Female labor force participation, labor market dynamic, and growth

Bustelo, Monserrat; Flabbi, Luca; Piras, Claudia; Tejada, Mauricio Working Paper Female labor force participation, labor market dynamic, and growth IDB Working Paper Series, No. IDB-WP-966 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Bustelo, Monserrat; Flabbi, Luca; Piras, Claudia; Tejada, Mauricio (2019) : Female labor force participation, labor market dynamic, and growth, IDB Working Paper Series, No. IDBWP-966, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0001449 This Version is available at: https://hdl.handle.net/10419/208160 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. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode Female Labor Force Participation, Labor Market Dynamic, and Growth Monserrat Bustelo Luca Flabbi Claudia Piras Mauricio Tejada IDB WORKING PAPER SERIES Nº IDB-WP-00966 January 2019 Social Sector (SCL) Inter-American Development Bank January 2019 Female Labor Force Participation, Labor Market Dynamic, and Growth Monserrat Bustelo Luca Flabbi Claudia Piras Mauricio Tejada Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Female labor force participation, labor market dynamic, and growth / Monserrat Bustelo, Luca Flabbi, Claudia Piras, Mauricio Tejada. p. cm. — (IDB Working Paper Series ; 966) Includes bibliographic references. 1. Women-Employment-Latin America. 2. Women in development-Latin America. 3. Labor market-Latin America. 4. Gross domestic product-Latin America. 5. Economic development-Latin America. I. Bustelo, Monserrat. II. Flabbi, Luca. III. Piras, Claudia. IV. Tejada, Mauricio. V. Inter-American Development Bank. Gender and Diversity Division. VI. Series. IDB-WP-966 Keywords: Female labor force participation; Labor market frictions; Search and matching; Nash bargaining; Informality. JEL classification: J24, J3, J64, O17 Copyright © 2019 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 Female Labor Force Participation, Labor Market Dynamic, and Growth Monserrat Bustelo Luca Flabbi Claudia Piras Mauricio Tejada January 2019 Abstract The labor force participation of women is lower than the labor force participation of men. This empirical regularity is particularly acute in Latin America and the Caribbean (LAC). In terms of labor market productivity and growth potential, these lower participation rates constitute a reserve of untapped resources. Providing an estimate of the impact that increased female participation in the labor force has on labor market outcomes and GDP is therefore crucial but challenging. Two issues are of importance: sample selection and equilibrium effects. We develop a labor market model that is able to address these issues. We estimate the model on the microdata of five LAC countries. We find that both a childcare policy and a policy increasing women’s productivity generate a positive impact on female participation and significant increases in GDP per capita. Our results suggest that relatively modest policies that are able to increase the participation of women in the labor market can provide a significant increase in GDP. However, we are not able to take into account the fiscal costs necessary to implement the policies or the possible negative externalities on household production . Keywords: Female labor force participation; Labor market frictions; Search and matching; Nash bargaining; Informality. JEL classification: J24, J3, J64, O17 * We thank Norbert Schady and seminar participants for their useful comments. All errors are our own. The findings, interpreta tions, and conclusions expressed in this paper are those of the authors and do not necessarily represent the views of the Inter-American Development Bank, its executive directors, or the governments they represent . Contact Information: Monserrat Bustelo, Inter-American Development Bank, 1300 New York Avenue, N.W., Washington, DC 20577. E-mail: [email protected]. Luca Flabbi, Department of Economics, University of North Carolina - Chapel Hill, 107 Gardner Hall, CB 3305 Chapel Hill, NC 27599-3305. E-mail: [email protected]. Claudia Piras, Inter-American Development Bank, 1300 New York Avenue, N.W., Washington, DC 20577. Email: [email protected]. Mauricio Tejada, Department of Economics, Universidad Alberto Hurtado, Erasmo Escala 1835 Office 211, Santiago, Chile. E-mail: [email protected]. 1 1 Introduction 1.1 Motivation The labor force participation of women is lower than the labor force participation of men. This empirical regularity is found in virtually all countries 1 and holds true particularly in Latin America and the Caribbean (LAC). For example, Busso and Fonseca (2015) showed that the average rate for female participation in LAC labor forces in 2010 was about 65% compared to about 76% in the United States. There are important differences between LAC countries, with values ranging from 50% in Honduras and Mexico to 70% in Peru and Uruguay. In terms of labor market productivity and growth potential, these lower participation rates constitute a reserve of untapped resources. If these resources could be brought to the market, the production generated by the increased labor force is likely to have substantial positive impacts on the GDP. The potential positive impact of bringing more women to the labor market has increased over time as women continue to acquire more human capital with each passing generation. For example, schooling completed among women is now higher than men in all high-income economies and in many LAC economies. Argentina, Brazil, Colombia, and Uruguay all report a positive gender gap in years of schooling completed; that is, women have on average more years of schooling completed than men. The aggregate average for LAC in 2012 is a small, positive gender gap in favor of women in contrast to women having half a year less than men—a negative gap—in 1992. 2 The objective of this paper is to provide estimates of changes in GDP implied by policies that increase the labor force participation of women on five LAC countries: Argentina, Chile, Colombia, Mexico, and Peru. 1 See, for example, Blau and Kahn (2013), who show a gender difference in employment rates in a large sample of high-income countries, or Olivetti and Petrongolo (2008), who show a gender difference in participation rates in a large sample of OECD countries. On average, participation rates for men are about 90, while participation rates for women are about 75%. 2 See Marchionni (2015) for more details. The aggregate result is strongly driven by younger generations: The 25-34 age group shows a strong positive gap in favor of women; 35-44, a small positive gap; and 45-54, a strong negative gap. All data refer to 2012. 2 1.2 Challenges Estimating the impact of an increase in female labor force participation on labor market outcomes and GDP is challenging. Two issues are of importance when considering such a counterfactual exercise: 1. sample selection and 2. equilibrium effects. Sample selection refers to the difference in the type of individuals who are participating in the labor market with respect to those who are not. When we observe men and women who are currently producing labor in the market, earning wages, and contributing some level of productivity to the country’s economy, we have to consider that a large proportion of women do not work. Therefore, the women who are currently working may differ from those who would enter the labor force as a result of policies designed to increase female labor force participation. For example, if the women who are currently working are more productive than those who are not, we could overestimate the impact of increasing female labor force participation. The opposite would be true if the women currently working are less productive that those who are not. Equilibrium effect refers to the change in equilibrium prices and quantities that may result from a change in the labor market environment. The wage distribution and employment proportion observed in a given moment in an economy are the result of the meeting of labor demand and labor supply in the market. Wages and earnings are the prices realized as a result of this meeting; they may be called equilibrium prices. A significant increase in female labor supply implies a large increase in the amount of labor offered in the market. As a result of an increase in supply, wages and earnings will change. This is the first consequence, labeled here as a short-run equilibrium effect. Eventually, labor demand will also adjust because firms may decide to change their production mix and post more or fewer jobs at various skill levels. This demandside behavior has the potential to change wages and earnings. This is the second consequence, and it is categorized as a long-run equilibrium effect. Both effects make it challenging to quantitatively evaluate the impact of an increase in female labor market participation by only observing wages and earnings before the increase is taking place. This is due to the fact that the 3 observed data are extracted from an equilibrium that is different than the one realized after the increase in participation is taking place. 1.3 Approach A possible approach that is able to take sample selection and equilibrium effects into account consists of specifying an economic model in which the channels generate the effects. Microlevel data for each specific country can then be collected to estimate the parameters of the model. Finally, the estimated model can be used to perform counterfactual experiments in which the quantitative impact of an increase in female labor force participation can be estimated by taking selection and equilibrium effects into account. We propose such an approach by developing and estimating a search model of the labor market. The model captures the specific characteristics of LAC labor markets, including the high level of informality and self-employment. Labor force participation decisions are integrated in the labor market dynamic, taking sample selection into account because the optimal decisions implemented by the agents are sensitive to the policy parameters. Moreover, workers’ decision rules can be explicitly characterized by acknowledging some of the short-run equilibrium effects we described above. Long-run equilibrium effects can be potentially integrated in this setting if firm side data were available. As a first step, we will only use worker side data and limit our analysis to short-run equilibrium effects and some selections effects. Search models of the labor market are widespread and influential 3 because they introduce labor market dynamics, equilibrium unemployment, and noncompetitive features as a tractable and empirically relevant model of the market. Their use in answering policy questions using microdata has a long tradition. For example, Eckstein and Wolpin (1995) studied returns to schooling; Ahn, Arcidiacono, and Wessels (2011) and Flinn (2006) evaluated the employment and welfare impact of minimum-wage legislation; Dey and Flinn (2005) analyzed the impact of employer-provided health insurance; Flabbi (2010) investigated the effects of affirmative-action legislation; and Cahuc, Postel-Vinay, and Robin (2006) evaluated the impact of workers’ bargaining power. Recent contributions have used this approach to answer policy questions in LAC. Tejada (2017) focused on the distortions of introducing 3 For a survey of theoretical literature, see Rogerson, Shimer, and Wright (2005). For a survey of empirical literature, see Eckstein and van den Berg (2007). 