Who suffers during recessions in Brazil?

Cravo, Túlio; Schimanski, Caroline Working Paper Who suffers during recessions in Brazil? IDB Working Paper Series, No. IDB-WP-975 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Cravo, Túlio; Schimanski, Caroline (2019) : Who suffers during recessions in Brazil?, IDB Working Paper Series, No. IDB-WP-975, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0001640 This Version is available at: https://hdl.handle.net/10419/208165 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 Who Suffers During Recessions in Brazil? Túlio Cravo Caroline Schimanski IDB WORKING PAPER SERIES Nº (IDB-WP-00975) March 2019 Labor Markets Division Inter-American Development Bank March 2019 Who Suffers During Recessions in Brazil? Túlio Cravo Caroline Schimanski Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Cravo, Túlio. Who suffers during recessions in Brazil? / Túlio Cravo, Caroline Schimanski. p. cm. — (IDB Working Paper Series ; 975) Includes bibliographic references. 1. Recessions-Brazil. 2. Labor market-Brazil. 3. Business cycles-Brazil. I. Schimanski, Caroline. II. Inter-American Development Bank. Labor Markets Division. III. Title. IV. Series. IDB-WP-975 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. http://www.iadb.org 2019 Working Paper | IDB-WP-00975 Who Suffers During Recessions in Brazil? Túlio A. Cravo Caroline Schimanski 2019 Working Paper | IDB-WP-00975 Contents 1. Introduction .................................................................................................... 4 2. Data ............................................................................................................... 6 3. Methodology .................................................................................................. 8 4. Results and Discussion ................................................................................ 10 4.1 Descriptive Results ............................................................................... 10 4.2 Regression Results ............................................................................... 15 5. Conclusion and policy implications .............................................................. 23 6. References ................................................................................................... 25 7. Appendix ...................................................................................................... 29 Working Paper | IDB-WP-00975 Who Suffers During Recessions in Brazil? Túlio A. Cravo I Caroline Schimanski II Abstract While the relationship between business cycles and employment is a topic of continuing interest, it has received limited attention in the literature focusing on developing countries. This study adds to the literature as it analyzes the heterogeneous correlations of the business cycle with different age, education, and ethnic groups by gender in a developing country setting, controlling for other characteristics. Using data from Brazil’s monthly PME employment surveys between 2002 and 2016, regressions are estimated to assess how business cycles affect employment in specific demographic groups. The results provide evidence of large heterogeneities among demographic groups. Interestingly, unemployment rates in levels across demographic groups are not necessarily aligned with the sensitivity of these demographic groups’ (un)employment and participation rates to economic crises. JEL Codes: E24, J21, J23, J24, J46 Keywords: Labor Market, Brazil, Developing Countries, Business Cycles, Heterogeneous Effects, Employment Dynamics I Corresponding author: Tulio A. Cravo, Senior Labor Market Specialist, Inter-American Development Bank, tcravo@iadb. II Caroline Schimanski, Consultant, UNU-WIDER, [email protected] Acknowledgements: We are thankful for comments from Eduardo Pontual Ribeiro, Paulo Jacinto and one anonymous reviewer. Working Paper | IDB-WP-00975 1. Introduction One of the most perverse results of recessions is employment destruction. Understanding employment dynamics in different demographic groups across business cycles is therefore paramount to the design of public policy that are better targeting specific demographic groups. In recent years, some studies, mainly in developed countries, provided empirical evidence on the dynamics of employment for specific demographic groups. Most papers on the heterogeneity of labor market outcomes across demographic groups address how wages, unemployment, and employment vary in a static manner. For instance, Card and Lemieux (1997) show that for the US and Canada unemployment rates vary across demographic groups, and Card and Krueger (1993) estimate large earning differences in the US between whites and blacks. For Brazil, Garcia et al. (2009) and Salardi (2012) show the existence of wage gaps based on gender and race in favor of men and whites, and that the racial wage gap is larger than the gender wage gap. Bourginon et al. (2007) measure the impact of inequality of opportunities for different demographic groups on income and show that current earnings are determined by parental characteristics and household