.0087 σx,h 0.8359 0.0172 1.1051 0.0120 µy,h 2.2615 0.0145 2.4129 0.0125 σy,h 0.7120 0.0072 0.7338 0.0107 α0.4813 0.0135 0.4813 0.0135 β1,h 0.5615 0.0034 0.6705 0.0101 β0,h 0.9166 0.0055 0.9082 0.0081 ch0.1089 0.0026 0.0856 0.0014 ξh-17.509 0.6445 -20.141 0.7263 ψh0.0965 0.0061 0.0983 0.0038 ιh0.3157 0.0494 0.4218 0.0365 ζh1.7730 0.1721 2.1080 0.1857 νh-79.649 4.8504 -100.75 9.4442 δ0.0073 0.0011 0.0073 0.0011 Predicted Values: Eh(x) 23.59 0.4607 25.08 0.3271 SDh(x) 23.72 1.1345 38.79 1.1453 Eh(y) 12.37 0.1879 14.62 0.1845 SDh(y) 10.05 0.2254 12.35 0.3151 E(k) 137.62 17.062 137.62 17.062 Note: Bootstrap standard errors based on 120 replications reported. For the definition of the parameters, see Section 3.1 and Section 4. All the parameters are schooling-specific with the exception of αand δ, which we report under both schooling columns at the jointly estimated value. Low Schooling is defined as having completed at most junior secondary (9th grade); High Schooling is defined as having completed at most high school (12th grade). 44 Table 3: Estimates of the Returns to Schooling Low Schooling High Schooling Relative h= 0 h= 1 Difference Accepted Wages and Income: Formal: Eh[w1|˜x(y, h)≤x]21.442 27.636 0.289 Informal: Eh[w0|x∗ 0(y, h)≤x < ˜x(y, h)] 17.396 21.253 0.222 Self-Employed: Eh[y|y∗(h)≤y]19.652 25.438 0.294 Offered Wages and Income: Formal: Eh[w1]15.226 17.021 0.118 Informal: Eh[w0]17.592 20.292 0.153 Self-Employed: Eh[y]12.366 14.617 0.182 Labor Market Value: Z(h) 329.60 397.73 0.207 Note: Simulated samples of 10,000 worker-level observations for each schooling group based on the estimates reported in Table 2. For the definition of Zsee equation (6). Low Schooling is defined as having completed at most junior secondary (9th grade); High Schooling is defined as having completed at most high school (12th grade). 45 Table 4: Model Fit: Conditional Moments Low Schooling High Schooling Moments Model Data Model Data Proportions: Formal Employees 0.512 0.500 0.546 0.523 Informal Employees 0.220 0.223 0.182 0.193 Self-employed 0.254 0.235 0.241 0.238 Unemployed 0.014 0.042 0.030 0.045 Wages and Income: Formal Employees: Mean 21.442 24.001 27.636 30.355 Formal Employees: Standard Deviation 12.742 11.926 19.555 18.309 Informal Employees: Mean 17.396 18.138 21.253 21.792 Informal Employees: Standard Deviation 8.810 9.908 10.746 15.291 Self-employed: Mean 19.652 21.591 25.438 24.054 Self-employed: Standard Deviation 13.077 12.683 14.659 15.448 Durations: Unemployed: Mean 5.063 2.387 5.695 3.708 Self-employed: Mean 113.1 133.7 112.5 122.9 Quintiles: Proportions: Informal Employee - Q1 0.421 0.412 0.356 0.448 Informal Employee - Q2 0.151 0.243 0.277 0.233 Informal Employee - Q3 0.158 0.141 0.130 0.123 Informal Employee - Q4 0.155 0.123 0.137 0.096 Informal Employee - Q5 0.116 0.080 0.100 0.101 Mean Wages: Formal Employees - Q1 11.250 11.207 14.008 12.528 Formal Employees - Q2 14.196 16.901 17.212 19.070 Formal Employees - Q3 17.735 21.432 21.441 25.696 Formal Employees - Q4 23.162 27.506 28.637 33.767 Formal Employees - Q5 40.858 42.930 56.843 58.317 Informal Employees - Q1 11.132 10.463 14.537 11.359 Informal Employees - Q2 14.159 16.509 16.700 19.070 Informal Employees - Q3 17.728 21.104 21.648 25.696 Informal Employees - Q4 22.916 27.529 28.598 33.767 Informal Employees - Q5 36.599 42.773 47.266 58.317 