4 multiple labor contracts, and Bobba, Flabbi, and Levy (2017) assessed the effects of noncontributory benefits, informality, and long-term impacts on education. Adapting this approach to labor markets in LAC is important to consider in the variety of labor market states present in the region. We model the large informal sector as composed by self-employed and informal employees, but we keep them in separate labor market states to capture the systematic differences in their observed labor market dynamic. Individuals are allowed to move freely between labor market states and may choose to do so as a result of shocks and new opportunities. An additional step is needed to adapt the framework to the study of female labor force participation: a labor supply decision. We introduce an endogenous participation decision as a function of individual heterogeneity over out-of-labor-market market utility, which is allowed to very observable characteristics. This is considered the most important in determining its value: the presence of young children in the household. The endogeneity of the decision will make it sensitive to policy variables, allowing for the evaluation of policy experiments that consider individuals’ optimal responses. Finally, we embed in the model measures able to capture the potential impact on GDP and aggregate welfare. We accomplish this by introducing a match-specific productivity distribution that is affected by policy variables and by optimal individual behavior. This approach dates to at least Eckstein and Wolpin (1995). In gender literature, Flabbi (2010) used this to evaluate affirmative-action policies in favor of women. In LAC, Tejada and Perticara (2016) employed this method to estimate the presence of discrimination against women. The proposed approach has two main advantages. First, we are able to deal with the two main challenges described above: sample selection and equilibrium effects. Sample selection is explicitly modeled because the participation decision is endogenous. Estimates of the out-of-labor-market utility’s heterogeneity will allow for a quantitative assessment of the importance of this channel. Equilibrium effects are taken into account through two features: the optimal reservation values rules and the endogenous accepted-wage distribution. 11 3.2 Value Functions The full formal representation of the model is presented in Appendix A. Here, we just briefly mention that the stationarity of the environment allows for a recursive characterization of the dynamic. For example, we can write the discounted value of an unemployed worker of type i as follows: 𝜌𝑈 𝑖 = 𝑏 𝑖 + 𝜆 𝑖𝐹 ∫𝑚𝑎𝑥[𝐸 𝑖𝐹 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝐹 (𝑥)+𝜆 𝑖𝐼 ∫𝑚𝑎𝑥[𝐸 𝑖𝐼 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝐼 (𝑥) +𝜆 𝑖𝑆 ∫𝑚𝑎𝑥[𝐸 𝑖𝑆 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝑆 (𝑥)−(𝜆 𝑖𝐹 +𝜆 𝑖𝐼 +𝜆 𝑖𝑆 ) 𝑈 𝑖 (1) The interpretation is intuitive. When a worker is unemployed, she receives utility b i every period. Moreover, she has the possibility of meeting an employer offering a formal or an informal job (with probability λ iF and λ iI , respectively). Finally, the unemployed worker can take advantage of a self-employment opportunity with probability λ iS . Every time she receives a job opportunity, either as an employee or as self-employed, she has the possibility to reject or accept the offer, as represented by the max operator over the possible labor market states. The trade-off involved in the decision is as follows. If the worker accepts the offer, she receives labor income, but if she rejects, this person may receive an even better offer in the future. All future offers are realized only when meeting a specific employer or self-employment opportunity. Therefore, the unemployed agent can only have an expectation of what those offers will be: The integral operator over the appropriate distributions define these expectations. When a worker meets an employer, they both realize the potential productivity of that specific worker at that firm. We denote it by x. Based on this, they split the revenue the usual way: The worker receives wages, and the firm keeps the profit, which will be equal to the revenue x less than the wage paid to the worker. In addition, firms hiring legally have to pay the social security contribution τ, while firms hiring illegally set aside the illegality cost c. The actual wage paid to the workers is decided by bargaining; that is, the employee and firm engage in making offers and counteroffers while contemplating their outside 12 options. Their outside options are the state they will be in if they reject the offer. For the worker, this will be unemployment, and for the firm, this is the state of having a vacancy open at the firm. Additionally, workers and firms may have a stronger or weaker bargaining power, which includes other factors that may put the agents in a stronger bargaining position. We denote this parameter with β. 3.3 Wage Determination When a worker meets an employer, they both realize the potential productivity of that specific worker at that firm. We denote it by x. Based on this, they split the revenue the usual way: The worker receives wages, and the firm keeps the profit, which will be equal to the revenue x less than the wage paid to the worker. In addition, firms hiring legally have to pay the social security contribution τ, while firms hiring illegally set aside the illegality cost c. The actual wage paid to the workers is decided by bargaining; that is, the employee and firm engage in making offers and counteroffers while contemplating their outside options. Their outside options are the state they will be in if they reject the offer. For the worker, this will be unemployment, and for the firm, this is the state of having a vacancy open at the firm. Additionally, workers and firms may have a stronger or weaker bargaining power, which includes other factors that may put the agents in a stronger bargaining position. We denote this parameter with β. The details for the solution of this bargaining problem are given in Appendix A. Here, we only mention that we assume the axiomatic Nash-bargaining solution, which leads to the following reasonably intuitive wage schedules: 𝑤 𝑖𝐹 (𝑥)=𝛽 𝑥 (1+𝜏)+(1−𝛽)𝜌𝑈 𝑖 (2) 𝑤 𝑖𝐼 (𝑥)=𝛽(𝑥−𝑐)+(1−𝛽)𝜌𝑈 𝑖 (3) Wages increase with the worker’s productivity x. However, the productivity is decreased either by the contribution rate τ or by the illegality cost c. Moreover, the higher the worker’s outside option (ρU i ), the higher the wage. Finally, the higher the worker’s 13 bargaining power β, the higher the portion of the productivity x the worker will receive through the wage. In conclusion, when a woman meets an employer offering a formal job generating productivity x, she will receive a wage w WF (x). When the offer is for an informal job, she will receive a wage w WI (x). When the job offer is self-employment, she will receive the entire productivity x. The same is true for men, but their wage schedules may potentially have different parameters and therefore other outside options. 3.4 Equilibrium The equilibrium of the model has a simple structure. Agents have to make two discrete choices. The first concerns labor market participation: Either they participate in the labor market looking for a job (state U) or they stay out enjoying the utility of out-of-labor-market activities (state NP). Because women receive different utilities from these activities (z), women receiving a relatively high utility will stay out, while women receiving a relatively low utility will enter the market. The threshold for staying out or coming in is determined by the indifference point between the two states (i.e., by the specific 𝑧 𝑖∗ ) such that: 𝑁𝑃 𝑖 (𝑧 𝑖∗ )= 𝑈 𝑖 ⇒𝑧 𝑖∗ = 𝜌𝑈 𝑖 In conclusion, all the women with 𝑧 𝑊 < 𝑧 𝑊 ∗ participate in the labor market; all those with 𝑧 𝑊 > 𝑧 𝑊 ∗ stay out. The same is true for men but at different parameters. The second discrete choice the agents have to make concerns the labor market state decision: Either they accept a job offer or they reject it and continue searching. Again, we can identify a threshold. If the productivity and therefore the wage is high enough, they will accept. If not, they will continue to search for a better offer. As before, the threshold is identified by the indifference point between the two alternatives (i.e., by the specific 𝑥 𝑖𝑗 ∗ ) such that: 𝑈 𝑖 = 𝐸 𝑖𝐹 (𝑥 𝑖𝐹 ∗ )⇒𝑥 𝑖𝐹 ∗ =(1+𝜏)𝜌𝑈 𝑖 (4) 𝑈 𝑖 = 𝐸 𝑖𝐼 (𝑥 𝑖𝐼 ∗ )⇒𝑥 𝑖𝐼∗ =𝜌𝑈 𝑖 + 𝑐 (5) 𝑈 𝑖 = 𝐸 𝑖𝑆 (𝑥 𝑖𝑆 ∗ )⇒𝑥 𝑖𝑆 ∗ =𝜌𝑈 𝑖 (6) 14 Notice that these thresholds have some economic interpretations. Employee relationships require higher productivity to be acceptable because the worker has to share with the employer. Moreover, the employer has to pay contribution or illegality costs, so the thresholds are increasing in those parameters. In conclusion, every time an unemployed woman receives a wage offer that is higher than 𝑤 𝑤𝑗 (𝑥 𝑤𝑗 ∗ ) or a self-employment opportunity with income higher than 𝑥 𝑤𝑆 ∗ , she will accept. Otherwise, she will keep searching. Men, on the other hand, tend to have analogous behavior on different parameters. These relatively simple, threshold-based, optimal decision rules can be incorporated in the value functions that then can be solved as a function of the primitive parameters. Finally, the optimal decision rules, solved value function, and steady state conditions can be used to determine the equilibrium levels of nonparticipation ( 𝑛 𝑝𝑖 ) , unemployment ( 𝑢 𝑖 ), and employment ( 𝑒 𝑖,𝑗 ) for each gender. Again, the details are in Appendix A. 4 Identification and Estimation Method We discuss identification and estimation based on the model we developed and the data at our disposal. As described in Section 2, we have information on labor market states, hourly wages or earnings (w), and ongoing unemployment duration (u). Each piece of information is available for each gender and for each of the three education groups we consider: primary, secondary, and tertiary schooling level. The combination of our model and our data allows for the derivation of the likelihood contributions of each observation in our sample (see Appendix B). From the likelihood contributions, it is possible to formally prove the identification of the structural parameters of the model under some common distributional assumptions about the matchspecific productivity x and the out-of-labor-market utility z (Flinn & Heckman, 1982). The only parameters we have to normalize are the discount rate ρ, which we fix at 5% a year, and the Nash-bargaining parameter β, which we fix at the symmetric bargaining value of 0.5. Although the theoretical identification of β is assured by the model’s implications and by the distributional assumptions, its empirical identification is challenging without 15 demand-side information, 9 and that is why we simply calibrate the parameter to the value of symmetric Nash bargaining. This is a restriction in our context because it forces us to set the same Nash-bargaining parameter for men and women. Previous literature has shown that differences in β by gender are likely to be present and often interpreted as capturing discrimination or gender-specific attitudes toward negotiation. 