income. Besides, Foguel et al. (2000) estimate the existence of a wage gap between the public and private sector and show an overrepresentation of women, older, and more educated workers and an underrepresentation of non-whites in the public sector. Another line in the literature addresses how labor market measures across demographic groups are affected during business cycles. Bredemeier and Winkler (2015) find that economic downturns in the US have a greater effect on workers who are young, women, less educated, and blue-collar. Similarly, Bachman et al. (2015) find that in Europe men and young persons are more responsive to business cycles. Kelly et al. (2016) show a rising gap between immigrants and natives in Ireland and Burnette (2017) between Native Americans and non-natives during economic crises. Reis (2017) shows that in Brazil the probability of transition from unemployment to employment varies among demographic groups and over Working Paper | IDB-WP-00975 the business cycle. According to Dureya and Arends-Kuenning (2003), children in Brazil are less likely to drop out of school to work during an economic crisis. However, none of these studies control extensively for other demographic characteristics. A different but important way to look at heterogeneity in the labor market is to analyze how different demographic groups respond to business cycles, while controlling for other demographic characteristics and structural changes in the composition of demographic groups. Hoynes et al. (2012) are the first to do so for different population groups in the US, providing an overall analysis about how these groups respond to business cycles in the US. Hoynes et al. (2012) use state-level panel data and provide results for the relationship between a business cycle indicator and labor market outcomes for each demographic group while controlling for other demographic characteristics. This allows the authors to isolate the sensitivity of a specific population group to cycles. Such analysis provides an idea of whether business cycles reinforce the heterogeneity of demographic groups’ labor market outcomes discussed in the literature (e.g., Salardi (2014)) and is an important guide for public policy related to issues such as gender and racial discrimination, youth employment, and educational and training policies during cycles, individually targeted at specific demographic groups. However, the validity of their results for other countries is unclear. To date there is a general lack of studies on how economic cycles affect different demographic groups controlling for other demographic factors, particularly for developing countries. However, business cycle effects on labor market outcomes can have much further reaching impacts on people’s household income in developing countries, which are characterized by more volatile economies. As such, academics and policy-makers can benefit greatly from deeper knowledge about the factors influencing employment dynamics in different demographic groups during business cycles in developing countries. This paper aims to fill this gap in the literature by exploring Brazil’s monthly PME employment surveys from March 2002 to February 2016, covering three economic downturns, to provide Working Paper | IDB-WP-00975 rate for all age and education levels, genders, and ethnicities, whereas changes in participation rates vary in direction. Table 1: Heterogeneity in Labor Market Outcomes in February 2016 Unemployment Rate (%) Employmen t rate (%) Usual Weekly Hours Estimate s Hourly Income Monthly Income (in Feb. 2016 Reais) All 8.0 % 49.6 % 38 12.9 2087 Race-Gender White men 6.5 % 58.8 % 40 17.3 2896 White women 7.9 % 42.2 % 36 14.6 2222 Non-White men 7.9 % 58.3 % 40 10.6 1798 Non-White women 10.3 % 42.2 % 37 9.0 1396 Education Less than middle school graduate 6.8 % 28.1 % 38 7.4 1204 Middle school grad/High school dropout/incomplete 10.9 % 43.0 % 39 8.0 1309 High school grad/College dropout/incomplete 9.3 % 61.6 % 39 10.4 1707 College grad/Master/PhD 4.2 % 72.8 % 37 28.4 4382 Age Group 10-17 38.3 % 3.5 % 29 5.8 703 18-24 20.5 % 49.2 % 38 7.5 1189 25-39 8.1 % 74.9 % 39 12.3 2010 40-59 4.6 % 67.8 % 39 14.1 2294 60 and older 1.8 % 18.3 % 37 17.8 2760 Formality Formal 38.4 % 39 14.0 2312 Informal 11.3 % 36 8.8 1329 Source: Authors' calculations based on PME survey February 2016 Note: Formality is based on the social security status definition following Henley et al. (2009). While based on larger increases of the unemployment rate for white men and white women than for non-whites, white workers appear at first side harder hit by the recession. This phenomenon/observation is affected by varying degrees of changes in the participation rate by race. The twice as high percentage point decrease in the labor market participation rate for non-whites compared to whites may well more than overcompensate for changes in the rate of people out of work, and as a result let the change in the unemployment rate for non-whites appear lower. In terms of education, percentage point increases in net unemployment rates 12 are highest for the lower educated groups. By age group, rather than the youngest group, the second youngest group between 18 and 24 years old appears 12 Net unemployment rate changes are estimated by subtracting the percentage point changes in the participation rate from the percentage point changes in the unemployment rate. This means that for the less than middle school educated group, 3.4 – (-3.2) = 6.6 Working Paper | IDB-WP-00975 most affected. In terms of formality of employment, the recession leads to a slower decrease in the employment rate in the informal sector. 