Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Mauritius Addressing Inequality through More Equitable Labor Markets Mauritius Addressing Inequality through More Equitable Labor Markets This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Photography: www.shutterstock.com. Cover design: Circle Graphics, Inc. Design and Layout: Circle Graphics, Inc. Additional material relating to this report can be found on the World Bank Mauritius website (www. worldbank.org/mauritius). The material includes a fact sheet. © 2017 International Bank for Reconstruction and Development / International Development Association or The World Bank Group 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 www.worldbankgroup.org Contents Acknowledgments............................................................................................................. ix Abbreviations and Acronyms.............................................................................................. x EXECUTIVE SUMMARY........................................................................................................ 1 INTRODUCTION................................................................................................................. 17 CHAPTER 1  Inequality in Mauritius: Stylized Facts......................................................... 25 CHAPTER 2  Drivers of Growing Inequality in Household Labor Income........................ 37 Trends in Household Demographics....................................................................37 2.1  Trends in Labor Market Factors..........................................................................41 2.2  Explaining Changes in Equivalent Household Labor Income Inequality..............46 2.3  CHAPTER 3  The Role of Gender Inequality.................................................................... 53 Women’s Labor Market Participation..................................................................53 3.1  Working Women: What Do They Do?.................................................................58 3.2  Gender Wage Gap in the Public and Private Sector..............................................66 3.3  CHAPTER 4  Rising Inequality in Wages Among Individuals............................................ 75 4.1  Stylized Facts.......................................................................................................75 4.1.1  Trends In Overall Wage Inequality............................................................75 4.1.2 Inequality Between and Within Demographic Groups...............................79 4.2 Effects of Changes in Wages and Workforce Composition on Rising Inequality�������������������������������������������������������������������������������������������82 4.3 The Role of Labor Market Forces.......................................................................86 4.4 Changes in Workforce Composition....................................................................88 4.5 The Role of Shifts in the Relative Supply of Labor..............................................92 4.5.1  The Role of Foreign Labor........................................................................92 4.6 The Role of Relative Demand Shifts....................................................................96 4.7 The Role of Remuneration Orders....................................................................101 4.7.1 The Role of Remuneration Orders on Employment and Working Hours��������������������������������������������������������������������������������105 4.8 Skill Mismatch among the Employed and Rising Unemployment......................107 REFERENCES ................................................................................................................... 115 APPENDIX A.................................................................................................................... 119 APPENDIX B .................................................................................................................... 123 APPENDIX C.................................................................................................................... 125 APPENDIX D.................................................................................................................... 127 APPENDIX E .................................................................................................................... 131 ivContents List of Boxes Box I.1. Household Income: Choices, Markets, and Institutions..................................... 18 Box I.2. The Definition of Income..................................................................................... 19  o Household Surveys Underestimate Inequality? Box 1.1. D The Challenge of Top Incomes........................................................................... 34 Box 3.1. Day-Care Centers and Preprimary Schools........................................................ 58 Box 4.1. The Reweighting Approach................................................................................ 84 Box 4.2. Relative Labor Supply and Relative Labor Demand: A Simplified Framework............. 90 Box 4.3. Wage Setting in the Private Sector.................................................................. 102 nternational Standard Classification of Occupations: Definitions Box 4.4. I of Skill Levels.................................................................................................... 107 Box B.1. Data Overview and Definitions of Labor Market Variables............................. 123 List of Figures The Incomes of the Poorest Households Grew, but Less Rapidly Figure ES.1.  Than the Incomes at the Top, 2001–15.......................................................... 3 nequality in Household Income Widened, Particularly Figure ES.2. I in Labor Income, 2001–15.............................................................................. 3  arnings of Heads: The Main Contributor to Expanding Inequality Figure ES.3. E in Household Labor Income, 2001–15........................................................... 4  ourly Wage Inequality Widened Mostly in the Upper Tail, Figure ES.4. H Especially among Men, 2001–15.................................................................... 