10 Even if we have to impose the restriction, it is worth remembering that the presence of endogenous and gender-specific outside options (U i ) still allows the wages to capture differences in bargaining power between men and women. Because the outside option enters directly in the wage equations, a lower outside option for a given gender in a schooling group translates into lower wages at the same productivity compared to the other gender. 11 Following previous literature, we assume that the match-specific productivity distribution G ij (x) is lognormal with parameters (µ ij , σ ij ). Each set of parameters is allowed to be different by country and education group on top of gender i and type of employment j. Additionally, we assume that the out-of-labor-market utility distribution Q i (z) is a negative exponent with parameter γ iκ . The subscript iκ denotes that the parameter is not only a function of gender i but also of the presence of young children in a household. We consider three age groups: households with at least one child aged 5 or younger (κ = k5); households with at least one child older than 5 but 13 or younger (κ = k13); and households where no children are aged 13 or younger (κ = other). After a preliminary analysis, we concluded that the estimates on men were not sensitive to the presence of children; therefore, we introduce these differences only on the women’s specifications. As with the productivity distributions, each set of parameters is allowed to be different by country and education group. Finally, we allow for the presence of measurement errors in wages. We assume classic measurement error: Observed wages w o are equal to the true wage w up to a multiplicative measurement error 𝜖 . We assume the log of 𝜀 is normal with mean zero and variance 𝜎 𝑀𝐸 2 . 9 For a formal discussion, see Flinn (2006). For an example on implementation using demand-side information, see Cahuc et al. (2006). 10 See, for example, Bartolucci (2013). Eckstein and Wolpin (1999) and Borowczyk-Martins, Bradley, and Tarasonis (2017) are examples of a similar strategy applied to racial gaps instead of gender gaps. 11 See equations 2 and 3. 16 5 Estimation Results The complete parameter estimates are reported in Appendix C. The estimates are quite precise, typically more so the higher the education level and the larger the sample size. The estimates also report significant differences for many parameters by gender, country, and education. Three comments about those differences are worth mentioning. First, Colombia has the lowest arrival rates in the formal/informal sector, and the differences with respect to the other countries are statistically significant. Additionally, in all countries, the biggest (and statistically significant) differences between arrival rates in the formal/informal sector of men and women are in the group of workers with primary education. Second, in all countries, except Peru, the termination rate of formal/informal jobs is lower in the group of workers with tertiary education. The differences with other educational groups are evident and statistically significant for Argentina and Chile. With respect to gender, termination rates of formal/informal jobs are in general higher for women, but the differences are statistically significant for all educational groups for Argentina. Finally, productivity is lower for women in formal jobs. The differences, however, are statistically significant for Argentina (in primary and secondary educational groups), Chile, and Mexico. Only in the cases of Argentina and Peru, in the tertiary educational group, are women more productive than men, but the difference is not statistically significant. In informal jobs, in turn, women are more productive than men only in the case of Argentina and for all educational groups. However, the difference is statistically significant only for the secondary educational group. In Chile (primary education) and Colombia (secondary education), women are also more productive than men in informal jobs and the differences in this case are statistically significant. Among the structural parameters, the parameter γ iκ is of interest because it is the parameter governing the distribution of the utility when not participating in the labor market. As expected, the presence of young children in the household increases the value of out-of-labor-market activities. The difference may be substantial. For example, among tertiary educated women in Colombia, the average value of out-of-labor-market activities when children younger than 5 are present is almost 30% higher than when no children younger than 13 are present. 17 Tables 6 through 10 report the implications of the parameter estimate on productivity and wages. The top panel of each table reports expected value ( E [ x ]) and standard deviation ( SD [ x ]) of the match-specific productivity in formal employment, informal employment, and self-employment. They describe the primitive productivity distributions that we denoted with G ij ( x ) in the formal modal, and they represent the potential output of a given match between a worker and a firm. Some of these matches are realized (accepted) and some are not, depending on the optimal decision rules of the agents (see Section 3.4). The bottom panel of each table reports the expected value and standard deviation of the accepted wages in formal employment and informal employment and of the realized labor income in self-employment. Notice that the relation between the top panel and the bottom panel involves two steps. The first step is the mapping between a specific value of productivity x and the wage paid to the worker w . This relation is governed by the equilibrium equations 2 and 3. The second step is the optimal decision rule: Not all the matches are acceptable. Only matches with productivity higher than the appropriate reservation values—as defined in equations 4 and 5—are realized in equilibrium. In the case of the self-employed, the mapping between productivity and realized labor income only involves the second step. Finally, the middle panel of each table reports the implied GDP per worker ( GDP W ) and GDP per capita ( GDP C ) for each schooling and education group. It is a useful measure to evaluate the policy experiments, and it represents the total value of the production of a given group in the economy. It does take into account that (a) agents may spend time in different labor market states, including unemployment; (b) agents may be less or more productive if they work formally or informally; and (c) some agents may not participate in the labor market at all. The formal definition of the measures GDP W and GDP C as a function of the model parameters is given in Appendix A. The first relevant result reported in the top panel was expected: Productivity increases with education in all countries and for men and women. For example, the average productivity of formal male employees in Peru is about 6% higher if they complete secondary school with respect to primary and about 45% higher if they complete tertiary school with respect to secondary. The second result is less obvious: The average gender gap in productivity is sometimes very different from the average gender gap in wages. If the gender gap in wages typically favors men, then that is not always true of the gap in productivity. For example, in 18 Peru, the average productivity of women with tertiary education working as formal employees is about 10% higher than the average productivity of the corresponding group of men. The gap increases to almost 30% when considering informal employees and decreases to about 3% among the self-employed. 12 It is important to notice that a gender gap in productivity in favor of women rarely translates into a similar gap in accepted wages. Again, looking at tertiary educated women in Peru, the last column of the bottom panels in Tables 6 through 10 show almost identical accepted wages between men and women working as employees and actually a significantly lower average self-employment income for women with respect to men. Even if women may have on average higher productivity, they may decide to accept lower wages as a result of different arrival rates of offers, different values of the outside option while bargaining, and different values of out-of-labor-market activities. The bottom panels are useful to assess gender gaps in accepted wages but also to judge how well the estimated model fits the data. That is why each table reports not only the simulated moments (denoted by Model) but also the sample moments (denoted by Data). The fit of the model is quite good on the means, but in some instances, it is unable to fit the standard deviations. Goodness of fit on the other labor market variables— including participation rates and labor market dynamic over the other labor markets states—are reported in Appendix C. 6 Policy Experiments 6.1 Definition We propose two policy experiments that may clarify the reasons behind and the loss implied by the lower labor market participation of women with respect to men. Women may decide to participate less than men either because the value of nonparticipation is higher or because the benefit of participating in the market is lower. The first experiment relates to the first component—the value of nonparticipation—and the 12 The gender gaps are reported in the third column of each gender-education group. 19 second experiment to the second component—gender asymmetries in labor market opportunities. Opinion surveys and economic literature indicate that women value time outside the labor market more than men. 13 Our estimates show this to be the case because the average value of nonparticipation E(z) is estimated to be higher for women than men in all education groups. Many factors may impact this difference, such as preferences, household production, abilities, and attitudes. One major component seems to be childcare and child-rearing. Women still invest a higher amount of hours in childcare than men and their labor market participation is significantly affected by fertility outcomes (Burda, Hamermesh, & Weil, 2013). Many policy tools may have an impact on this value. For example, good and affordable childcare provisions may decrease the benefit of mothers’ time in child-rearing and induce them to work more. Numerous policies focus on providing good and affordable childcare, using either a voucher system that provides subsidies to parents who use childcare or a direct public provision of the service. 14 To map this policy in our model, we change the parameters governing the flow utility of nonparticipation z. Recall that this value is heterogeneous in the population, but it is distributed with the cdf Q(z). We estimate specific Q(z) for women with young children. Specifically, we allow the distribution of values of nonparticipation to be different between women with children 5 or younger, children between the age of 5 and 13, and without children younger than 13. Because childcare provision policies are more likely to affect mothers with young children, Policy Experiment 1 reduces the average value of nonparticipation for those mothers in half. Formally, it is equivalent to doubling the parameter γ k5 . Reducing the value in half is arbitrary but, as we will show when discussing the results, seems to generate labor supply responses in line with some estimates available in the literature. Still, the reduction in half is more a reference point than an attempt to mimic specific policies implemented in the region. To gain more flexibility in this respect and to study the 13 For example, Scandura and Lankau (1997) show that women more than men value flexible working arrangements in order to perform activities not related with the labor market. 