13 Table A.2 in the Appendix illustrates that this pattern also largely holds when considering each metropolitan area separately. Figure 1: (In)formal Employment Rate and Share of Informal Workers over Time (based on Social Security Def.) Source: Authors' calculations based on raw PME surveys from March 2002 to February 2016 Note: (In)formality is here defined based on social security contributions following Henley et al. (2009). Employment rate Informal (Informal workers/whole working age population), Share of Informal Workers (Informal Workers/all occupied), Employment Rate Formal (Formal workers/whole working age population) Looking at the formality case in more detail, a plot of employment rate differentials over time, displayed in Figure 1, shows that the informal employment rate is generally decreasing over time and the formal employment rate is thus generally rising. This is an indicator of an ongoing structural change. All recession periods—2003, 2008, and the most recent recession—have led to a slowing of the pace at which the share of informal employment has been decreasing. 14 This underlines the importance of controlling for time trends in equation (1). 13 Here, formality is defined as dependent on whether a person is contributing to social security following Henley et al. (2009). 14 Using the alternative informality definition as used by the Brazilian Institute of Applied Economic Research (IPEA), classifying instead all those as informal, who do not have a legal employment booklet (Carteira Assinada), are self-employed, or are unpaid workers similarly shows a decreasing trend of the share of informality over the whole period and is in line with findings by IPEA (2015). Only starting from the beginning of 2014, we observe 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 2002_3 2002_6 2002_9 2002_12 2003_3 2003_6 2003_9 2003_12 2004_3 2004_6 2004_9 2004_12 2005_3 2005_6 2005_9 2005_12 2006_3 2006_6 2006_9 2006_12 2007_3 2007_6 2007_9 2007_12 2008_3 2008_6 2008_9 2008_12 2009_3 2009_6 2009_9 2009_12 2010_3 2010_6 2010_9 2010_12 2011_3 2011_6 2011_9 2011_12 2012_3 2012_6 2012_9 2012_12 2013_3 2013_6 2013_9 2013_12 2014_3 Employment Rate Informal Share of Informal workers Employment Rate Formal Working Paper | IDB-WP-00975 Table 2: Peak to Trough Changes March 2014 to Feb 2016 Recession Levels (by Gender, Race Education, Age, and Formality Status) Δ Unemployment Rate (in percentage points) Δ Employment Rate (in percentage points) Δ Participation Rate (in percentage points) All 3.1 -3.4 -1.7 Race-Gender White men 3.8 -3.6 -1.2 White women 3.3 -1.7 -0.1 Non-White men 2.6 -4.9 -3.2 Non-White women 2.5 -3.1 -2.0 Education Less than middle school graduate 3.4 -4.1 -3.2 Middle school grad/High school dropout/incomplete 4.4 -4.0 -2.0 High school grad/College dropout/incomplete 3.5 -4.8 -2.6 College grad/Master/PhD 1.6 -3.7 -2.4 Age Group 10-17 10.3 -2.0 -1.8 18-24 8.3 -7.4 -2.6 25-39 3.3 -3.5 -0.7 40-59 2.6 -1.5 0.4 60 and older 1.1 -2.1 -2.0 Formality Formal -1.8 Informal -1.6 Source: Authors' calculations based on seasonally adjusted (ArimaX11) PME surveys, March 2014 and February 2016 Note: Formality is based on the social security status definition following Henley et al. (2009) Nevertheless, a static analysis of differences in levels or changes in levels of various rates likely provides biased results, as discussed in Barnichon and Mesters (2017) and Hoynes et al. (2012). First, an analysis relying on levels cannot control for other characteristics. Second, it does not account for changes in the composition of distinct demographic groups, such as population growth of certain groups or rising labor force participation of women over time. Therefore, the next section provides estimates that control for both of these potential sources of bias. differences in the trend of the share of informality between the two measures, whereby the share of informality following the IPEA definition increases starting with the start of the recession in 2014 (see Appendix Figure A.1), whereas it only decreases to less than the trend when following our definition based on social security contributions. Working Paper | IDB-WP-00975 4.2 Regression Results In separate regressions, we estimate each major demographic group’s sensitivity of labor market measures to national and metropolitan unemployment rates, as explained in section 3. This allows us to analyze the sensitivity of unemployment, employment, and participation rate for each specific demographic group over the business cycle, for which the national and metropolitan area unemployment rates act as a proxy. Importantly, we