Aggregate Moment: Model Data Labor Share 0.415 0.419 Note:Model columns report moments computed on the simulated sample of 10,000 worker-level observations for each schooling group based on the estimates reported in Table 2. Sample columns report moments computed on the estimation sample extracted from the four quarters of 2005 of the Mexican labor force survey (ENOE) and on 2005 AMECO data for the labor share. Low Schooling is defined as having completed at most junior secondary (9th grade); High Schooling is defined as having completed at most high school (12th grade). 46 Table 5: Universal Social Security Benefits Benchmark Policy 1: Policy 2: B0=b1= 4.04 B0=b1= 4.04 τ= 1; t= 0.15 Panel A: Low Schooling h= 0 Proportions: Self-employed 0.254 0.253 0.250 Informal Employee 0.220 0.478 0.217 Formal Employee 0.512 0.257 0.518 Unemployed 0.014 0.012 0.015 Mean Wages and Income: Self-Employed Income 19.652 19.810 20.314 Informal Employee 17.396 17.127 13.181 Formal Employee 21.442 27.801 26.217 Panel B: High Schooling h= 1 Proportions: Self-employed 0.241 0.256 0.122 Informal Employee 0.182 0.418 0.000 Formal Employee 0.546 0.297 0.827 Unemployed 0.030 0.029 0.051 Mean Wages and Income: Self-Employed Income 25.438 24.792 34.617 Informal Wages 21.253 20.206 0.000 Formal Wages 27.636 36.403 36.404 Panel C: Aggregate Statistics Social Security Benefits: Formal Lump-sum = b14.039 4.946 4.039 Formal Proportional (at mean w) = τ∗t∗¯w14.472 5.883 4.823 Informal and Unemployed = B02.257 4.039 4.039 Outcomes: Share with High Schooling 0.395 0.364 0.697 Labor Market Value = Z356.5 380.0 480.9 Value of Production (normalized) 1.000 0.985 1.167 Cost: Fiscal Cost = [B0/(Value of Production)] 0.028 0.080 0.031 Note: Simulation results computed on the simulated sample of 10,000 worker-level observations for each schooling group based on the estimates reported in Table 2. The benchmark case sets the institutional parameters at the values for the Mexican labor market in 2005 – see Appendix C for details. Policy 1 increases the monetary value of per-capita non-contributory social security benefits (B0) to the benchmark level of of the lump-sum portion of contributory benefits (b1). Policy 2 eliminates the lump-sum redistributive component (τ= 1), accordingly decrease social security taxes (t(1 −τ)), and provides non-contributory social security benefits (B0=b1) to all workers. 47 Figure 1: Observed Wages Density Functions 0.02 .04 .06 Frequency 020 40 60 80 100 120 Hourly Wages Formal Employees Informal Employees (a) Low Schooling 0.01 .02 .03 .04 .05 Frequency 020 40 60 80 100 120 Hourly Wages Formal Employees Informal Employees (b) High Schooling Note: Data extracted from the four quarters of 2005 of the Mexican labor force survey (ENOE). The Figure shows the empirical densities of the hourly wages (in Mexican Pesos). Low Schooling is defined as having completed at most junior secondary (9th grade); High Schooling is defined as having completed at most high school (12th grade). The Formal status of the job is defined according to whether or not workers report having access to health care through their employers. 48 Figure 2: Equilibrium Representation x Ff F0[x, y, h] F1[x, y, h] 0x x∗ 1 x∗ 0˜x Unfilled Filled Informal Filled Formal Note: Illustrative figure, not based on actual data. For the definitions of F0,F1,x∗ 0,x∗ 1, and ˜x, see equations (13), (14), (19), (20), and (18). 