14 Examples of specific policies in the region include construction of preprimary school facilities in Argentina (Berlinski & Galiani, 2007); subsidized provision of after-school care in Chile (Martınez & Perticara, 2017); and a large subsidized childcare program in Colombia (Bernal & Fernandez, 2013). 20 possible nonlinearity of the policy impacts, we also present selected results on the same policy where we vary the average value of nonparticipation for mothers with children 5 or younger over a broader range: from a 25% to a 75% decrease. Gender asymmetries in labor market opportunities are the results of many components, including the gender wage gap, differences in promotions and labor market careers, asymmetries in search intensity, and occupational choices. Some of these differences may be due to differences in preferences and attitudes, but others may relate to issues affected by policies such as human capital accumulation, gender discrimination, and occupational choices. For example, a policy that gives incentives to women to enroll in STEM or an affirmative-action policy aiming at reducing discrimination can both be seen as policies boosting women productivities. In this spirit, Policy Experiment 2 increases the average productivity of women in the three sectors by 10%. Because productivity is represented in our model by the distributions G i,j (x), formally, the experiments change the parameters µ Wj and σ Wj for j = F, I, S so that the new average productivity E Wj (x) is 10% higher. We chose 10% to ease the calculation of the elasticities, but it is worth noting that, in many cases, a 10% increase is enough to close the gender gap in productivity. This is true in most countries among workers who completed secondary and tertiary education. 15 Among workers with only primary education, the gaps are instead typically larger, ranging from 20% to 30%, and therefore a 10% increase is not enough to generate the same average productivity between men and women. As in the previous experiment, 10% is a useful but arbitrary reference point. To study the impact on a broader range of values, we also implement experiments changing average productivity over a grid of values ranging from 1% to 20%. 15 A notable exception is Chile, which is registering the largest gender gap in productivity in the tertiary education group. We estimate the average productivity of women to be about 20% lower than the average productivity of men. See the last column of Table 7. 27 References Ahn, T., Arcidiacono, P., & Wessels, W. (2011). The distributional impacts of minimum wage increase when both labor supply and labor demand are endogenous. Journal of Business & Economic Statistics, 29(1), 12–23. Anton, A., Hernandez, F., & Levy, S. (2012). The End of Informality in Mexico? The InterAmerican Development Bank, Wahington DC. Baker, M., Gruber, J., & Milligan, K. (2008). Universal child care, maternal labor supply, and family well-being. Journal of Political Economy, 116(4), 709–745. Bartolucci, C. (2013). Gender wage gaps reconsidered: A structural approach using matched employer-employee data. Journal of Human Resources, 48(4), 998–1034. Berlinski, S., & Galiani, S. (2007). The effect of a large expansion of pre-primary school facilities on preschool attendance and maternal employment. Labour Economics, 14(3), 665–680. Retrieved from http://www.sciencedirect.com/science/article/pii/S092753710700005X Bernal, R., & F ern ´ andez, C. (2013). Subsidized c hildcare and c hild dev elopmen t in Colom bia: Effects of hogares comunitarios de bienestar as a function of timing and length of exposure. Social Science & Medicine, 97, 241–249. Blau, D., & Currie, J. (2006). Pre-school, day care, and after-school care: Who’s minding the kids? Handbook of the Economics of Education, 2, 1163–1278. Blau, F. D., & Kahn, L. M. (2013). Female labor supply: Why is the US falling behind? (Technical report). National Bureau of Economic Research. Bloemen, H. G. (2008). Job search, hours restrictions, and desired hours of work. Journal of Labor Economics, 26, 137–179. Bobba, M., Flabbi, L., & Levy, S. (2017). Labor market search, informality and schooling investments (Discussion Paper 11170). Institute for the Study of Labor (IZA). Borowczyk-Martins, D., Bradley, J., & Tarasonis, L. (2017). Racial discrimination in the US labor market: Employment and wage differentials by skill. Labour Economics, 49, 106–127. Burda, M., Hamermesh, D. S., & Weil, P. (2013). Total work and gender: Facts and possible explanations. Journal of Population Economics, 26(1), 239–261. Busso, M., & Fonseca, D. R. (2015). Facts and determinants of female labor supply in Latin America. In L. Gasparini and M. Marchionni (Eds.), Bridging gender gaps? Argemtina: CEDLAS. 28 Cahuc, P., Postel-Vinay, F., & Robin, J. M. (2006). Wage bargaining with on-the-job search: Theory and evidence. Econometrica, 74(2), 323–364. Cuberes, D., & Teignier, M. (2016). Aggregate effects of gender gaps in the labor market: A quantitative estimate. Journal of Human Capital, 10(1), 1–32. Dey, M. S., & Flinn, C. J. (2005). An equilibrium model of health insurance provision and wage determination. Econometrica, 73(2), 571–627. Eckstein, Z., & van den Berg, G. J. (2007). Empirical labor search: A survey. Journal of Econometrics, 136(2), 531–564. Eckstein, Z., & Wolpin, K. I. (1995). Duration to first job and the return to schooling: Estimates from a search-matching model. Review of Economic Studies, 62(2), 263– 86. Eckstein, Z., & Wolpin, K. I. (1999). Estimating the effect of racial discrimination on first-job wage offers. The Review of Economics and Statistics, 81(3), 384–392. Felfe, C., Lechner, M., & Thiemann, P. (2016). After-school care and parents’ labor supply. Labour Economics, 42, 64–75. Flabbi, L. (2010). Gender discrimination estimation in a search model with matching and bargaining. International Economic Review, 51(3), 745–783. Flabbi, L., & Mabli, J. (2018). Household search or individual search: Does it matter? Journal of Labor Economics, 36(1), 1–46. Flinn, C. (2006). Minimum wage effects on labor market outcomes under search, bargaining and endogenous contact rates. Econometrica, 73, 1013–1062. Flinn, C., & Heckman, J. (1982). New methods for analyzing structural models of labor force dynamics. Journal of Econometrics, 18(1), 115–168. Kanbur, R. (2009). Conceptualizing informality: Regulation and enforcement. Indian Journal of Labour Economics, 52(1), 33–42. Levy, S. (2008). Good intentions, bad outcomes: Social policy, informality, and economic growth in Mexico. Washington, DC: Brooking Institution Press. Marchionni, M. (2015). A changing scenario: Education, family, and economic environment. In L. Gasparini and M. Marchionni (Eds.), Bridging gender gaps? Argentina: CEDLAS. Mar t ´ ınez, C., & P erticar ´ a, M. (2017). Childcare effects on maternal emplo ymen t: Evidence from Chile. Journal of Development Economics, 126, 127–137. 29 Meghir, C., Narita, R., & Robin, J. M. (2015). Wages and informality in developing countries. American Economic Review, 105(4), 1509–46. Olivetti, C., & Petrongolo, B. (2008). Unequal pay or unequal employment? A crosscountry analysis of gender gaps. Journal of Labor Economics, 26(4), 621–654. Petrongolo, B., & Pissarides, C. A. (2001). Looking into the black box: A survey of the matching function. Journal of Economic Literature, 39(2), 390–431. Rogerson, R., Shimer, R., & Wright, R. (2005). Search-theoretic models of the labor market: A survey. Journal of Economic Literature, 43(4), 959–988. Scandura, T. A., & Lankau, M. J. (1997). Relationships of gender, family responsibility and flexible work hours to organizational commitment and job satisfaction. Journal of Organizational Behavior, 377–391. Strategy and Co. (2012). Empowering the third billion: Women and the world of work in 2012. Tejada, M. M. (2017). Dual labor markets and labor protection in an estimated search and matching model. Labour Economics, 46, 26–46. Tejada, M., & Perticara, M. (2016). Sources of gender wage gaps for skilled workers in Latin American countries (Working Paper 317). ILADES, Universidad Alberto Hurtado. 30 Table 1: Argentina - Descriptive Statistics Labor Market N Prop. t ¯ u w ¯ σ w N Prop. t ¯ u w ¯ σ w States Men Women Education Group: Primary K ≤ 5 1750 0.44 5 < K ≤ 13 1091 0.28 Education Group: Secondary Unemployed 190 0.04 - - - 219 0.05 - - - Formal Emp. 2460 0.54 - 5.10 2.36 1426 0.30 - 4.66 2.19 Informal Emp. 665 0.14 - 2.84 1.65 712 0.15 - 2.78 1.78 Self-Emp. 1043 0.23 - 3.52 2.77 565 0.12 - 3.16 3.21 Non Part. 229 0.05 - - - 1837 0.39 - - - K ≤ 5 772 0.42 5 < K ≤ 13 485 0.26 Education Group: Tertiary Unemployed 140 0.03 - - - 252 0.04 - - - Formal Emp. 2555 0.59 - 6.73 3.35 3455 0.53 - 6.64 3.03 Informal Emp. 374 0.09 - 4.17 2.96 640 0.10 - 3.89 2.77 Self-Emp. 914 0.21 - 5.21 4.36 812 0.12 - 5.23 4.77 Non Part. 335 0.08 - - - 1344 0.21 - - - K ≤ 5 506 0.38 5 < K ≤ 13 292 0.22 Note. Wage distributions are trimmed at the top and bottom 1 percentile by gender, education group, and type of job, and they are reported in U.S. dollars as of December 2016 (exchange rate = 15.8620 Argentinian pesos/U.S. dollar). A worker is categorized as informal if he or she reports not having benefits of social security. K means proportion of women with the presence of children in the household with respect to non participating w omen. Unemploymen t durations ( t ¯ u ) are only observed in time intervals. Unemployed 400 0.05 - - - 311 0.04 - - - Formal Emp. 2594 0.34 - 4.49 2.14 1070 0.14 - 3.78 1.75 Informal Emp. 1773 0.24 - 2.48 1.33 1584 0.21 - 2.60 1.56 Self-Emp. 2030 0.27 - 3.00 2.27 726 0.10 - 2.37 2.18 Non Part. 737 0.10 - - - 3946 0.52 - - - 31 Table 2: Chile - Descriptive Statistics Labor Market N Prop. t ¯ u w ¯ σw N Prop. t ¯ u w ¯ σw States Men Women Education Group: Primary Unemployed 873 0.07 2.55 - - 776 0.05 2.09 - - Formal Emp. 5807 0.46 - 2.68 1.11 2703 0.17 - 2.13 0.68 Informal Emp. 865 0.07 - 2.31 1.12 403 0.03 - 2.00 1.38 Self-Emp. 3073 0.25 - 2.63 2.02 1871 0.12 - 2.33 2.29 Non Part. 1882 0.15 - - - 10176 0.64 - - - K ≤ 5 3201 0.31 5 < K ≤ 13 2710 0.27 Education Group: Secondary 5 < K ≤ 13 K ≤ 5 1314 0.39 5 < K ≤ 13 769 0.23 Note. Wage distributions are trimmed at the top and bottom 1 percentile by gender, education group, and type of job, and they are reported in U.S. dollars as of December 2016 (exchange rate = 667.17 Chilean pesos/U.S. dollar). A worker is categorized as informal if he or she reports not having benefits of social security. K means proportion of women with the presence of children in the household with respect to nonparticipating women. Unemployed 1002 0.07 2.89 - - 980 0.05 2.67 - - Formal Emp. 9995 0.65 - 3.26 1.58 7052 0.39 - 2.57 1.04 Informal Emp. 715 0.05 - 2.80 1.71 531 0.03 - 2.37 1.56 Self-Emp. 2717 0.18 - 3.46 3.11 2203 0.12 - 2.84 2.76 Non Part. 892 0.06 - - - 7504 0.41 - - - K ≤ 5 Education Group: 3067 2071 Tertiary 0.41 0.28 Unemployed 778 0.06 3.35 - - 802 0.05 2.93 - - Formal Emp. 8510 0.66 - 7.31 5.92 9246 0.60 - 5.50 3.73 Informal Emp. 446 0.03 - 5.73 5.46 497 0.03 - 4.98 3.79 Self-Emp. 1966 0.15 - 8.09 9.04 1442 0.09 - 6.20 6.67 Non Part. 1278 0.10 - - - 3401 0.22 - - - 32 Table 3: Colombia - Descriptive Statistics Labor Market N Prop. t ¯ u w ¯ σw N Prop. t ¯ u w ¯ σw States Men Women Education Group: Primary Unemployed 607 0.06 3.14 - - 828 0.07 4.56 - - Formal Emp. 1784 0.18 - 1.31 0.41 669 0.06 - 1.17 0.23 Informal Emp. 1311 0.13 - 1.08 0.39 935 0.08 - 0.87 0.36 Self-Emp. 5487 0.55 - 1.12 0.66 4199 0.35 - 0.80 0.57 Non Part. 758 0.08 - - - 5429 0.45 - - - K ≤ 5 1870 0.34 5 < K ≤ 13 1552 0.29 Education Group: Secondary Unemployed 577 0.06 4.05 - - 984 0.09 5.22 - - Formal Emp. 3656 0.41 - 1.45 0.54 2246 0.21 - 1.31 0.38 Informal Emp. 819 0.09 - 1.13 0.41 932 0.09 - 0.98 0.35 Self-Emp. 3496 0.39 - 1.40 0.91 3084 0.29 - 1.07 0.84 Non Part. 408 0.05 - - - 3335 0.32 - - - K ≤ 5 1272 0.38 5 < K ≤ 13 970 0.29 Education Group: Tertiary Unemployed 840 0.09 5.33 - - 1611 0.12 6.02 - - Formal Emp. 4551 0.50 - 3.06 2.24 5885 0.44 - 2.77 1.94 Informal Emp. 422 0.05 - 1.41 0.79 562 0.04 - 1.28 0.68 Self-Emp. 2775 0.30 - 2.99 2.73 3027 0.23 - 2.60 2.34 Non Part. 583 0.06 - - - 2167 0.16 - - - K ≤ 5 893 0.41 5 < K ≤ 13 516 0.24 Note. Wage distributions are trimmed at the top and bottom 1 percentile by gender, education group, and type of job, and they are reported in U.S. dollars as of December 2016 (exchange rate = 3009.86 Colombian pesos/U.S. dollar). A worker is categorized as informal if he or she reports not having benefits of social security. K means proportion of women with the presence of kids in the household with respect to nonparticipating women. 33 Table 4: Mexico - Descriptive Statistics Labor Market N Prop. t ¯ u w ¯ σw N Prop. t ¯ u w ¯ σw States Men Women Education Group: Primary Unemployed 328 0.03 1.24 - - 182 0.01 1.50 - - Formal Emp. 2412 0.24 - 1.42 0.59 1063 0.07 - 1.14 0.44 Informal Emp. 3480 0.35 - 1.22 0.52 1177 0.08 - 1.04 0.63 Self-Emp. 2415 0.24 - 1.67 1.14 2248 0.15 - 1.18 1.04 Non Part. 1413 0.14 - - - 10430 0.69 - - - K ≤ 5 3727 0.36 5 < K ≤ 13 2902 0.28 Education Group: Secondary Unemployed 1076 0.04 1.95 - - 713 0.02 1.87 - - Formal Emp. 11929 0.46 - 1.59 0.75 6235 0.19 - 1.39 0.69 Informal Emp. 6401 0.25 - 1.29 0.66 2991 0.09 - 1.15 0.67 Self-Emp. 4770 0.18 - 1.99 1.58 4001 0.12 - 1.67 1.63 Non Part. 1832 0.07 - - - 18215 0.57 - - - K ≤ 5 7809 0.43 5 < K ≤ 13 5532 0.30 Education Group: Tertiary Unemployed 782 0.06 2.73 - - 647 0.04 2.61 - - Formal Emp. 7078 0.57 - 3.02 1.85 7227 0.42 - 2.86 1.63 Informal Emp. 1389 0.11 - 2.09 1.57 1380 0.08 - 2.02 1.48 Self-Emp. 1897 0.15 - 3.17 2.90 1474 0.09 - 2.64 2.62 Non Part. 1239 0.10 - - - 6358 0.37 - - - K ≤ 5 2115 0.33 5 < K ≤ 13 1545 0.24 Note. Wage distributions are trimmed at the top and bottom 1 percentile by gender, education group, and type of job, and they are reported in U.S. dollars as of December 2016 (exchange rate = 20.52 Mexican pesos/U.S. dollar). A worker is categorized as informal if he or she reports not having access to health care. K means proportion of women with the presence of children in the household with respect to nonparticipating women. 34 Table 5: Peru - Descriptive Statistics Labor Market N Prop. t ¯ u w ¯ σw N Prop. t ¯ u w ¯ σw States Men Women Education Group: Primary Unemployed 60 0.02 1.04 - - 102 0.02 0.74 - - Formal Emp. 631 0.18 - 1.95 0.99 192 0.03 - 1.37 0.54 Informal Emp. 981 0.29 - 1.52 0.77 581 0.09 - 1.01 0.62 Self-Emp. 1447 0.42 - 1.86 1.68 3198 0.52 - 1.06 1.20 Non Part. 319 0.09 - - - 2059 0.34 - - - K ≤ 5 1014 0.49 5 < K ≤ 13 533 0.26 Education Group: Secondary Unemployed 121 0.02 1.14 - - 94 0.02 0.72 - - Formal Emp. 1659 0.33 - 2.28 1.23 485 0.11 - 1.79 1.04 Informal Emp. 1023 0.20 - 1.62 0.84 641 0.15 - 1.21 0.71 Self-Emp. 1966 0.39 - 2.21 2.10 1670 0.39 - 1.50 1.72 Non Part. 270 0.05 - - - 1429 0.33 - - - K ≤ 5 716 0.50 5 < K ≤ 13 384 0.27 Education Group: Tertiary Unemployed 236 0.04 1.31 - - 259 0.04 1.12 - - Formal Emp. 3685 0.57 - 3.82 2.60 3061 0.44 - 3.54 2.16 Informal Emp. 627 0.10 - 2.18 1.57 730 0.11 - 1.77 1.32 Self-Emp. 1588 0.24 - 3.48 3.88 1380 0.20 - 2.27 2.96 Non Part. 383 0.06 - - - 1468 0.21 - - - K ≤ 5 717 0.49 5 < K ≤ 13 361 0.25 Note. Wage distributions are trimmed at the top and bottom 1 percentile by gender, education group, and type of job, and they are reported in U.S. dollars as of December 2016 (exchange rate = 3.395 soles/U.S. dollar). A worker is categorized as informal if he or she reports not having access to health care. K means proportion of women with the presence of children in the household with respect to nonparticipating women. 35 Table 6: Argentina - Productivity and Wages Primary Secondary Tertiary M W W/M M W W/M M W W/M E[xF ] Model 4.493 3.788 0.843 5.134 4.731 0.922 6.753 6.689 0.990 SD ( x F ) Model 0.024 0.021 0.849 0.010 0.018 1.783 0.009 0.005 0.631 E[xI] Model 2.494 2.638 1.058 2.853 2.796 0.980 4.677 4.259 0.911 SD[xI] Model 0.680 1.070 1.574 1.524 1.761 1.156 8.133 5.026 0.618 E[xS] Model 3.013 2.413 0.801 3.525 3.197 0.907 5.263 5.428 1.031 SD[xS] Model 1.758 1.880 1.069 2.204 2.781 1.261 3.931 4.786 1.217 GDP W Model 7.189 7.035 0.979 9.025 8.151 0.903 13.462 14.111 1.048 GDP C Model 6.107 3.113 0.510 8.201 4.627 0.564 11.980 10.647 0.889 E [ w | eF ] Data 4.492 3.783 0.842 5.095 4.662 0.915 6.728 6.642 0.987 Model 4.524 3.769 0.833 5.161 4.760 0.922 6.749 6.700 0.993 SD [ w | eF ] Data 2.140 1.749 0.817 2.361 2.189 0.927 3.354 3.035 0.905 Model 2.169 1.773 0.818 2.541 2.448 0.964 3.443 3.230 0.938 E[w|eI] Data 2.477 2.597 1.048 2.845 2.783 0.978 4.167 3.892 0.934 Model 2.499 2.640 1.057 2.841 2.810 0.989 4.565 4.287 0.939 SD[w|eI] Data 1.329 1.559 1.173 1.645 1.782 1.083 2.957 2.774 0.938 Model 1.402 1.695 1.209 2.146 2.524 1.176 6.630 5.910 0.891 E [ w | eS ] Data 2.997 2.365 0.789 3.520 3.156 0.897 5.207 5.228 1.004 Model 3.034 2.434 0.802 3.524 3.185 0.904 5.284 5.432 1.028 SD [ w | eS ] Data 2.269 2.184 0.962 2.771 3.206 1.157 4.360 4.770 1.094 Model 2.517 2.296 0.912 3.056 3.521 1.152 5.322 6.019 1.131 Note. E [ x ] is the average productivity, SD ( x ) is the standard deviation of productivity, GDP W is the GDP per worker, GDP C is the GDP per capita, E [ w | e ] is the average wage conditional on the employment status e , and finally SD [ w | e ] is the standard deviation of wages conditioning in the employment status e . 36 Table 7: Chile - Productivity and Wages Primary Secondary Tertiary M W W/M M W W/M M W W/M E[xF ] Model 5.080 4.936 0.972 5.823 5.134 0.882 13.382 10.585 0.791 SD ( x F ) Model 0.030 0.410 13.464 0.029 0.021 0.740 1.888 0.115 0.061 E[xI] Model 0.916 4.115 4.490 0.692 0.580 0.838 1.018 0.806 0.792 SD[xI] Model 2.301 1.852 0.805 1.511 1.711 1.132 3.111 6.594 2.119 E[xS] Model 2.034 2.345 1.153 0.785 1.429 1.821 4.441 3.217 0.724 SD[xS] Model 1.630 2.243 1.376 1.420 2.706 1.906 5.734 5.427 0.946 GDP W Model 4.206 3.715 0.883 5.265 4.489 0.853 12.261 10.059 0.820 GDP C Model 3.279 1.161 0.354 4.614 2.405 0.521 10.319 7.312 0.709 E [ w | eF ] Data 2.676 2.126 0.794 3.262 2.566 0.787 7.312 5.501 0.752 Model 2.698 2.121 0.786 3.254 2.594 0.797 7.210 5.481 0.760 SD [ w | eF ] Data 1.107 0.679 0.613 1.577 1.039 0.659 5.921 3.730 0.630 Model 1.114 0.634 0.569 1.463 0.998 0.682 5.620 3.600 0.641 E[w|eI] Data 2.315 2.004 0.866 2.798 2.372 0.848 5.730 4.983 0.870 Model 2.346 1.993 0.850 2.913 1.885 0.647 6.280 5.542 0.882 SD[w|eI] Data 1.122 1.381 1.232 1.707 1.560 0.914 5.458 3.787 0.694 Model 2.413 1.122 0.465 2.863 1.846 0.645 8.766 9.735 1.110 E [ w | eS ] Data 2.632 2.328 0.885 3.457 2.842 0.822 8.091 6.199 0.766 Model 2.666 2.337 0.877 3.449 2.956 0.857 8.037 6.569 0.817 SD[w|eS ] Data 2.020 2.289 1.133 3.110 2.764 0.889 9.040 6.670 0.738 Model 2.092 2.370 1.133 3.348 3.594 1.073 9.589 8.445 0.881 Note. E [ x ] is the average productivity, SD ( x ) is the standard deviation of productivity, GDP W is the GDP per worker, GDP C is the GDP per capita, E [ w | e ] is the average wage conditional on the employment status e , and finally SD [ w | e ] is the standard deviation of wages conditioning in the employment status e . 43 Figure 4: Childcare Provision Policy: Impact on GDP per Capita Note: Figure reports percent changes in GDP per capita as a result of Policy Experiment 1: reducing by half the average value of nonparticipation for mothers with children aged 5 or younger. Lightly colored bars represent the effect on GDP considering differences in average weekly hours worked by men and women. See Section 6 for more details. 44 Figure 5: Childcare Provision Policy: Impact on Female Participation Rates Note: Figure reports percent changes in female participation rates as a result of Policy Experiment 1: a range between 25% and 75% of reductions in the average value of nonparticipation for mothers with children aged 5 or younger. See Section 6 for more details. 45 Figure 6: Childcare Provision Policy: Impact on GDP per Capita Note: Figure reports percent changes in GDP per capita as a result of Policy Experiment 1: a range between 25% and 75% of reductions in the average value of nonparticipation for mothers with children aged 5 or younger is considered. See Section 6 for more details. 46 Figure 7: Increased Female Productivity Policy: Impact on Female Participation Rates Note: The overall length of the column is the post policy participation rate. The darker red segment is the impact of the policy. The number on top of each column is the percent change in participation rates post policy (i.e., the length of the darker red segment). The policy reported is Policy Experiment 2: increasing the average productivity of women by 10%, keeping the variance of the productivity constant. See Section 6 for more details. 47 Figure 8: Increased Female Productivity Policy: Impact on GDP per Capita Note: Figure reports percent changes in GDP per capita as a result of Policy Experiment 2: increasing the average productivity of women by 10%. Lightly colored bars represent the effect on GDP considering differences in average weekly hours worked by men and women. See Section 6 for more details. 48 Figure 9: Increased Female Productivity Policy: Impact on GDP per Capita by Channel Note: Figure reports percent changes in GDP per capita as a result of Policy Experiment 2: increasing the average productivity of women by 10%. See Section 6 for more details. The overall increase is decomposed in this portion due to the 10% productivity increase (pure productivity effect) and the portion due to the increase in participation resulting from the productivity increase (labor force effect). 49 Figure 10: Increased Female Productivity Policy: Impact on Female Participation Rates Note: Figure reports percent changes in participation rates as a result of Policy Experiment 2: A range between 1% and 20% increasing the average productivity of women is considered. See Section 6 for more details. 50 Figure 11: Increased Female Productivity Policy: Impact on GDP per Capita Note: Figure reports percent changes in GDP per capita as a result of Policy Experiment 2: A range between 1% and 20% increasing the average productivity of women is considered. See Section 6 for more details. 