control for demographic characteristics to account for any differences in the composition of demographic groups. The point estimates for 72 regressions estimated using national aggregated data and a panel for the main metropolitan areas of the country are presented in Figure 2. The first row of Figure 2 shows results using national level data (Figures a to c), and the second row shows the metropolitan level data (Figures d to f). Each point in the figure is the coefficient 𝛽1 of equation (1). For instance, Figures 2a and 2d show the results for the regressions capturing the sensitivity of group specific unemployment to the cyclical condition (expressed by the national unemployment rate in a and by the metropolitan area unemployment rate in d) for each age, education, and race/gender group. Controlling for metropolitan areas has the advantage of bringing more heterogeneity into the data and adds observations to each major demographic group’s regression. As we generally observe similar patterns at national and metropolitan area levels, this implies that the sensitivity is not affected by the fewer observations at the national level and smaller within group sample sizes at the metropolitan area level. However, the metropolitan area level coefficients have smaller confidence intervals. The observed smaller confidence intervals of the coefficients at the metropolitan area level and level differences of the coefficients as compared to the national level suggest the existence of inter-metropolitan area differences in the sensitivity of certain demographic groups’ responsiveness to business cycles, highlighting the importance of controlling for metropolitan areas in the regression to obtain more precise estimates. Therefore, results are presented at both levels. Working Paper | IDB-WP-00975 Figure 2: Plots of Coefficients using specification with year trend National Level: Impact of national unemployment (model specification with year trend) a) on group unemployment b) on group employment c) on group participation Metropolitan Level: Impact of metropolitan unemployment (model specification with metropolitan area specific year trend) d) on group unemployment e) on group employment f) on group participation Source: Authors' calculations based on seasonally adjusted (ArimaX11) PME surveys, March 2002 to February 2016 Note: The coefficients are displayed as dots, whereas the bars represent the 95% confidence intervals. -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 Working Paper | IDB-WP-00975 The first point estimate in Figure 2a indicates that a 1 percentage point increase in the unemployment rate leads to a 3 percentage point increase in the unemployment rate of the demographic group between 10 and 17 years old. The results for other age groups indicate that older workers are less sensitive to aggregate unemployment. This is in line with findings by Hoynes et al. (2012) in the US and Bachman et al. (2015) in Europe, who likewise find that younger age groups are harder hit by crisis. Next, results for demographic groups constructed based on education and race/gender, are presented. Less educated workers (less than middle school attainment) are not very sensitive to labor market conditions, a result that is in line with the raw data presented in Table 1. A possible explanation could be that firms can save more in terms of wages as presented in Table A1a when dismissing the slightly more educated rather than the lower paid least educated workers. Alternatively, one may argue that the category of the least educated are employed in basic but more essential positions. This would be in line with findings by Reis (2017) that the least educated face relatively shorter unemployment spells compared to the more educated. This is also in line with a report in The Economist (2016) that routine middle-educated jobs are most affected by automation over time, rather than unskilled, non-routine work such as cleaning. As also found for the US in Hoynes et al. (2012), the group that suffers more is made up of workers who have completed middle school or are high school dropouts; the responsiveness of more qualified workers is of lesser intensity. This is moreover in line with findings by Haltiwanger et al. (2017) in the US that younger and less than high school educated workers are more likely to be dismissed during economic downturns than older workers. 