49 Figure 3: Wage Schedules and Overlap x ww0(x;y, h) w1(x;y, h) ˜x(y, h) w0(˜x;y, h) w1(˜x;y, h) x0(y, h)x00(y, h) Note: Illustrative figure, not based on actual data. For the definitions of w0(x;y, h), w1(x;y, h), x0(y, h), x00(y, h) and ˜x(y, h), see equations (17), (16), (21), (22), and (18). 50 Figure 4: Simulated Accepted Wage Distributions and Overlap 0.1 .2 .3 Frequency 010 20 30 40 50 60 Hourly Wages Formal Employees Informal Employees (a) Outside Option is Unemployment 0.02 .04 .06 .08 Frequency 010 20 30 40 50 60 Hourly Wages Formal Employees Informal Employees (b) Outside Option is Self-employment Note: Simulated sample of 10,000 worker-level observations for the high schooling group based on the estimates reported in Table 2. The Figure shows the empirical densities of the accepted hourly wages (in Mexican Pesos), separately for formal employees and informal employees. 51 Figure 5: Overlap and Identification of β1,h and ch x ww0 0(x;y, h) w0(x;y, h) w0 1(x;y, h) w1(x;y, h) ˜x(y, h) w0(˜x;y, h) w0 0(˜x;y, h) w1(˜x;y, h) w0 1(˜x;y, h) Note: Illustrative figure, not based on actual data. For the definitions of w0(x;y, h), w1(x;y, h), ˜x(y, h), see equations (17), (16), (18). The wage schedules resulting by changing β1,h and chare denoted by w0 0(x;y, h) and w0 1(x;y, h). 52 Figure 6: Policy Impacts on Informality Rate .05 .1 .15 .2 .25 0.2 .4 .6 .8 Low Schooling High Schooling Overall (a) Changes in t– Share of Informal Employees 0.1 .2 .3 .4 .5 0 1 2 3 4 Low Schooling High Schooling Overall (b) Changes in B0– Share of Informal Employees .1 .2 .3 .4 .5 0.2 .4 .6 .8 Low Schooling High Schooling Overall (c) Changes in t– Share of Informal Employment 0.2 .4 .6 .8 0 1 2 3 4 Low Schooling High Schooling Overall (d) Changes in B0– Share of Informal Employment Note: Top panel reports the proportion of informal employees, the bottom panel the proportion of informal employees and the self-employed. Simulated samples of 10,000 worker-level observations for each schooling group based on the estimates reported in Table 2. The Overall dotted line is computed as a weighted average between the the high schooling group and the low schooling group, where the weights are defined by the equilibrium frequencies. The vertical lines are set at the institutional values for the Mexican labor market in 2005. See Appendix C for details. 53 Table B.2: Roll-out of the Seguro Popular Program and Pre-determined Labor Market Characteristics (1) (2) (3) (4) (5) (6) (7) ln(wf) ln(wi) ln(wse) Formal Informal Self-Empl Unempl Seguro Popular in 2005 (1=yes) 0.014 0.083 0.043 -0.038 0.024 0.013 0.001 (0.042) (0.068) (0.060) (0.023) (0.017) (0.016) (0.003) Complete Secondary (1=yes) 0.187 0.182 0.189 -0.034 -0.015 0.048 0.001 (0.016) (0.034) (0.020) (0.007) (0.006) (0.008) (0.003) Mean Dep. Var. 0.494 0.154 0.327 0.024 Number of Obs 10077 3061 6534 20803 20803 20803 20803 Number of Clusters 217 190 217 238 238 238 238 Note: OLS estimates. Standard errors clustered at the municipality level are reported in parenthesis. Data is drawn from the Mexican labor market survey (ENE, 2001) and matched at the municipality-level with the roll-out of the Seguro Popular program. VI Table B.3: Unconditional Moments: Municipalities With Seguro Popular in 2005 Low Schooling