51 Appendix A Model A.1 Environment The environment is characterized by stationarity, continuous time, and infinitely lived individuals. There are two types of workers: i = M, W. There are five mutually exclusive states in which agents may be classified: nonparticipation (NP), unemployment (U), formal employment (F), informal employment (I), or self-employment (S). When nonparticipating, workers receive a flow utility z, where z ∼ Q i (z) with i = M, W. Only unemployed workers can search for a job and receive job offers. While searching for a job, workers receive a flow (dis)utility b i with i = M, W. Job opportunities arrive at a genderand employment-type specific Poisson rate λ ij , with i = M, W and j = F, I,S. If a job is accepted, subsequent job termination is possible and exogenous. Termination shocks arrive at a genderand employment-type specific Poisson rate δ ij . A job opportunity is characterized by a match-specific productivity x where x ∼ G ij (x). The flow pay for employees is w ij (x), where w ij is a specific genderand labor-related wage schedule determined by bargaining. Formal jobs are subject to a payroll social security contribution, collected at the proportional rate τ and withdrawn at the source by firms. Collected contributions are not redistributed among workers and are just a sunk cost. Informal jobs do not pay social security contributions, but they face the risk of being caught, which involves a penalty. The penalty has to be paid by the firm. This penalty is modeled as a constant flow cost c. The future is discounted at a rate ρ common to all the agents in the economy. A.2 Value Function The gender-specific flow value for an individual deciding not to participate in the labor market ρNPi(z) is the flow utility received by the individual for engaging in nonmarket activities z, that is: 𝜌𝑁𝑃 𝑖 (𝑧)=𝑧, 𝑖=𝑀,𝑊 (𝐴.1) 52 If the individual decides to participate, the gender-specific flow value of participating in the labor market is characterized by the flow value of an unemployed individual searching for a job ρU i , that is: 𝜌𝑈 𝑖 =𝑏 𝑖 +𝜆 𝑖𝐹 ∫𝑚𝑎𝑥[𝐸 𝑖𝐹 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝐹 (𝑥)+𝜆 𝑖𝐼 ∫𝑚𝑎𝑥[𝐸 𝑖𝐼 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝐼 (𝑥) +𝜆 𝑖𝑆 ∫𝑚𝑎𝑥[𝐸 𝑖𝑆 (𝑥),𝑈 𝑖 ]𝑑𝐺 𝑖𝑆 (𝑥)−(𝜆 𝑖𝐹 +𝜆 𝑖𝐼 +𝜆 𝑖𝑆 )𝑈 𝑖 , 𝑖=𝑀,𝑊 (𝐴.2) In each instant of time, an unemployed individual receives a flow utility b i and may meet a formal or an informal potential employee or receive a self-employment job opportunity. Meetings with formal firms, informal firms, and self-employment opportunities arrive at Poisson rates λ iF , λ iI , and λ iS , respectively. If a job opportunity of any type arrives, a match-specific productivity x is drawn from the corresponding productivity distribution G ij (x), with i = M, W and j = F, I, S. Note that the formality status in this model is assumed to be exogenous. For any type of job, the worker accepts the potential match if and only if the value of working at that productivity E ij (x) is higher than the value of staying in the unemployment state U i (this is captured by the maximum operator). Finally, if no job opportunity arrives, the individual remains unemployed and continues searching for a job. The gender-specific flow values of working as a formal employee, as an informal employee, or as self-employed in a match with specific productivity x are ρE iF (x), ρE iI (x), and ρEiS(x), respectively (with i = M, W), and are characterized as: 𝜌𝐸 𝑖𝐹 (𝑥)= 𝑤 𝑖𝐹 (𝑥)+𝛿 𝑖𝐹 [𝑈 𝑖 −𝐸 𝑖𝐹 (𝑥)] (𝐴.3) 𝜌𝐸 𝑖𝐼 (𝑥)= 𝑤 𝑖𝐼 (𝑥)+𝛿 𝑖𝐼 [𝑈 𝑖 −𝐸 𝑖𝐼 (𝑥)] (𝐴.4) 𝜌𝐸 𝑖𝑆 (𝑥)= 𝑥+ 𝛿 𝑖𝑆 [𝑈 𝑖 −𝐸 𝑖𝑆 (𝑥)] (𝐴.5) 59 𝑌 𝑖𝑝𝑤 =𝑌 𝑀𝑝𝑤 𝑁 𝑀 𝑤 𝑁 𝑀 𝑤 +𝑁 𝑊 𝑤 +𝑌 𝑊 𝑝𝑤 𝑁 𝑊 𝑤 𝑁 𝑀 𝑤 +𝑁 𝑊 𝑤 𝑌 𝑖𝑝𝑐 =𝑌 𝑀𝑝𝑐 𝑁 𝑀 𝑁 𝑀 +𝑁 𝑊 +𝑌 𝑊 𝑝𝑐 𝑁 𝑊 𝑁 𝑀 +𝑁 𝑊 Where 𝑁 𝑖𝑤 and 𝑁 𝑖 , with i=M,W, represent the number of working individuals (in either F, I, or S) and the total number of individuals, respectively. B Estimation and Identification The model is estimated by maximum likelihood methods and using supply-side data of the labor market. Data on (outgoing) unemployment duration t, wages in both formal (w F ) and informal (w I ) jobs, earning in self-employment activities (w S ), and individuals’ labor market status were used. B.1 Estimation Conditional to the model, the probability of observing an individual k in the nonparticipation state is P(z>z ∗ ). Given that z ∼ Q(z) and z ∗ =ρU i , the contribution to the likelihood of non-participation information is: P i ( k ∈ N P) = 1 − Q ( ρU i ) (B.1) To find the contribution of the unemployment duration information to this likelihood, we first define the total hazard rate out of unemployment as: h i = λ iF [1 − G iF ( x ∗ iF ) ] + λ iI [1 − G iI ( x ∗ iI ) ] + λ iS [1 − G iS ( x ∗ iS ) ] = hiF + hiI + hiS That is, the hazard rate is the probability that a match is formed once an individual meets a potential employer of any type of job (formal or informal) or self-employment opportunity. Recall that the match is formed only if the productivity drawn from the match 60 is greater than the corresponding reservation productivity. The hazard rate, conditional on the model, does not depend on the duration, and therefore the unemployment duration follows a negative exponential distribution with a coefficient equal to the hazard rate. Given that the unemployment duration is observed only for individuals who are actively participating in the labor market and are currently unemployed, the contribution of unemployment duration has to be weighted by the probability of participation ( Q ( ρU i )) and of being unemployed (the unemployment rate u i ). Given these considerations, the contribution of the unemployment duration information to the likelihood function is: f i,u ( t i,k , k ∈ U, k ∈ / N P ) = h i exp( − h i t i,k ) u i Q ( ρU i ) (B.2) To derive the contribution of wages to the likelihood function, it is necessary to make three considerations with respect to the data on wages. First, we have information about wages but not on productivity. Second, the observed wages are those related to matches already formed, and therefore they are accepted wages. Finally, we only observe data for those individuals who are currently employed (in one of the different types of jobs). To take into account these considerations under the structure of the theoretical model, we proceed in the following way. In the first step, we map the unconditional wage-cumulative distribution from the unconditional productivity-cumulative distribution ( G ij ( x )) using the wage equations for the formal and the informal sectors (in the case of those self-employed, the mapping is 1:1). In the second step, we construct the truncated version of the density wage distributions, taking into account the optimal decisions summarized in the reservation productivities ( x ∗ ij ). In the fi nal step, the truncated w age distributions are w eigh ted b y the probability of participation ( Q ( ρU i )) and of being employed (the employment rate in each type of job, e ij ). Under these considerations, wages’ contribution to the likelihood function in the cases of formal and informal sectors and self-employment, respectively, are: 𝑓 𝑒 𝑖𝐹 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝐹 ∗ ,𝑘∈𝐹,𝑘∉𝑁𝑃) =1+𝜏 𝛽𝑔 𝑖𝐹 ((1+𝜏)(𝑤 𝑖,𝑘 −(1−𝛽)𝜌𝑢 𝑖 𝛽) 1−𝐺 𝑖𝐹 ((1+𝜏)𝜌𝑈 𝑖 )𝑒 𝑖𝐹 𝑄(𝜌𝑈 𝑖 ) (𝐵.3) 61 𝑓 𝑒 𝑖𝐼 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝐼∗ ,𝑘∈𝐼,𝑘∉𝑁𝑃) =1𝛽𝑔 𝑖𝐼 ((𝑤 𝑖,𝑘 +𝛽𝑐−(1−𝛽)𝜌𝑢 𝑖 𝛽) 1−𝐺 𝑖𝐼 (𝜌𝑈 𝑖 +𝑐)𝑒 𝑖𝐼 𝑄(𝜌𝑈 𝑖 ) (𝐵.4) 𝑓 𝑒 𝑖𝑆 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝑆 ∗ ,𝑘∈𝑆,𝑘∉𝑁𝑃)= 𝑔 𝑖𝐼 (𝑤 𝑖,𝑘 ) 1−𝐺 𝑖𝑆 (𝜌𝑈 𝑖 )𝑒 𝑖𝑆 𝑄(𝜌𝑈 𝑖 ) (𝐵.5) Putting all the contributions together, the logarithm of the joint likelihood function to be maximized is: 𝑙𝑛𝐿(𝑤 𝑘 ,𝑡 𝑘 ,𝑈,𝐹,𝐼,𝑆,𝑁𝑃,𝜃) = ∑ {∑𝑙𝑛𝑃 𝑖 𝑘∈𝑁𝑃 (𝑘∈𝑁𝑃)+∑𝑙𝑛𝑓 𝑖,𝑢 𝑘∈𝑈 (𝑡 𝑖,𝑘 ,𝑘∈𝑈,𝑘∉𝑁𝑃 ) 𝑖=𝑀,𝑊 +∑𝑓 𝑒 𝑖𝐹 𝑘∈𝐹 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝐹 ∗ ,𝑘∈𝐹,𝑘∉𝑁𝑃 ) +∑𝑓 𝑒 𝑖𝐼 𝑘∈𝐼 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝐼∗ ,𝑘∈𝐼,𝑘∉𝑁𝑃 ) +∑𝑓 𝑒 𝑖𝑆 𝑘∈𝑆 (𝑤 𝑖,𝑘 ,𝑤 𝑖,𝑘 ≥𝑤 𝑖𝑆 ∗ ,𝑘∈𝑆,𝑘∉𝑁𝑃 )} Where 𝜃 is the vector of primitive parameters of the model. Using the contributions defined in equations (B.1) to (B.5) and making some algebraic manipulations, the logarithm of the joint likelihood function becomes: 62 𝑙𝑛𝐿(𝑤 𝑘 ,𝑡 𝑘 ,𝑈,𝐹,𝐼,𝑆,𝑁𝑃,𝜃) = ∑{𝑁 𝑁𝑃 ln(1−𝑄(𝜌𝑈 𝑖 ))+(𝑁 𝑈 +𝑁 𝐹 +𝑁 𝑆 +𝑁 𝐼 )𝑙𝑛𝑄(𝜌𝑈 𝑖 )+𝑁 𝑈 𝑙𝑛ℎ 𝑖 𝑖=𝑀,𝑊 +𝑁 𝑈 𝑙𝑛𝑢 𝑖 +𝑁 𝐹 𝑙𝑛𝑒 𝑖𝐹 +𝑁 𝐼 𝑙𝑛𝑒 𝑖𝐼 +𝑁 𝑆 𝑙𝑛𝑒 𝑖𝑆 −ℎ 𝑖 ∑𝑡 𝑖,𝑘 𝑘∈𝑈 +∑ln(1+𝜏 𝛽𝑔 𝑖𝐹 ((1+𝜏)(𝑤 𝑖,𝑘 −(1−𝛽)𝜌𝑢 𝑖 𝛽) 1−𝐺 𝑖𝐹 ((1+𝜏)𝜌𝑈 𝑖 )) 𝑘∈𝐹 +∑ln(1𝛽𝑔 𝑖𝐼 ((𝑤 𝑖,𝑘 +𝛽𝑐−(1−𝛽)𝜌𝑢 𝑖 𝛽) 1−𝐺 𝑖𝐼 (𝜌𝑈 𝑖 +𝑐)) 𝑘∈𝐼 +∑ln(𝑔 𝑖𝐼 (𝑤 𝑖,𝑘 ) 1−𝐺 𝑖𝑆 (𝜌𝑈 𝑖 )) 𝑘∈𝑆 } Finally, we make the following parametric assumptions about the gender-specific distribution Q i (z) and the genderand job-specific distributions G ij (x) (j = F, I, S). For the former, we assume a negative exponential distribution with parameter γ i , that is: Q i ( z ) = 1 − exp( − γ i z ) , z > 0 While for the latter we assume a log normal distribution with location and scale parameters µ ij and σ ij ; that is, the density function of G ij ( x ) is: 𝑔 𝑖𝑗 (𝑥)=1 𝜎 𝑖𝑗 𝑥𝜙(ln(𝑥)−𝑢 𝑖𝑗 𝜎 𝑖𝑗 ), 𝑥>0 Where 𝜙(.) is the standard normal density function. B.2 Identification The identification strategy closely follows Flinn and Heckman (1982). On one hand, the identification of the mobility parameters, hazard rates, and arrival rates of the termination shocks rely on the unemployment duration information and the steady-state equilibrium conditions. On the other hand, the identification of the productivity distributions (in all types 63 of jobs) relies on the idea of uniquely recovering the productivity and entire wage distributions from a truncated distribution with a known truncation point (the observed wage distributions). This can be done if the assumed distributions for the productivities meet what Flinn and Heckman (1982) called the recoverability condition. Starting with the mobility parameters and taking the first-order conditions of the maximization problem of the logarithm of the likelihood function with respect to the hazard rates: ℎ 𝑖𝐹 : 𝑁 𝑈 ℎ 𝑖 +𝑁 𝑈 ℎ 𝑖 𝜕𝑢 𝑖 𝜕ℎ 𝑖𝐹 +𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕ℎ 𝑖𝐹 +𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕ℎ 𝑖𝐹 +𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕ℎ 𝑖𝐹 −∑𝑡 𝑖,𝑘 =0 (𝐵.6) 𝑘∈𝑈 ℎ 𝑖𝐼 : 𝑁 𝑈 ℎ 𝑖 +𝑁 𝑈 𝑢 𝑖 𝜕𝑢 𝑖 𝜕ℎ 𝑖𝐼 +𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕ℎ 𝑖𝐼 +𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕ℎ 𝑖𝐼 +𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕ℎ 𝑖𝐼 −∑𝑡 𝑖,𝑘 =0 (𝐵.7) 𝑘∈𝑈 ℎ 𝑖𝑆 : 𝑁 𝑈 ℎ 𝑖 +𝑁 𝑈 𝑢 𝑖 𝜕𝑢 𝑖 𝜕ℎ 𝑖𝑆 +𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕ℎ 𝑖𝑆 +𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕ℎ 𝑖𝑆 +𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕ℎ 𝑖𝐼 −∑𝑡 𝑖,𝑘 =0 (𝐵.8) 𝑘∈𝑈 And with respect to the arrival rates of termination shocks: 𝛿 𝑖𝐹 : 𝑁 𝑈 𝑢 𝑖 𝜕𝑢 𝑖 𝜕𝛿 𝑖𝐹 + 𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕ℎ 𝑖𝐹 + 𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕ℎ 𝑖𝐹 + 𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕𝛿 𝑖𝐹 =0 (𝐵.9) 𝛿 𝑖𝐼 : 𝑁 𝑈 𝑢 𝑖 𝜕𝑢 𝑖 𝜕𝛿 𝑖𝐼 +𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕𝛿 𝑖𝐼 +𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕𝛿 𝑖𝐼 +𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕𝛿 𝑖𝐼 =0 (𝐵.10) 𝛿 𝑖𝑆 : 𝑁 𝑈 𝑢 𝑖 𝜕𝑢 𝑖 𝜕𝛿 𝑖𝑆 +𝑁 𝐹 𝑒 𝑖𝐹 𝜕𝑒 𝑖𝐹 𝜕𝛿 𝑖𝑆 +𝑁 𝐼 𝑒 𝑖𝐼 𝜕𝑒 𝑖𝐼 𝜕𝛿 𝑖𝑆 +𝑁 𝑆 𝑒 𝑖𝑆 𝜕𝑒 𝑖𝑆 𝜕𝛿 𝑖𝑆 =0 (𝐵.11) Where 𝜕𝑋 𝜕𝑌 is the partial derivative of the steady state condition X (u i , e iF , e iI , and e iS in equations A.18 to A.21) with respect