5  ourly Wages Grew More among the More Highly Educated Figure ES.5. H and Young Men, 2004–15............................................................................... 6  ervices, High-Skilled Occupations, and Employment among Figure ES.6. S the More Highly Educated Grew, 2004–15.................................................... 7  hanges in the Relative Supply of Workers, by Gender, Figure ES.7. C Education, and Experience, 2004–15............................................................. 8 ndex of Shifts in Relative between- and within-Industry Labor Figure ES.8. I Demand, by Gender and Education, 2004–15............................................... 9  emuneration Orders Modestly Increased Hourly Wage Figure ES.9. R Inequality, Particularly in the Lower Tail....................................................... 10 Labor Market Participation, by Gender and Educational Figure ES.10.  Attainment among Women, 2004–15........................................................ 11 Women Are Paid Unequally in the Private Sector, but Are Figure ES.11.  More Well Paid in the Public Sector, 2004–15........................................... 12 Figure ES.12. The Share of Overeducated Youth Is on the Rise, 2006–15...................... 13 Contents v Youth Unemployment Is High, and Unemployed Youth Figure ES.13.  Are Increasingly More Well Educated, 2006–15........................................ 14 Figure BI.1.1. Choices, Markets, and Institutions Affect Labor Income.......................... 18 Density of Log-Household Income with and without Housing Figure BI.2.1.  Rental Value, 2015...................................................................................... 20 Growth Incidence Curve, Household Consumption and Income, Figure 1.1.  2001–15.......................................................................................................... 26 Figure 1.2. Inequality in Mauritius and in the Rest of the World, 2000–12...................... 26 Trends of Selected Percentiles of Household Consumption Figure 1.3.  and Income, 2001–15..................................................................................... 27 Figure 1.4. Measures of Household Income Inequality, 2001–15.................................... 28 Figure 1.5. Gini Coefficient, by Income Source, 2001 and 2015...................................... 29 Total Household Income, Changes and Shares of Components, Figure 1.6.  2001–15.......................................................................................................... 30 Total Household Income, Changes and Shares of Components, Figure 1.7.  by Quintile, 2001–15...................................................................................... 31 Decomposition of Total Household Income Inequality, by Income Figure 1.8.  Source, 2001–15............................................................................................. 33 Figure B1.1.1. Top 1 Percent of the Fiscal Income Share, 1999–2011............................. 34 Figure 2.1. Household Size and Composition, 2001–15.................................................. 38 Double Earners and Dispersion of Household Labor Income, Figure 2.2.  by Family Type, 2001–15................................................................................ 40 Labor Force Status of Household Members of Married-Couple Figure 2.3.  Households, 2001–15..................................................................................... 42 Labor Force Status of Household Members of Single-Headed Figure 2.4.  Households, 2001–15..................................................................................... 44 Figure 2.5. Female Participation Rate, by Age-Group and Family Type, 2001–15.......... 45 Figure 2.6. Monthly and Hourly Labor Income, by Gender, 2001 and 2015................... 46 The Gender Gap in Hourly Earnings and Trends in Hours Worked, Figure 2.7.  by Gender and Education, 2001–15............................................................... 47 Figure 2.8. Trends in Individual Labor Income Inequality, by Gender, 2001–15.............. 47 Trends in Selected Percentiles of Household Labor Income Figure 2.9.  by Family Type, 2001–15................................................................................ 48 Step-Wise Decomposition of Household Labor Income, Figure 2.10.  2001 and 2015............................................................................................... 49 Labor Market Status of Spouses, by Quintile of Household Labor Figure 2.11.  Income, 2001–15.......................................................................................... 51 Labor Market Participation Rates, Mauritius and the Rest Figure 3.1.  of the World, 2004–15.................................................................................... 54 Figure 3.2. Labor Market Participation Rates, by Gender and Cohort, 2004–15............ 55 viContents Labor Market Participation Rates, by Gender and Age-Group, Figure 3.3.  2004–15.......................................................................................................... 56 Figure 3.4. Female Labor Force Participation Rates, by Marital Status, 2004–15........... 56 Female Labor Force Participation Rates, by Educational Attainment, Figure 3.5.  2004–15.......................................................................................................... 57 Counterfactual Participation Rate, by Gender, and Oaxaca-Blinder Figure 3.6.  Decomposition of the Gap, 2004–15............................................................. 59 Figure 3.7. Employment Ratio and Unemployment Rate, by Gender, 2004–15.............. 59 Employment Ratio and Unemployment Rate, by Gender Figure 3.8.  and Age-Group, 2004–15............................................................................... 60 Employment Category Distribution and Share of Wage Workers Figure 3.9.  in the Public Sector, by Gender, 2004–15...................................................... 61 Figure 3.10. Sectoral Distribution of Wage Workers, by Gender, 2004–15..................... 63 Occupational Distribution of Wage Workers, by Gender Figure 3.11.  and Main Sector, 2004–15............................................................................ 64 Educational Distribution of Wage Workers, by Gender Figure 3.12.  