High Schooling Moment Model Data Weight Model Data Weight Share Self-employed 0.259 0.241 0.005 0.241 0.238 0.007 Share Formally Employed 0.429 0.486 0.006 0.485 0.514 0.007 Share Informally Employed 0.297 0.236 0.006 0.245 0.206 0.006 Share Unemployed 0.016 0.037 0.003 0.028 0.042 0.003 Mean Informal Wages 5.043 4.252 0.114 5.074 4.480 0.179 SD Informal Wages 8.873 9.009 0.182 10.141 11.119 0.380 Mean Formal Wages 9.860 11.651 0.183 14.681 15.690 0.309 SD Formal Wages 14.195 14.615 0.168 21.229 20.294 0.348 Mean Self-empl Income 5.332 5.408 0.143 6.276 5.983 0.203 SD Self-empl Income 11.272 11.531 0.216 13.871 13.218 0.334 U Duration (months) 0.076 0.095 0.014 0.162 0.159 0.023 SE Duration (months) 30.965 33.017 0.996 27.432 29.786 1.116 Share Informally Employed - Q1 0.162 0.094 0.005 0.143 0.092 0.005 Share Informally Employed - Q2 0.037 0.061 0.005 0.033 0.046 0.004 Share Informally Employed - Q3 0.038 0.033 0.004 0.029 0.026 0.003 Share Informally Employed - Q4 0.034 0.029 0.003 0.025 0.020 0.003 Share Informally Employed - Q5 0.025 0.018 0.002 0.015 0.021 0.003 Mean Informal Wages - Q1 1.950 0.973 0.054 2.204 1.049 0.071 Mean Informal Wages - Q2 0.584 1.005 0.080 0.603 0.883 0.078 Mean Informal Wages - Q3 0.739 0.695 0.074 0.700 0.669 0.075 Mean Informal Wages - Q4 0.848 0.797 0.072 0.794 0.675 0.096 Mean Informal Wages - Q5 0.921 0.781 0.081 0.773 1.204 0.143 Mean Formal Wages - Q1 1.116 1.063 0.023 1.525 1.220 0.043 Mean Formal Wages - Q2 1.358 1.626 0.046 1.823 1.969 0.095 Mean Formal Wages - Q3 1.645 2.046 0.044 2.277 2.707 0.075 Mean Formal Wages - Q4 2.140 2.715 0.070 3.049 3.534 0.085 Mean Formal Wages - Q5 3.601 4.201 0.089 6.007 6.260 0.166 Aggregate Moment: Model Data Labor Share 0.415 0.419 VII Table B.4: Unconditional Moments: Municipalities Without Seguro Popular in 2005 Low Schooling High Schooling Moment Model Data Weight Model Data Weight Share Self-employed 0.267 0.224 0.007 0.244 0.239 0.010 Share Formally Employed 0.551 0.527 0.009 0.626 0.545 0.011 Share Informally Employed 0.171 0.198 0.007 0.100 0.165 0.009 Share Unemployed 0.011 0.051 0.004 0.030 0.052 0.005 Mean Informal Wages 3.281 3.632 0.152 2.422 3.607 0.243 SD Informal Wages 8.245 8.609 0.274 8.317 10.425 0.635 Mean Formal Wages 11.555 12.675 0.259 16.187 16.320 0.455 SD Formal Wages 13.757 14.746 0.224 19.670 19.770 0.502 Mean Self-empl Income 5.452 4.440 0.177 5.886 5.193 0.266 SD Self-empl Income 11.699 9.951 0.293 12.486 11.556 0.468 U Duration (months) 0.066 0.109 0.012 0.182 0.186 0.033 SE Duration (months) 29.523 28.462 1.252 27.747 28.245 1.546 Share Informally Employed - Q1 0.052 0.085 0.006 0.011 0.078 0.007 Share Informally Employed - Q2 0.024 0.043 0.005 0.028 0.036 0.006 Share Informally Employed - Q3 0.034 0.027 0.004 0.018 0.020 0.004 Share Informally Employed - Q4 0.032 0.025 0.003 0.025 0.015 0.003 Share Informally Employed - Q5 0.029 0.017 0.003 0.018 0.015 0.003 Mean Informal Wages - Q1 0.570 0.902 0.079 0.162 0.896 0.099 Mean Informal Wages - Q2 0.338 0.727 0.082 0.413 0.690 0.105 Mean Informal Wages - Q3 0.585 0.571 0.074 0.365 0.531 0.096 Mean Informal Wages - Q4 0.744 0.693 0.094 0.665 0.531 0.114 Mean Informal Wages - Q5 1.045 0.740 0.125 0.819 0.960 0.200 Mean Formal Wages - Q1 1.211 1.153 0.041 1.635 1.363 0.049 Mean Formal Wages - Q2 1.521 1.880 0.068 1.960 2.130 0.090 Mean Formal