17 Interestingly, our results suggest that the unemployment rate of women is more responsive to labor market conditions than the unemployment rate of men for the white and non-white population. The 17 Hoynes et al. (2012) and Haltiwanger et al. (2017) do not further subdivide the less than high school educated into less than middle school educated workers. Working Paper | IDB-WP-00975 responsiveness of white and non-white males to unemployment is the same, indicating that labor market conditions do not appear to affect these groups differently. Despite the indication from Table 1 that non-white males have a higher level of unemployment, this does not seem to be related to higher sensitivity during business cycles (controlling for other demographic characteristics). Corresponding regressions using employment and participation rates as dependent variables can help us further investigate some results found so far. The unemployment rate of the demographic group between 10 and 17 years old is the most sensitive to changes in unemployment among all demographic groups. This pattern is observed at national (Figure 2a) and metropolitan level (Figure 2d). Nevertheless, Figures b and e show that the employment rate of the same demographic group is not as sensitive to the unemployment rate as expected. This might be explained by the increase in the participation rate of this demographic group when unemployment increases, as indicated in Figures 2c and 2f. This result suggests that youngsters are forced into the labor market to help their families. This result opposes findings from an earlier analysis on Brazil by Dureya and Arends-Kuenning (2003) that children are not more likely to work during economic crisis as their opportunity costs likewise fall. The joint analysis also helps us to better understand the similar sensitivity of white and non-white males to unemployment. At the national level, we show that the sensitivity of the white male and non-white male unemployment rate is the same; however, Figures 2b and 2c show that the mean sensitivity of the employment and participations rates of white males to business cycles is slightly lower than that of non-white males. This suggests that the sensitivity of the unemployment rate of non-white males would be higher if non-white males were not deciding to withdraw from active labor market participation at a higher rate than white males. The same also holds in an alternative specification of the model that controls for year and month fixed effects instead of a year trend (see Figure A.2). Using the metropolitan area Panel d in Figure 2, we likewise show that white male and non-white male sensitivity to the business cycle in terms of Working Paper | IDB-WP-00975 unemployment is similar. However, Figures e and f show that unlike the national level case, the employment and participation rates of white males are at the mean more sensitive to cycles than non-white males. This result also holds when considered apart from the year trend adding year-month fixed effects to the model to control, in a more detailed manner, for structural and compositional changes over time (see Figure A.2). While controlling for year-month fixed effects has the advantage of removing further bias from changes in the long run demographic structure, it also partly removes the macroeconomic cyclicality in which we are particularly interested. Besides, these alternative specification confidence intervals are much larger; thus, any potential difference in sensitivity is less precisely estimated. Moreover, including year-month fixed effects, these findings differ from the clear larger changes in the unemployment and participation rates for non-white males presented in Tables 2 and A.2 for the national and metropolitan area level respectively and highlight the importance of controlling for regional heterogeneity and demographic characteristics. A persistent wage gap in favor of men and white Brazilians throughout the income distribution, as estimated by Garcia et al. (2009), may be an explanation for why men and white workers are more likely to drop out of the labor market. Their larger potential wealth and savings from the higher wages earned can be used to bridge inactive periods during crises. Further support for a continued racial and gender wage gap during this sample period is also provided, as earlier mentioned in Table A.1. Particularly in Rio de Janeiro, white men’s monthly income is more than double the amount non-white women earn and almost double the earnings of non-white men. Comparing the sensitivity of white and non-white women’s unemployment rates to cycles in Figure 2 shows similar interesting deviations from the static findings of Tables 1. Figure 2a suggests that unemployment rate for white women is at the mean slightly more responsive to cycles when controlling for other characteristics, rather than that of non-white women, who are the ones who experience according to Table 1 higher levels of unemployment. This pattern Working Paper | IDB-WP-00975 persists for the metropolitan area model in Figure 2d, but the difference in sensitivity diminishes, leaving white women only very minimally more if not equally sensitive than non-white women to cycles. As in the case of men, we also consider a different specification controlling for year and month fixed effects at national and metropolitan area levels respectively, presented in Panels a and c of Figure A.2 in the Appendix. These panels show very similar sensitivity levels compared to those in the main specification. While not in line with the findings in terms of static levels, these findings are in line with the larger static change in the unemployment rate of white women between the peak and trough of the most recent crisis presented in Table 2. However, the sensitivity of the employment and participation rate to changes in the unemployment