to the parameter Y (the hazard rates and the termination shock Poisson rates). Note that equations (B.6) to (B.11) represent a nonlinear system of six equations with six unknowns for each gender type (h iF , h iI h iS , δ iF , δ iF , and δ iF ). These parameters are identified if the solution of this system of equations is unique. Thus, in the context of nonlinear systems of equations, there is a possibility of multiple solutions. To 64 overcome this possibility, we follow Bobba et al. (2017) and restrict the solutions to those that satisfy λ iF = λ iI and δ iF = δ iI , that is, for waged jobs, formal and informal, the arrival rates of meetings and terminations are the same. With respect to the productivity distributions, we assume, as discussed in the previous subsection, that they take a lognormal form. As discussed by Eckstein and van den Berg (2007), this parametrization meets the recoverability condition and belongs to a log locationscale family; therefore, the location and the scale of the original distribution should be identified from the location and the scale of the truncated distribution. To see this in the context of the distribution of the different types of jobs, we re-parametrize the observed wage distribution for the case of formal jobs in the following way: 1+𝜏 𝛽𝑔 𝑖𝐹 ((1+𝜏)(𝑤 𝑖,𝑘 −(1−𝛽)𝜌𝑢 𝑖 𝛽) 1−𝐺 𝑖𝐹 ((1+𝜏)𝜌𝑈 𝑖 )=1 𝑤 𝑖,𝑘 𝜎 𝑖𝐹,0 ∅ 𝑖𝐹 (ln (𝑤 𝑖,𝑘 −𝜇 𝑖𝐹,0 𝜎 𝑖𝐹,0 ) 1−Φ 𝑖𝐹 (ln (𝜌𝑈 𝑖 )−𝜇 𝑖𝐹,0 𝜎 𝑖𝐹,0 ) Where: 𝜇 𝑖𝐹,0 =(1−𝛽)𝜌𝑈 𝑖 +𝛽 (1+𝜏)𝜇 𝑖𝐹 (𝐵.12) 𝜎 𝑖𝐹,0 =𝛽 (1+𝜏)𝜎 𝑖𝐹 (𝐵.13) That is, µ iF, 0 and σ iF, 0 are the mean (location) and standard deviation (scale) of the observed wage distribution, respectively, and µ iF and σ iF are the mean (location) and standard deviation (scale) of the productivity distribution. From (B.12) and (B.13), it follows immediately that if ρU i , β and τ are known, then µ iF and σ iF are uniquely identified from the data on wages in the formal sector. The parameters β and τ are not identified, and therefore they are just set. We set β at 0 . 5 for all countries, while in the case of τ, we use information about the payroll contributions of each country. Using the same re-parametrization for the observed wage distribution for the case of informal jobs, we have: 65 1𝛽𝑔 𝑖𝐼 ((𝑤 𝑖,𝑘 +𝛽𝑐−(1−𝛽)𝜌𝑈 𝑖 𝛽) 1−𝐺 𝑖𝐼 (𝜌𝑈 𝑖 +𝑐)=1 𝑤 𝑖,𝑘 𝜎 𝑖𝐼,0 ∅ 𝑖𝐼 (ln (𝑤 𝑖,𝑘 −𝜇 𝑖𝐼,0 𝜎 𝑖𝐼,0 ) 1−Φ 𝑖𝐼 (ln (𝜌𝑈 𝑖 )−𝜇 𝑖𝐼,0 𝜎 𝑖𝐼,0 ) Where: 𝜇 𝑖𝐼,0 =(1−𝛽)𝜌𝑈 𝑖 +𝛽(𝜇 𝑖𝐼 − 𝑐) (𝐵.14) 𝜎 𝑖𝐼,0 =𝛽𝜎 𝑖𝐼 (𝐵.15) In this case, µ iI and σ iI are uniquely identified from the data if ρU i , β, and c are known, which means that the cost of informality has to be set using additional sources of information in order to identify the productivity distribution in the informal sector. We use the ratio between the cost of informality and the average wage in the formal sector estimated by Bobba et al. (2017) for the case of Mexico, and we use that ratio to set c across countries. Finally, the re-parametrization of observed wage distribution for the case of self-employed workers gives: 𝑔 𝑖𝐼 (𝑤 𝑖,𝑘 ) 1−𝐺 𝑖𝑆 (𝜌𝑈 𝑖 )=1 𝑤 𝑖,𝑘 𝜎 𝑖𝑆,0 ∅ 𝑖𝑆 (ln (𝑤 𝑖,𝑘 −𝜇 𝑖𝑆,0 𝜎 𝑖𝑆,0 ) 1−Φ 𝑖𝑆 (ln (𝜌𝑈 𝑖 )−𝜇 𝑖𝑆,0 𝜎 𝑖𝑆,0 ) Where: 𝜇 𝑖𝑆,0 =𝜇 𝑖𝑆 (𝐵.16) 𝜎 𝑖𝑆,0 =𝜎 𝑖𝑆 (𝐵.17) Given that there is no bargaining involved in self-employment, the location and scale of the productivity distribution in equations (B.16) and (B.18) are identified one to one from their counterparts in the observed wage distribution provided that ρU i is known. Flinn and Heckman (1982) showed that the minimum observed wage is a strongly consistent nonparametric estimator of the reservation wage. This estimator is typically used in the literature to estimate ρU i . However, because the model in this paper indicates that w iF ( x ∗ iF ) = w iI ( x ∗ iI ) = x ∗ iI = ρU i , the Flinn and Hec k man (1982) estimator implies that min 𝑤 𝑖𝐹 0 = min 𝑤 𝑖𝐼0 = min 𝑤 𝑖𝑆 0 = ρU i but nothing guarantees that these equalities hold in 66 the data. Instead, we attempt to estimate ρU i jointly with all the other parameters, maximizing the likelihood function. The problem that arises in this case is that ρU i determines the reservation productivities, which in turn are the truncation parameters in the accepted wage distributions in all types of job and changing this parameter in the maximization process of the likelihood function alters its support and violates one of the regularity conditions of the estimation method. To avoid this problem and because it is likely that wages are measured with error (particularly in self-employment), we introduce measurement error in the estimation. We assume that the measurement error E is multiplicative, and, therefore, the observed wages can be expressed as wo = w × E. The assumptions we make about the measurement error are threefold: (1) the measurement error is gender specific; (2) we use a lognormal distribution for the measuremen t error: 𝜐 𝑖 (𝜖)= 1 𝜖𝜎 𝜖𝑖 𝜙( 𝑙𝑛𝜖−𝜇 𝜖𝑖 𝜎 𝜖𝑖 ) , where 𝜙 (.) is the standard normal density function, i=M, W; (3) we assume that the conditional expectation of the observed wages is equal to the true wages, that is E [ w o | w ] = w , which implies that E[E|w] = 1. All these assumptions together imply that the parameters 𝜇 𝜖 𝑖 and 𝜎 𝜖 𝑖 satisfy 𝜎 𝜖 𝑖 =√−2𝜇 𝜖 𝑖 with i=M,W, and therefore only one parameter of the measurement error has to be estimated. Using the measurement error, the implied density functions of observed wages that should be used in the contributions of wages in all types of jobs to the likelihood function are: 𝑓𝑒𝑖𝐹 0(𝑤𝑖,𝑘 0)=∫ 1 𝑤𝑖 𝜌𝑈 𝑖 𝑣𝑖(𝑤𝑖,𝑘 0 𝑤𝑖)𝑓𝑒𝑖𝐹(𝑤𝑖,𝑤𝑖≥ 𝜌𝑈 𝑖 ,𝑘∈𝐹,𝑘∉𝑁𝑃 ) 𝑑 𝑤𝑖 (𝐵.18) 𝑓𝑒𝑖𝐼 0(𝑤𝑖,𝑘 0)=∫ 1 𝑤𝑖 𝜌𝑈 𝑖 𝑣𝑖(𝑤𝑖,𝑘 0 𝑤𝑖)𝑓𝑒𝑖𝐼(𝑤𝑖,𝑤𝑖≥ 𝑤 𝑖𝐼∗ ,𝑘∈𝐹,𝑘∉𝑁𝑃 ) 𝑑 𝑤𝑖 (𝐵.19) 𝑓𝑒𝑖𝑆 0(𝑤𝑖,𝑘 0)=∫ 1 𝑤𝑖 𝜌𝑈 𝑖 𝑣𝑖(𝑤𝑖,𝑘 0 𝑤𝑖)𝑓𝑒𝑖𝑆(𝑤𝑖,𝑤𝑖≥ 𝑤 𝑖𝑆 ∗ ,𝑘∈𝑆,𝑘∉𝑁𝑃 ) 𝑑 𝑤𝑖 (𝐵.20) 67 Finally, to identify the parameter γ i in Q i ( z ), the assumed distribution must be invertible with respect to its parameter, and the negative exponential distribution meets this requirement. The first-order condition of the maximum likelihood estimation gives the following estimator for this parameter: 𝛾 𝑖 =𝑙𝑛(𝑁 𝑖 𝑁 𝑖,𝑁𝑃 ) 𝜌𝑈 𝑖 Where N i is the total number of individuals and N i ,NP is the number of individuals who are not participating in the labor market by gender. To analyze the influence of the presence of children in the household on the participation rates (in particular in the γ i parameter), we divided nonparticipating individuals into three groups: first, those that have kids 5 years old or younger in the household (k5); second, those that have kids between 5 and 13 years old (k13); and third, the remaining nonparticipants (other). It can be shown that if Pr[NP ∩ k5] + Pr[NP ∩ k13] + Pr[NP ∩ other] = Pr[NP], the estimator of the parameter γ by group is: 𝛾 𝑖𝑔 =𝑙𝑛(𝑁 𝑖𝑔 𝑁 𝑖,𝑁𝑃 𝑔 ) 𝜌𝑈 𝑖 Where 𝑁 𝑖𝑔 is the total numbers of individuals in the group g and 𝑁 𝑖,𝑁𝑃 𝑔 is the number of individuals who are not participating in the group g by gender. C Complete Estimation Results Tables C.4, C.7, C.10, and C.13 report the estimated structural parameters of the model for each country, gender, and education group. Tables C.5, C.8, C.11, and C.14 report the implications for the labor market dynamics and the distribution across labor market states. As mentioned in the main text, we perform two policy experiments. Tables C.6, C.9, C.12, and C.15 report the impact of the policy experiments on a variety of labor market outcomes along with the same outcomes reported at benchmark. Finally, Tables C.16, C.17, and C.18 report aggregated results on participation rates and GDP per capita; the figures presented in the main text are based in these tables. 68 Table C.1: Argentina - Estimated Parameters Primary Secondary Tertiary Men Women Men Women Men Women ρU 0.2010 0.1481 1.7531 1.4020 1.8737 1.6045 (0.0407) (0.0955) (0.0545) (0.0469) (0.0994) (0.0566) λ F 0.1290 0.1270 0.2149 0.1825 0.2095 0.2009 (0.0068) (0.0060) (0.0117) (0.0056) (0.0057) (0.0060) λ S 0.0991 0.0492 0.1434 0.1192 0.0855 0.0496 (0.0149) (0.0069) (0.0164) (0.2133) (0.0043) (0.0022) δF 0.0235 0.0298 0.0166 0.0286 0.0115 0.0147 (0.0011) (0.0014) (0.0009) (0.0009) (0.0003) (0.0004) δ S 0.0194 0.0212 0.0106 0.0056 0.0100 0.0114 (0.0011) (0.0030) (0.0011) (0.0019) (0.0003) (0.0005) µ F 2.5652 2.3973 2.5337 2.4788 2.8459 2.8579 (0.0109) (0.0164) (0.0106) (0.0168) (0.0115) (0.0110) σ F 0.0055 0.0056 0.0023 0.0044 0.0015 0.0009 (0.0014) (0.0118) (0.0015) (0.0015) (0.0005) (0.0006) µI 1.6267 1.6492 0.2905 0.7025 -0.8285 -0.7050 (0.0096) (0.0199) (0.0543) (0.0203) (0.1254) (0.0513) σI 0.2555 0.3702 0.8894 0.8820 1.6094 1.6250 (0.0228) (0.0172) (0.0568) (0.0372) (0.0864) (0.0342) µS 0.9628 0.6250 0.3670 -1.1566 1.1756 1.0534 (0.1563) (0.0325) (0.2372) (1.3264) (0.0894) (0.1054) σS 0.5374 0.7032 0.8134 1.2801 0.7668 0.8915 (0.0491) (0.0196) (0.0784) (0.2740) (0.0433) (0.0536) σM E 0.4533 0.4495 0.4626 0.4834 0.4778 0.4574 (0.0055) (0.0095) (0.0057) (0.0077) (0.0061) (0.0054) γ 11.5668 4.4577 1.7097 0.6790 1.3644 0.9826 γk5 - 3.6072 - 0.5685 - 0.8183 γk13 - 4.7808 - 0.7131 - 1.0216 γother - 5.3368 - 0.7787 - 1.0859 b -16.2883 -4.6048 -14.1671 -10.3582 -2.3621 -1.6530 c 0.4717 0.4717 0.5350 0.5350 0.4710 0.4710 Likelihood -21,279 -11,291 -13,751 -9,427 -13,581 -17,417 N 7534 7637 4587 4759 4318 6503 Note: Bootstrap standard errors (based on 25 replications) in parentheses. Nonestimated parameters: β = 0.5, τ = 0.48, and ρ = 0.062. 75 Table C.8: Colombia - Labor Market Dynamics and States Primary Secondary Tertiary M W W/M M W W/M M W W/M hu Data 0.318 0.219 0.690 0.247 0.192 0.776 0.188 0.166 0.886 Model 0.322 0.220 0.683 0.247 0.206 0.834 0.188 0.166 0.886 hu→eF Model 0.075 0.038 0.508 0.144 0.076 0.527 0.099 0.087 0.883 hu→e I Model 0.075 0.038 0.508 0.033 0.076 2.323 0.009 0.008 0.912 hu→eS Model 0.173 0.144 0.834 0.071 0.054 0.771 0.079 0.070 0.885 u Data 0.066 0.125 1.890 0.068 0.136 2.012 0.098 0.145 1.486 Model 0.066 0.125 1.898 0.068 0.129 1.913 0.098 0.145 1.486 eF Data 0.194 0.101 0.520 0.428 0.310 0.725 0.530 0.531 1.002 Model 0.168 0.121 0.718 0.427 0.214 0.501 0.530 0.531 1.002 eI Data 0.143 0.141 0.988 0.096 0.129 1.342 0.049 0.051 1.032 Model 0.168 0.121 0.718 0.097 0.214 2.209 0.049 0.051 1.034 eS Data 0.597 0.633 1.060 0.409 0.426 1.041 0.323 0.273 0.845 Model 0.597 0.633 1.060 0.409 0.443 1.084 0.323 0.273 0.845 np Data 0.076 0.450 5.907 0.046 0.315 6.919 0.064 0.164 2.572 Model 0.076 0.450 5.907 0.046 0.315 6.919 0.064 0.164 2.572 76 Table C.9: Colombia - Policy Experiments Benchmark P. Exp. No. 1 P. Exp. No. 2 M W W/M W W/M W W/M Primary u 0.066 0.125 1.898 0.125 1.898 0.125 1.899 eF 0.168 0.121 0.718 0.121 0.718 0.121 0.718 eI 0.168 0.121 0.718 0.121 0.718 0.121 0.718 eS 0.597 0.633 1.06 0.633 1.06 0.633 1.06 np 0.076 0.45 5.907 0.369 4.836 0.024 0.309 hu 0.322 0.22 0.683 0.22 0.683 0.22 0.683 GDPW 1.714 1.301 0.759 1.302 0.759 1.432 0.835 GDPC 1.48 0.626 0.423 0.719 0.486 1.223 0.827 E[w|eF] 1.3 1.242 0.955 1.248 0.96 1.379 1.061 E[w|eI] 1.087 0.84 0.772 0.843 0.775 0.978 0.9 E[w|eS] 1.131 0.839 0.741 0.835 0.738 0.925 0.818 Res. W. 0.095 0.021 0.221 0.021 0.221 0.098 1.036 Secondary u 0.068 0.129 1.913 0.129 1.913 0.132 1.95 eF 0.427 0.214 0.501 0.214 0.501 0.218 0.511 eI 0.097 0.214 2.209 0.214 2.209 0.218 2.251 eS 0.409 0.443 1.084 0.443 1.084 0.432 1.058 np 0.046 0.315 6.919 0.238 5.232 0.249 5.461 hu 0.247 0.206 0.834 0.206 0.834 0.204 0.824 GDPW 2.041 1.821 0.892 1.843 0.903 2.048 1.003 GDPC 1.817 1.086 0.598 1.222 0.673 1.336 