and Main Sector, 2004–15............................................................................ 67 Unconditional Gender Differentials in Hourly Wages, by Quantile and Figure 3.13.  Sector, 2004–15............................................................................................ 69 Oaxaca-Blinder Decomposition of the Gender Wage Differential Figure 3.14.  at the Mean, by Sector, 2004–15................................................................. 70 Oaxaca-Blinder Decomposition of the Gender Wage Differential Figure 3.15.  at Selected Percentiles, by Sector, 2004–15................................................ 71 Composition of Employment, by Category, and of Household Figure 4.1.  Labor Income, by Source, 2001–15................................................................ 76 Trends in Monthly Earnings, Weekly Hours Worked, Figure 4.2.  and Hourly Wages, Selected Percentiles, 2004–15........................................ 77 Figure 4.3. Hourly Wage Inequality, by Year, 2001–15..................................................... 78 Figure 4.4. Hourly Wage Inequality, by Year and Gender, 2001–15................................ 79 Education, Experience, and Gender Hourly Wage Differentials, Figure 4.5.  2004–15.......................................................................................................... 80 Changes in Relative Hourly Wages, by Gender and Education, Figure 4.6.  2004–15.......................................................................................................... 80 Changes in Real Hourly Wages, by Gender and Experience, Figure 4.7.  2004–15.......................................................................................................... 81 Figure 4.8. Overall and Residual Wage Inequality, by Year, 2004–15.............................. 83 Figure 4.9. Actual and Counterfactual Wage Inequality, 2004–15................................... 84 Figure 4.10. Actual and Counterfactual Wage Inequality, by Gender, 2004–15.............. 85 Figure 4.11. Actual and Counterfactual Residual Wage Inequality, 2004–15.................. 86 Distribution of the Employed Population, by Industry, Occupation, Figure 4.12.  and Year, 2001–15......................................................................................... 87 Contents vii Distribution of Total (Ages 16+) and Employed Population, Figure 4.13.  by Educational Level and Year, 2001–15...................................................... 89 Figure B4.2.1. The Relative Supply and Relative Demand Model.................................... 90 Changes in the Relative Supply of Workers, by Gender, Education, Figure 4.14.  and Experience, 2004–15............................................................................. 91 Figure 4.15. Price Versus Quantity Changes, All Workers, by Period, 2004–15.............. 93 Figure 4.16. Price Versus Quantity Changes, by Gender, 2004–15................................. 93 Figure 4.17. Foreign Workers, Overall and by Sector, 2004–15...................................... 94 Foreign Workers, by Gender, Sector, and Country of Origin, Figure 4.18.  2004– or 2005–15......................................................................................... 95 Sectoral and Occupational Distribution of Employment, Figure 4.19.  by Demographic Group, Average, 2004–15................................................ 97 Sectoral and Occupational Distribution of Total Employment, Figure 4.20.  by Period, 2004–15 and Period Average..................................................... 98 The Between, Within, and Overall Labor Demand Shift Index, Figure 4.21.  by Demographic Group and Period, 2004–15............................................. 99 Framework for Establishing a Minimum Wages in the Figure B4.3.1.  Private Sector.......................................................................................... 102 Figure 4.22. Changes in the Real Hourly Minimum Wage, by RO, 2004–14.................. 103 Figure 4.23. Changes in Average Real Earned Hourly Wages, by RO, 2004–14........... 103 Estimates of the Effect of Remuneration Orders on Hourly Wage Figure 4.24.  Inequality, 2004–15..................................................................................... 104 Figure 4.25. Trends in Education Mismatch, 2006–15.................................................... 108 Trends in the Education Mismatch, by Gender and Age-Group, Figure 4.26.  2006–15...................................................................................................... 109 Trends in Unemployment Rates, by Age-Group and Gender, 2006–15......... 109 Figure 4.27.  Distribution of Unemployed Youth (Ages 16–29), by Education Figure 4.28.  and Quintile of Household Income, 2006–15............................................ 111 Growth Incidence Curve, Total Household Income: Figure B.1.  Comparing HBS and CMPHS Data, 2007 and 2012.................................... 124  emographic and Labor Market Factors and Changes in Household Figure C.1. D Labor Income Inequality, 2001–15............................................................... 125 Sectorial Distribution of Wage Workers, by Gender Figure D.1.  and Main Sector, 2004–15........................................................................... 127 Figure D.2. Educational Distribution of all Wage Workers, 2004–15............................. 129 Earned Mean Wages and Legislated Mean RO Wages in Figure E.1.  Covered Sectors, 2004–14........................................................................... 131 viiiContents List of Tables Table I.1. List of Methodologies Adopted in the Analysis............................................... 21 Assortative Mating: Double-Earner Couples, by the Husband and Wife’s Table 2.1.  Labor Income Decile, 2001 and 2015.............................................................. 40 Table 4.1. Estimates of the Effects of Minimum Wages on Employment...................... 105 Table 4.2. Estimates of the Effects of Minimum Wages on Hours Worked................... 106 Table E.1. Number of Wage Rates Specified within Remuneration Orders, 2016......... 132 Estimation Approach to Determining RO Worker Coverage Table E.2.  