rate in Figures 2b and c and Figures 2e and f suggest the opposite from what the static percentage point change in those rates in Table 2 may suggest. Non-white women’s employment rates are, when controlling for other characteristics, less responsive to changes in the overall national or metropolitan area unemployment rate than that of white women. In addition, the participation rate of non-white women always tends to be less reduced or even increased when compared to white women. As in the case of white versus non-white men, also the lower sensitivity of non-white women’s participation rate may be the result of a greater need for non-white women to work to support the household, which can also explain non-white women’s generally higher labor market participation rate as presented in Table 1. A potential counter to the argument that lower total household incomes for non-white individuals results in fewer funds to bridge unemployment spells is to consider racial-intermarriage, whereby the income difference is more balanced at the household level. However, Telles (1993) and Ribeiro and Da Silva (2009) show that while racial intermarriage is becoming increasingly more common, representing one in three marriages by the year 2000, such marriages are largely of couples of similar skin color rather than between white and black individuals. Therefore, racial wage gaps may despite rising racial inter-marriage nevertheless still significantly affect total household income and savings. This finding may also Working Paper | IDB-WP-00975 provide support for the existence of an added worker effect for women, non-whites, and youngsters when other groups might be discouraged from participating in the labor force, as discussed for women in Lundberg (1985) for the case of the US and observed in a developing country context by Parker and Skoufias (2004) during the peso crisis in Mexico. Although Figures 2 and A.2 distinguish by race and gender, the responsiveness of the individual groups suggests that the gender component is more sensitive than the race component. In spite of lower gender than racial wage gaps estimated by Garcia et al. (2009), that may suggest differently, women’s unemployment rates are more sensitive to changes in the overall unemployment rate at the national level (Panel a) and at the metropolitan area level (Panel d). At the same time, women’s employment and participation rates appear less responsive to changes in the unemployment rate than that of men. The sensitivity of the unemployment rate for women seems therefore to a lesser extent, reduced through unemployed individuals becoming inactive, as in the case of men. The higher sensitivity of women’s unemployment rate is in line with Bahçe and Memiş’ (2014) findings for Turkey that women become more likely to be marginally attached to the labor market during economic downturns than men, but contradicts findings for Europe by Bredemeier and Winkler (2015) and Hoynes et al. (2012) for the US. Indeed, Bachman et al. (2015) find that men are more likely to become unemployed during recessions. However, the lower responsiveness of the participation rate of women than men during recessions in Brazil contradicts their findings that among the marginally attached, women in Turkey are more likely to move out of the labor force than men. Hence, there appears to be no internationally consistent pattern. Brazilian women’s greater responsiveness in terms of unemployment might be partly explained by the rapid increase of women participating in the labor market. Among other factors, this increase has been attributed to the wider availability of child care facilities (Holanda Barbosa and Melo Costa, 2017). Moreover, these authors’ findings that the probability of women’s participation in the labor market increases with the availability of daycare facilities Working Paper | IDB-WP-00975 Salardi P (2012) Chapter 5 in: An Analysis of Pay and Occupational Differences by Gender and Race in Brazil - 1987 to 2006. (submitted for Doctoral Dissertation, University of Sussex). Retrieved from: http://sro.sussex.ac.uk/45204/1/Salardi%2C_Paola.pdf Telles E (1993) Racial Distance and Region in Brazil: Intermarriage in Brazilian Urban Areas Latin American Research Review, 28(2): 141-162 The Economist (2016, June 25th) The impact on jobs Automation and anxiety-Will smarter machines cause mass unemployment? The Economist Retrieved from: https://wwweconomistcom/news/special-report/21700758-will-smarter-machinescause-mass-unemployment-automation-and-anxiety . Working Paper | IDB-WP-00975 7. Appendix Table A.1: Heterogeneous Labor Market Outcomes (by Region, Gender, Race Education, Age, and Formality Status in February 2016 Unemployment Rate (%) Employment rate (%) Usual Weekly Hours Monthly Income (in Feb. 2016 Reais) Category All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All 8 % 10 % 13 % 7 % 5 % 9 % 6 % 50 % 44 % 48 % 49 % 49 % 51 % 51 % 38 39 39 38 39 39 36 2087 1600 1609 2026 2365 2297 2256 Race-Gender White men 7 % 9 % 11 % 7 % 4 % 8 % 5 % 59 % 54 % 55 % 57 % 59 % 59 % 61 % 40 40 41 40 40 41 39 2896 2149 3060 3059 3378 3013 2615 White women 