0.735 E[w|eF] 1.452 1.336 0.92 1.331 0.917 1.488 1.025 E[w|eI] 1.105 0.974 0.882 0.972 0.88 1.094 0.99 E[w|eS] 1.405 1.23 0.875 1.29 0.918 1.423 1.012 Res. W. 0.797 0.329 0.412 0.328 0.412 0.396 0.497 Tertiary u 0.098 0.145 1.486 0.145 1.486 0.147 1.506 eF 0.53 0.531 1.002 0.531 1.002 0.538 1.015 eI 0.049 0.051 1.034 0.051 1.034 0.043 0.863 eS 0.323 0.273 0.845 0.273 0.845 0.272 0.843 np 0.064 0.164 2.572 0.111 1.738 0.103 1.628 hu 0.188 0.166 0.886 0.166 0.886 0.163 0.872 GDPW 5.2 4.786 0.92 4.782 0.92 5.311 1.021 GDPC 4.393 3.421 0.779 3.636 0.828 4.06 0.924 E[w|eF] 3.045 2.76 0.907 2.758 0.906 3.114 1.023 E[w|eI] 1.392 1.288 0.925 1.295 0.93 1.55 1.113 E[w|eS] 3.066 2.728 0.89 2.691 0.878 2.952 0.963 Res. W. 0.902 0.845 0.937 0.845 0.937 1.059 1.174 77 Table C.10: Mexico - Estimated Parameters Primary Secondary Tertiary Men Women Men Women Men Women ρU 0.0779 0.0866 0.9945 0.6806 1.4056 1.1648 (0.0342) (0.0048) (0.0093) (0.0079) (0.0704) (0.0306) λ F 0.2604 0.1790 0.2614 0.2914 0.2168 0.2750 (0.0247) (0.0166) (0.0100) (0.0206) (0.0093) (0.0198) λ S 0.2824 0.3073 0.3034 0.5872 0.1744 0.4172 (0.0138) (0.0236) (0.0249) (0.0992) (0.0141) (0.4063) δF 0.0290 0.0291 0.0236 0.0336 0.0240 0.0246 (0.0028) (0.0027) (0.0009) (0.0024) (0.0010) (0.0018) δ S 0.0384 0.0248 0.0247 0.0179 0.0442 0.0242 (0.0019) (0.0019) (0.0020) (0.0022) (0.0028) (0.0069) µ F 1.2960 1.0563 1.0638 1.0281 1.8190 1.8075 (0.0208) (0.0228) (0.0069) (0.0092) (0.0137) (0.0089) σ F 0.1146 0.1178 0.0034 0.0189 0.0180 0.0228 (0.1174) (0.0869) (0.0009) (0.0041) (0.1276) (0.0094) µ I 0.9046 0.6911 0.1910 -0.1791 -0.3012 -0.6902 (0.0157) (0.0215) (0.0067) (0.0089) (0.1209) (0.0515) σ I 0.1622 0.3504 0.4401 0.7646 0.9147 1.1594 (0.0893) (0.0451) (0.0153) (0.0121) (0.0911) (0.0369) µS 0.3908 -0.1133 -0.3031 -1.6270 0.5568 -1.2754 (0.0283) (0.0278) (0.1370) (0.2744) (0.1216) (0.8081) σS 0.5210 0.7612 0.8395 1.3079 0.7455 1.2789 (0.0487) (0.0291) (0.0489) (0.0728) (0.0540) (0.1591) σM E 0.3716 0.3206 0.4322 0.4432 0.5736 0.5552 (0.1504) (0.1041) (0.0028) (0.0038) (0.0233) (0.0045) γ 25.1856 4.2741 2.6676 0.8351 1.6379 0.8487 γk5 - 3.7244 - 0.6902 - 0.7738 γk13 - 4.6411 - 0.8890 - 0.8623 γother - 4.5132 - 0.9857 - 0.8958 b -13.7186 -9.0288 -3.464 -4.5471 -6.6944 -8.2247 c 0.1495 0.1495 0.1669 0.1669 0.2116 0.2116 Likelihood -18023 -9219 -53030 -30738 -31751 -28936 N 10048 15100 26008 32155 12385 17086 Note: Bootstrap standard errors (based on 25 replications) in parentheses. Nonestimated parameters: β = 0 . 5, τ = 0 . 33, and ρ = 0 . 056. 78 Table C.11: Mexico - Labor Market Dynamics and States Primary Secondary Tertiary M W W/M M W W/M M W W/M hu Data 0.804 0.665 0.827 0.512 0.535 1.047 0.366 0.383 1.045 Model 0.803 0.665 0.828 0.511 0.536 1.047 0.366 0.383 1.045 hu→eF Model 0.260 0.179 0.687 0.261 0.291 1.115 0.217 0.275 1.269 hu → e I Model 0.260 0.179 0.687 0.140 0.144 1.023 0.043 0.053 1.240 hu→eS Model 0.282 0.307 1.087 0.110 0.100 0.916 0.107 0.055 0.515 u Data 0.038 0.039 1.026 0.045 0.051 1.149 0.070 0.060 0.860 Model 0.038 0.039 1.025 0.045 0.051 1.149 0.070 0.060 0.860 eF Data 0.279 0.228 0.815 0.493 0.447 0.906 0.635 0.674 1.061 Model 0.341 0.240 0.703 0.493 0.443 0.899 0.635 0.673 1.060 e I Data 0.403 0.252 0.625 0.265 0.215 0.810 0.125 0.129 1.032 Model 0.341 0.240 0.703 0.265 0.219 0.825 0.125 0.129 1.036 eS Data 0.280 0.481 1.721 0.197 0.287 1.455 0.170 0.137 0.807 Model 0.280 0.481 1.721 0.197 0.287 1.455 0.170 0.137 0.807 np Data 0.141 0.691 4.912 0.070 0.566 8.042 0.100 0.372 3.720 Model 0.141 0.691 4.912 0.070 0.566 8.042 0.100 0.372 3.720 79 Table C.12: Mexico - Policy Experiments Benchmark P. Exp. No. 1 P. Exp. No. 2 M W W/M W W/M W W/M Primary u 0.038 0.039 1.025 0.039 1.025 0.039 1.033 eF 0.341 0.240 0.703 0.240 0.703 0.242 0.708 eI 0.341 0.240 0.703 0.240 0.703 0.242 0.708 eS 0.280 0.481 1.721 0.481 1.721 0.478 1.708 np 0.141 0.691 4.912 0.623 4.428 0.387 2.750 hu 0.803 0.665 0.828 0.665 0.828 0.661 0.822 GDPW 2.683 1.858 0.693 1.850 0.690 2.052 0.765 GDPC 2.218 0.552 0.249 0.671 0.302 1.209 0.545 E[w|eF] 1.420 1.126 0.793 1.130 0.796 1.305 0.919 E[w|eI] 1.216 1.032 0.849 1.031 0.848 1.202 0.989 E[w|eS] 1.700 1.203 0.708 1.194 0.703 1.338 0.787 Res. W. 0.078 0.087 1.111 0.087 1.111 0.222 2.853 Secondary u 0.045 0.051 1.149 0.051 1.149 0.052 1.177 eF 0.493 0.443 0.899 0.443 0.899 0.454 0.921 eI 0.265 0.219 0.825 0.219 0.825 0.221 0.834 eS 0.197 0.287 1.455 0.287 1.455 0.272 1.381 np 0.07 0.566 8.042 0.475 6.75 0.507 7.197 hu 0.511 0.536 1.047 0.536 1.047 0.526 1.029 GDPW 2.387 2.229 0.934 2.242 0.939 2.475 1.037 GDPC 2.121 0.917 0.432 1.116 0.526 1.157 0.545 E[w|eF] 1.587 1.39 0.876 1.392 0.877 1.554 0.979 E[w|eI] 1.294 1.136 0.878 1.136 0.878 1.28 0.989 E[w|eS] 1.968 1.71 0.869 1.745 0.887 1.923 0.977 Res. W. 0.995 0.681 0.684 0.681 0.684 0.814 0.818 Tertiary u 0.07 0.06 0.86 0.06 0.86 0.062 0.888 eF 0.635 0.673 1.06 0.673 1.06 0.695 1.095 eI 0.125 0.129 1.036 0.129 1.036 0.121 0.969 eS 0.17 0.137 0.807 0.137 0.807 0.122 0.715 np 0.1 0.372 3.72 0.299 2.985 0.293 2.927 hu 0.366 0.383 1.045 0.383 1.045 0.37 1.01 GDPW 5.194 5.193 1 5.196 1 5.824 1.121 GDPC 4.346 3.064 0.705 3.425 0.788 3.862 0.889 E[w|eF] 3.019 2.874 0.952 2.858 0.947 3.249 1.076 E[w|eI] 2.138 2.046 0.957 2.095 0.98 2.392 1.118 E[w|eS] 3.175 2.705 0.852 2.761 0.87 3.101 0.977 Res. W. 1.406 1.165 0.829 1.165 0.829 1.447 1.029 80 Table C.13: Peru - Estimated Parameters Primary Secondary Tertiary Men Women Men Women Men Women ρU 0.0800 0.0427 0.4793 0.1858 1.3229 0.5062 (0.0468) (0.0398) (0.2612) (0.0565) (0.0374) (0.0549) λ F 0.2838 0.0802 0.2595 0.1627 0.4895 0.4508 (0.0595) (0.0171) (0.0360) (0.0286) (0.0452) (0.0404) λ S 0.4567 2.0834 0.3805 1.0895 0.6617 0.6736 (0.0759) (1.1200) (0.0504) (0.4362) (0.2175) (0.0486) δF 0.0197 0.0325 0.0225 0.0214 0.0314 0.0390 (0.0028) (0.0073) (0.0038) (0.0035) (0.0029) (0.0035) δ S 0.0176 0.0963 0.0225 0.0376 0.0285 0.0666 (0.0018) (0.0382) (0.0031) (0.0206) (0.0052) (0.0060) µ F 1.6319 1.3617 1.6895 1.5393 2.0606 2.1616 (0.0196) (0.1866) (0.0612) (0.0275) (0.0133) (0.0163) σ F 0.0055 0.0286 0.0031 0.0035 0.0217 0.0040 (0.0021) (0.0108) (0.0008) (0.0010) (0.0088) (0.0012) µI 1.2022 0.8186 1.1664 0.9976 -0.5798 -1.5607 (0.0176) (0.0443) (0.0850) (0.0432) (0.0864) (0.1454) σI 0.0024 0.0009 0.0072 0.0039 1.0989 1.9203 (0.0009) (0.0264) (0.0024) (0.0011) (0.0729) (0.1155) µS 0.4670 -0.2500 0.6618 -0.0873 -0.3841 -0.6274 (0.0200) (0.4952) (0.1107) (0.2537) (0.4778) (0.1764) σS 0.6023 1.5907 0.4817 1.3992 1.1980 1.4102 (0.0297) (0.3640) (0.0593) (0.1706) (0.1254) (0.0736) σM E 0.5999 0.6495 0.5999 0.6495 0.5999 0.6495 - - - - (0.0075) (0.0117) γ 29.7031 25.5737 6.1053 5.9528 2.1426 3.0565 γk5 - 21.3520 - 4.9705 - 2.4312 γk13 - 29.7499 - 6.8259 - 3.3043 γother - 28.3009 - 6.7401 - 3.6857 b -18.0207 -36.1811 -17.338 -29.5757 -19.3774 -24.1667 c 0.2052 0.2052 0.2390 0.2390 0.2675 0.2675 Likelihood -7714 -7810 -12471 -7028 -18485 -16129 N 3438 6132 5039 4319 6519 6898 Note: Bootstrap standard errors (based on 25 replications) in parentheses. Nonestimated parameters: β = 0.5, τ = 0.24, and ρ = 0.067. 81 Table C.14: Peru - Labor Market Dynamics and States Primary Secondary Tertiary M W W/M M W W/M M W W/M hu Data 0.964 1.355 1.406 0.878 1.398 1.591 0.765 0.895 1.169 Model 1.024 2.173 2.122 0.899 1.276 1.420 0.765 0.910 1.189 hu→eF Model 0.284 0.080 0.283 0.259 0.163 0.627 0.490 0.451 0.921 hu→e I Model 0.284 0.080 0.283 0.259 0.163 0.627 0.084 0.112 1.338 hu→eS Model 0.457 2.013 4.407 0.380 0.951 2.504 0.192 0.347 1.810 u Data 0.019 0.025 1.302 0.025 0.033 1.282 0.038 0.048 1.240 Model 0.018 0.037 2.080 0.024 0.024 0.985 0.038 0.048 1.259 eF Data 0.202 0.047 0.233 0.348 0.168 0.482 0.601 0.564 0.939 Model 0.259 0.092 0.355 0.282 0.183 0.650 0.600 0.560 0.933 eI Data 0.315 0.143 0.454 0.215 0.222 1.034 0.102 0.134 1.316 Model 0.259 0.092 0.355 0.282 0.183 0.650 0.103 0.139 1.356 eS Data 0.464 0.785 1.692 0.412 0.578 1.402 0.259 0.254 0.982 Model 0.465 0.779 1.676 0.412 0.610 1.479 0.259 0.252 0.975 np Data 0.093 0.336 3.619 0.054 0.331 6.175 0.059 0.213 3.622 Model 0.093 0.336 3.619 0.054 0.331 6.175 0.059 0.213 3.622 82 Table C.15: Peru - Policy Experiments Benchmark P. Exp. No. 1 P. Exp. No. 2 M W W/M W W/M W W/M Primary u 0.018 0.037 2.08 0.037 2.08 0.045 2.529 eF 0.259 0.092 0.355 0.092 0.355 0.112 0.432 eI 0.259 0.092 0.355 0.092 0.355 0.112 0.432 eS 0.465 0.779 1.676 0.779 1.676 0.731 1.573 np 0.093 0.336 3.619 0.237 2.553 0 0.002 hu 1.024 2.173 2.122 2.173 2.122 1.714 1.673 GDPW 3.126 2.859 0.915 2.933 0.938 3.905 1.249 GDPC 2.785 1.828 0.657 2.155 0.774 3.727 1.338 E[w|eF] 2.11 1.57 0.744 1.554 0.736 1.901 0.901 E[w|eI] 1.611 1.059 0.657 1.062 0.659 1.324 0.822 E[w|eS] 1.908 2.857 1.497 2.995 1.569 4.053 2.124 Res. W. 0.08 0.043 0.533 0.043 0.533 0.339 4.237 Secondary u 0.024 0.024 0.985 0.024 0.985 0.026 1.068 eF 0.282 0.183 0.65 0.183 0.65 0.199 0.705 eI 0.282 0.183 0.65 0.183 0.65 0.199 0.705 eS 0.412 0.61 1.479 0.61 1.479 0.577 1.399 np 0.054 0.331 6.175 0.231 4.31 0.082 1.531 hu 0.899 1.276 1.42 1.276 1.42 1.156 1.286 GDPW 3.409 3.085 0.905 3.154 0.925 3.718 1.091 GDPC 3.148 2.015 0.64 2.367 0.752 3.324 1.056 E[w|eF] 2.437 1.996 0.819 1.955 0.802 2.256 0.926 E[w|eI] 1.736 1.346 0.775 1.32 0.76 1.608 0.926 E[w|eS] 2.166 2.705 1.249 2.846 1.314 3.468 1.601 Res. W. 0.479 0.186 0.388 0.186 0.388 0.42 0.876 Tertiary u 0.038 0.048 1.259 0.048 1.259 0.049 1.277 eF 0.6 0.56 0.933 0.56 0.933 0.568 0.946 eI 0.103 0.139 1.356 0.139 1.356 0.139 1.354 eS 0.259 0.252 0.975 0.252 0.975 0.244 0.943 np 0.059 0.213 3.622 0.139 2.37 0.135 2.292 hu 0.765 0.91 1.189 0.91 1.189 0.892 1.166 GDPW 6.21 6.464 1.041 6.493 1.046 7.227 1.164 GDPC 5.62 4.842 0.862 5.318 0.946 5.947 1.058 E[w|eF] 3.827 3.76 0.983 3.743 0.978 4.194 1.096 E[w|eI] 2.201 2.314 1.051 2.486 1.13 2.798 1.271 E[w|eS] 3.529 2.585 0.732 2.582 0.732 2.953 0.837 Res. W. 1.323 0.506 0.383 0.506 0.383 0.656 0.496 83 Table C.16: Policy Effects on GDP per Capita Policy Exp. No. 1 Policy Exp. No.2 Argentina Primary 6.7% 43.1% Secondary 5.3% 11.2% Tertiary 3.9% 15.2% Total 5.0% 22.0% Chile Primary 5.9% 35.3% Secondary 5.8% 9.0% Tertiary 3.7% 9.4% Total 4.6% 13.0% Colombia Primary 5.1% 32.3% Secondary 4.4% 8.7% Tertiary 3.4% 10.0% Total 3.9% 13.4% Mexico Primary 6.1% 32.7% Secondary 7.3% 8.9% Tertiary 5.8% 12.8% Total 6.4% 14.0% Peru Primary 7.4% 52.8% Secondary 5.3% 22.0% Tertiary 4.1% 10.3% Total 5.0% 20.5% Note: Percentage of GDP per capita with respect to the benchmark case. 84 Table C.17: Policy Effects on Participation Rates Benchmark Policy Exp. No. 1 Policy Exp. No. 2 M W W W Argentina Primary 0.902 0.483 0.578 0.981 Secondary 0.950 0.614 0.703 0.706 Tertiary 0.922 0.793 0.850 0.893 Total 0.762 0.804 0.900 Ratio w.r.t. Benchmark 1.056 1.182 Chile Primary 0.849 0.361 0.429 0.697 Secondary 0.942 0.589 0.678 0.660 Tertiary 0.902 0.779 0.841 0.850 Total 0.722 0.763 0.807 Ratio w.r.t. Benchmark 1.056 1.118 Colombia Primary 0.924 0.550 0.631 0.976 Secondary 0.954 0.685 0.762 0.751 Tertiary 0.936 0.836 0.889 0.897 Total 0.802 0.841 0.906 Ratio w.r.t. Benchmark 1.049 1.130 Mexico Primary 0.859 0.309 0.377 0.613 Secondary 0.930 0.434 0.525 0.493 Tertiary 0.900 0.628 0.701 0.707 Total 0.650 0.696 0.720 Ratio w.r.t. Benchmark 1.071 1.107 Peru Primary 0.907 0.664 0.763 1.000 Secondary 0.946 0.669 0.769 0.918 Tertiary 0.941 0.787 0.861 0.865 Total 0.817 0.865 0.930 Ratio w.r.t. Benchmark 1.059 1.139