Wages - Q3 1.920 2.234 0.118 2.442 2.793 0.115 Mean Formal Wages - Q4 2.532 2.837 0.091 3.364 3.695 0.134 Mean Formal Wages - Q5 4.371 4.571 0.112 6.787 6.340 0.234 Aggregate Moment: Model Data Labor Share 0.415 0.419 VIII Figure B.1: Policy Impacts on Job Arrival Rates .25 .3 .35 .4 .45 .5 0.2 .4 .6 .8 Partial Eq. - Low Schooling Partial Eq. - High Schooling General Eq. - Low Schooling General Eq. - High Schooling (a) Changes in t– Unemployed arrival rate .3 .35 .4 .45 .5 0 1 2 3 4 Partial Eq. - Low Schooling Partial Eq. - High Schooling General Eq. - Low Schooling General Eq. - High Schooling (b) Changes in B0– Unemployed arrival rate .025 .03 .035 .04 .045 0.2 .4 .6 .8 Partial Eq. - Low Schooling Partial Eq. - High Schooling General Eq. - Low Schooling General Eq. - High Schooling (c) Changes in t– Self-Employed arrival rate .025 .03 .035 .04 .045 0 1 2 3 4 Partial Eq. - Low Schooling Partial Eq. - High Schooling General Eq. - Low Schooling General Eq. - High Schooling (d) Changes in B0– Self-Employed arrival rate Note: Simulated samples of 10,000 worker-level observations for each schooling group based on the estimates reported in Table 2. The Partial Equilibrium solid lines are computed with exogenous contact rates. The General Equilibrium dashed lines are computed using the endogenous contact rates implied by equilibrium measures of vacancies and searchers. See equation (5) for details. The vertical lines represent the institutional values for the Mexican labor market in 2005. See Appendix C for details. IX CInstitutional Parameters The parameters {B0, τ, t}are set to the values determined by the institutional setting of the Mexican labor market. In particular: τ= 0.55 In order to derive the share of the bundle of additional benefits for Formal employees (τ), we follow calculations reported in Levy [2008], which are based on the current legislation in Mexico. Accordingly, for a worker who earns twice the minimum wage in 2007 (2,931 pesos), social security contributions amount to 864.30 pesos (almost 30% of the wage), of which 55% are attributable to spending categories that are proportional to the wage - notably, work-risk insurance (76.2 pesos), disability and life insurance (69.6 pesos), retirement pensions (184 pesos) and housing fund (146.6 pesos). t= 0.33 We rely on calculations reported in Anton et al. [2012] , which are based on official statistics reported by the Mexican Social Security Institute (IMSS). The authors decompose the average tax rate on formal labor (38%) into government subsidies (5%) and firms’ and workers’ contributions (33%). B0,1= 2.42 and B0,0= 1.92 Total spending in non-contributory social security programs for the year 2005 amounted to 133,090,002,747 pesos, of which 11,916,448,117 pesos were devoted to the Seguro Popular program. For the same year, we compute the total number of informal workers (25,035,508) and unemployed (1,353,561) by applying sampling weights to the nationallyrepresentative labor market survey used in our empirical analysis (ENOE). Assuming full time working hours over a period of one year (2,080 hours), we can compute the per capita hourly monetary benefits extended to the part of the labor force that is non-formally employed, separately for those who reside in municipalities with (B0,1) and without (B0,0) the Seguro Popular program. X