8 % 9 % 13 % 7 % 6 % 9 % 7 % 42 % 39 % 38 % 41 % 40 % 44 % 44 % 36 37 37 36 37 38 34 2222 1658 2113 2242 2705 2281 2061 Non-White men 8 % 11 % 11 % 7 % 5 % 9 % 8 % 58 % 52 % 57 % 57 % 59 % 62 % 58 % 40 40 40 40 41 41 37 1798 1543 1607 1874 2015 1835 1840 Non-White women 10 % 12 % 14 % 8 % 7 % 13 % 10 % 42 % 36 % 42 % 43 % 42 % 44 % 45 % 37 37 37 37 37 37 34 1396 1263 1288 1437 1518 1443 1372 Education Less than middle school graduate 7 % 8 % 9 % 5 % 5 % 8 % 6 % 28 % 27 % 30 % 30 % 27 % 28 % 30 % 38 39 38 38 39 39 38 1204 910 943 1224 1240 1389 1308 Middle school grad/High school dropout/incomplete 11 % 13 % 17 % 9 % 6 % 14 % 10 % 43 % 35 % 42 % 45 % 42 % 43 % 49 % 39 40 38 39 39 40 38 1309 1034 1008 1339 1283 1541 1358 High school grad/College dropout/incomplete 9 % 13 % 14 % 9 % 6 % 10 % 7 % 62 % 56 % 57 % 61 % 59 % 65 % 66 % 39 39 39 39 40 40 37 1707 1384 1389 1710 1802 1803 1987 College grad/Master/PhD 4 % 5 % 7 % 5 % 3 % 4 % 3 % 73 % 71 % 70 % 71 % 70 % 76 % 73 % 37 38 39 37 38 38 32 4382 3350 3702 4155 4935 4440 4964 Age Group 10-17 38 % 43 % 41 % 33 % 32 % 44 % 25 % 4 % 1 % 5 % 4 % 2 % 4 % 6 % 29 23 26 29 25 32 30 703 470 540 679 619 773 845 18-24 20 % 27 % 28 % 18 % 15 % 23 % 15 % 49 % 39 % 42 % 48 % 44 % 55 % 57 % 38 38 37 38 38 39 37 1189 1007 943 1170 1227 1333 1272 25-39 8 % 11 % 13 % 7 % 6 % 9 % 7 % 75 % 66 % 70 % 73 % 75 % 78 % 77 % 39 40 39 39 40 40 37 2010 1553 1544 1987 2260 2179 2205 40-59 5 % 6 % 7 % 4 % 3 % 6 % 3 % 68 % 61 % 66 % 66 % 69 % 69 % 68 % 39 39 39 39 39 39 37 2294 1742 1811 2264 2526 2529 2513 60 and older 2 % 0 % 6 % 2 % 1 % 2 % 3 % 18 % 18 % 16 % 17 % 21 % 17 % 18 % 37 37 37 36 37 39 35 2760 1893 1942 2453 3183 3198 2945 Formality Status Formal 38 % 32 % 33 % 40 % 37 % 41 % 42 % 39 40 41 39 40 40 37 2312 1840 1888 2179 2651 2482 2455 Informal 11 % 13 % 15 % 9 % 12 % 11 % 9 % 36 35 35 36 36 37 35 1329 999 987 1369 1514 1593 1366 Source: Authors' calculations based on PME survey February 2016 Note: Formality status is based on social security status definition Working Paper | IDB-WP-00975 Figure A.1: Informal Employment Rate and Share of Informal Workers over Time (based on alternative informality definition) Source: Author’s estimations based on PME surveys, March 2002 to February 2016 Note: As in IPEA (2015) Figure A.3 (January 2012=1). Informality is defined as those who do not have a legal employment booklet (Carteira Assinada), are self-employed, or are unpaid workers 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 2002_3 2002_5 2002_7 2002_9 2002_11 2003_1 2003_3 2003_5 2003_7 2003_9 2003_11 2004_1 2004_3 2004_5 2004_7 2004_9 2004_11 2005_1 2005_3 2005_5 2005_7 2005_9 2005_11 2006_1 2006_3 2006_5 2006_7 2006_9 2006_11 2007_1 2007_3 2007_5 2007_7 2007_9 2007_11 2008_1 2008_3 2008_5 2008_7 2008_9 2008_11 2009_1 2009_3 2009_5 2009_7 2009_9 2009_11 2010_1 2010_3 2010_5 2010_7 2010_9 2010_11 2011_1 2011_3 2011_5 2011_7 2011_9 2011_11 2012_1 2012_3 2012_5 2012_7 2012_9 2012_11 2013_1 2013_3 2013_5 2013_7 2013_9 2013_11 2014_1 2014_3 2014_5 2014_7 2014_9 2014_11 2015_1 2015_3 2015_5 2015_7 2015_9 2015_11 2016_1 Employment Rate Informal workers Share informal workers Working Paper | IDB-WP-00975 Table A.2: Peak to Trough Changes March 2014 to February 2016 - Recession Levels (by Gender, Race Education, Age, and Formality Status) by metropolitan area Δ Unemployment Rate Peak March 2014 to February 2016 Recession Level (in percentage points) Δ Employment Rate Peak March 2014 to February 2016 Recession Level (in percentage points) Δ Participation Rate Peak March 2014 to February 2016 Recession Level (in percentage points) Category All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All Recife Salvador Belo Horizonte Rio de Janeiro São Paulo Porto Alegre All 26 29 31 33 35 43 All 26 29 31 33 35 43 All 26 29 31 33 35 43 All 3.1 5.3 3.6 3.4 1.9 3.4 3.5 -3.4 -2.4 -3.0 -5.3 -2.5 -3.9 -2.6 -1.7 0.3 -1.0 -3.5 -1.6 -2.0 -0.5 Race-Gender White men 3.8 5.9 4.0 3.5 3.1 4.0 5.3 -3.6 -4.3 -2.3 -6.1 -3.6 -2.6 -2.0 -1.2 -0.9 0.5 -4.2 -2.3 0.1 1.0 White women 3.3 5.9 2.6 3.9 1.8 4.3 6.7 -1.7 -1.6 -1.5 -4.2 -0.4 -2.0 -3.8 -0.1 1.1 -0.7 -2.3 0.9 0.4 -0.7 Non-White men 2.6 5.7 6.6 3.8 1.6 2.6 3.0 -4.9 -5.7 -11.6 -6.1 -3.8 -5.4 -1.7 -3.2 -2.5 -7.8 -4.2 -3.1 -3.3 0.3 Non-White women 2.5 3.1 6.0 2.4 1.5 2.6 3.7 -3.1 1.1 -4.1 -4.6 -1.9 -3.8 -2.8 -2.0 2.6 -0.5 -3.7 -1.5 -2.9 -1.3 Education Less than middle school graduate 3.4 3.8 1.3 2.1 2.7 3.8 3.0 -4.1 -3.0 -3.0 -5.6 -3.2 -5.0 -2.3 -3.2 -1.9 -3.0 -4.8 -2.7 -3.8 -1.3 Middle school grad/High school dropout/incomplete 4.4 7.5 4.4 4.7 1.3 5.7 5.8 -4.0 -3.3 -1.2 -7.5 -2.5 -5.1 -4.6 -2.0 -0.5 0.8 -5.3 -2.1 -2.5 -1.5 High school grad/College dropout/incomplete 3.5 6.4 4.6 3.9 2.0 3.7 3.5 -4.8 -5.4 -4.0 -7.0 -4.8 -4.3 -4.2 -2.6 -1.7 -0.8 -4.3 -3.6 -1.9 -1.5 College grad/Master/PhD 1.6 2.1 1.7 2.2 1.7 1.0 2.2 -3.7 -0.9 -3.5 -4.7 -3.7 -2.7 -3.9 -2.4 0.8 -2.1 -2.9 -2.3 -2.1 -2.2 Age Group 10-17 10.3 14.6 4.2 13.5 11.4 9.9 3.3 -2.0 -1.4 -0.1 -2.9 -1.5 -2.3 -1.6 -1.8 -1.1 0.3 -2.5 -1.4 -2.3 -1.7 18-24 8.3 11.4 7.6 8.8 4.3 8.5 10.0 -7.4 -8.2 -5.2 -11.8 -4.6 -6.5 -7.8 -2.6 -2.4 -0.3 -7.1 -3.0 -0.5 -0.9 25-39 3.3 6.0 3.4 3.5 2.1 3.3 4.1 -3.5 -2.6 -2.8 -5.7 -3.2 -3.5 -3.0 -0.7 2.0 -0.3 -2.9 -1.3 -0.6 0.3 40-59 2.6 4.6 2.9 2.5 1.5 3.0 1.7 -1.5 -1.2 -1.5 -3.2 -1.7 -1.0 -1.1 0.4 1.7 0.9 -1.4 -0.7 1.7 0.2 60 and older 1.1 -0.8 5.0 1.5 0.9 0.8 2.4 -2.1 -2.2 -4.3 -3.2 0.3 -3.8 -0.2 -2.0 -2.3 -3.4 -3.0 0.2 -3.9 0.4 Formality Status Formal -1.8 -1.1 -2.7 -3.5 -0.8 -2.0 -1.2 Informal -1.6 -1.3 -0.1 -1.8 -1.6 -1.9 -1.3 Source: Authors’ calculations based on seasonally adjusted (ArimaX11) PME surveys March 2014 and February 2016 Note: Calculations are based on largest possible sample. Formality status is based on social security status definition Working Paper | IDB-WP-00975 Figure A.2: Plots of Coefficients National Level: Impact of national unemployment (model with year and month fixed effects) a) on group unemployment b) on group employment c) on group participation Metropolitan Level: Impact of metropolitan unemployment (model with year-month fixed effects and year trend) d) on group unemployment e) on group employment f) on group participation Source: Authors' calculations based on seasonally adjusted (ArimaX11) PME surveys March 2014 and February 2016. Note: These calculations are based on the largest possible sample. The coefficients are displayed as dots, whereas the bars represent the 95% confidence intervals. -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Working Paper | IDB-WP-00975 Figure A.3: Unemployment, Employment, and Participation Rates by Demographic Group over Time (seasonally adjusted rates) a) b) c) d) e) f) g) h) i) Source: Authors’ own estimations based on balanced sample raw data of PME surveys March 2002 till February 2016 0% 10% 20% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Unemployment Race non-white men_x11 Unemployment Race white men_x11 Unemployment Race non-white women_x11 Unemployment Race white women_x11 30% 40% 50% 60% 70% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Employment Race non-white men_x11 Employment Race white men_x11 Employment Race non-white women_x11 Employment Race white women_x11 30% 40% 50% 60% 70% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Participation Race non-white men_x11 Participation Race white men_x11 Participation Race non-white women_x11 Participation Race white women_x11 0% 5% 10% 15% 20% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Unemployment Rate College Graduate/Master/PhD_x11 Unemployment Rate High School/ College dropout_x11 Unemployment Rate Middle School less than High School_x11 Unemployment Rate Less than Middle School_x11 25% 45% 65% 85% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Employment Rate College Graduate/Master/PhD_x11 Employment Rate High School/ College dropout_x11 Employment Rate Middle School less than High School_x11 Employment Rate Less than Middle School_x11 25% 45% 65% 85% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Participation Rate College Graduate/Master/PhD_x11 Participation Rate High School/ College dropout_x11 Participation Rate Middle School less than High School_x11 Participation Rate Less than Middle School_x11 0% 10% 20% 30% 40% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Unemployment Rate age 40-59_x11 Unemployment Rate age 25-39_x11 Unemployment Rate age 18-24_x11 Unemployment Rate age 10-17_x11 Unemployment Rate age 60 and above_x11 0% 20% 40% 60% 80% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Employment Rate age 40-59_x11 Employment Rate age 25-39_x11 Employment Rate age 18-24_x11 Employment Rate age 10-17_x11 Employment Rate age 60 and above_x11 0% 20% 40% 60% 80% 2002_3 2002_10 2003_5 2003_12 2004_7 2005_2 2005_9 2006_4 2006_11 2007_6 2008_1 2008_8 2009_3 2009_10 2010_5 2010_12 2011_7 2012_2 2012_9 2013_4 2013_11 2014_6 2015_1 2015_8 Participation Rate age 40-59_x11 Participation Rate age 25-39_x11 Participation Rate age 18-24_x11 Participation Rate age 10-17_x11 Working Paper | IDB-WP-00975 Figure A.4: Plot of Coefficients: Metropolitan Level: Impact of metropolitan level unemployment on the informality share by race (model specification with metropolitan area specific time trend) Source: Authors' calculations based on seasonally adjusted (ArimaX11) PME surveys March 2014 and February 2016. Note: These calculations are based on the largest possible samples. The abbreviations ‘ipea’ and ‘soc.sec.’ indicate the estimation of models with the group informality share as the dependent variable, calculated either based on IPEA’s definition or based on the social security contribution status definition. The abbreviations ‘control educ’ and ‘control age’ indicate that these estimations only control for educational or age group respectively and not for the full array of demographic characteristics as in the models estimating the impact of metropolitan level unemployment on subgroup unemployment, employment, or participation rate. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 non-white (ipea, control educ) white (ipea, control educ) non-white (social sec., control educ) white (social sec., control educ) non-white (ipea, control age) white (ipea, control age) non-white (social sec., control age) white (social sec., control age)