Using CMPHS Data........................................................................................ 133 Table E.3. The Real Hourly Minimum Wage, by Remuneration Order, 2004–14........... 135  i Acknowledgments This report has been prepared by a core team comprising Pierella Paci (Task Team Leader and Practice Manager, GPV01, at the Decision Meeting stage), Marco Ranzani (co-TTL, Economist, GPV01), and Giuseppe Grasso (Consultant, GPV01), with contributions from Professor Haroon Bhorat (University of Cape Town), Zaakhir Asmal (University of Cape Town), and Associate Professor Verena Tandrayen- Ragoobur (University of Mauritius). The report has benefited from initial discussions held in Mauritius with government officials and rep- resentatives of the private sector and of academia. Staff at Statistics Mauritius were instrumental in the realization of the report. The team would like to thank particularly Mrs. Yasmin Cassimally (Director), Mrs. Set Fong Cheung Tung Shing (Deputy Director), and Mrs. Naseem Ramjan (Statistician, CMPHS Unit) for their excellent support and collaboration. The team would also like to thank Jamele Rigolini (Lead Economist, GSP01) and Toby Linden (Lead Education Specialist, GED01) for insightful comments. The team gratefully acknowledges the support of Pablo Fajnzylber (Practice Manager, GPV01, at the Concept Note review stage), Mark Lundell (Country Director, AFCS2), Carolin Geginat (Program Leader, AFCS2), Alex Sienaert (Senior Economist, GMF13), Mariella Beugue (Program Assistant, AFMMU), Siele Shifferaw Ketema (Program Assistant, GPV07), and Martin Buchara (Program Assistant, GPV01). The peer reviewers of the report are Eliana Carranza (Senior Economist, GPSJB), Carolina Mejia-Mantilla (Economist, GPV01), and Victor Sulla (Senior Economist, GPV07).  Abbreviations and Acronyms CMPHS Continuous Multi-Purpose Household Survey GDP gross domestic product ICT information and communication technology NRB National Remuneration Board OECD Organisation for Economic Co-operation and Development RO remuneration order  1 EXECUTIVE SUMMARY M auritius is often cited as one of the few African The period between 2001 and 2015 was characterized success stories, and with good reason. In the by substantial economic growth and yet limited shared aftermath of independence (1968), this small prosperity and increasing inequality. The Gini coefficient island nation in the Indian Ocean seemed to be bound for of household income increased from 0.37 in 2001 to economic failure because of its high poverty rate and numer- 0.42 in 2015 (+16.4 percent). The income ratio of the ous vulnerabilities, including high population growth, ethnic 90th percentile of the distribution to the 10th percentile tensions, substantial unemployment, and an economy greatly (P90/P10), which measures the distance between the upper- dependent on the production of sugar for international bound income value of households in the ninth decile (the markets. However, Mauritius was success­ ful in diversifying richest 10 percent of households) and the lower bound the economy and accomplishing an unprecedented struc- income value of households in the first decile (the poorest tural transformation. This made steady economic growth 10 percent of households) expanded by about 37 percent. possible, and this has significantly reduced poverty and This increase calls attention to the tails of the household placed the country solidly among the richest in the Africa income distribution, particularly at the bottom. Among the region. various sources of household income, income from labor is by far the largest component—representing an average of However, about a decade ago, this economic model about 80 percent of household income in 2015—and was encountered initial challenges. This was the by-product the source of income that most contributed to the observed of the loss of preferential access of the country’s sugar increase in income inequality. Inequality in household and textile production to the European Union and U.S. labor income is affected by two main groups of factors. markets and growing international competition for the First are demographics, including household composition, country’s low-cost industries. The government reacted household mix, household characteristics, and the degree to promptly by implementing a series of liberal economic which individuals marry within their own income group. reforms that temporarily brought the Mauritian economy Second are labor market factors, including labor force back on track. However, economic growth began to slow participation and inequality in individual labor income. again in 2010. While these factors played a role, the single most important contributor was certainly the expansion in the inequality in At the same time, inequality increased, threatening the individual earnings. The considerable growth in inequality standards of living of the poor. The Inclusiveness of Growth in individual earnings, largely wages, is attributable to the and Shared Prosperity report (World Bank 2015a) turned structural changes that have occurred in Mauritius over the spotlight on the expanding gap of inequality in house- the last 15  years. The economic structure continued a hold incomes that occurred between 2007 and 2012 and progressive shift away from traditional sectors, including on the negative impact on poverty. The report estimates agriculture and manufacturing, particularly textiles, toward that the incidence of absolute poverty between 2007 and services, notably, professional and financial services. This 2012 would have declined twice as quickly had growth economic transformation generated a substantial increase been shared more widely and inequality not worsened. in the demand for skilled workers that was not matched Building on these earlier findings, this study investigates by an equally rapid increase in the supply of skilled work- the driving forces behind the growing income inequality ers, notwithstanding significant improvements in edu- and identifies policy levers that could mitigate and, in the cational attainment among the Mauritian population. long run, possibly reverse the upward trend. In addition to market forces, labor market institutions, more precisely, the system of remuneration orders (a set This study takes a comprehensive approach to the deter- of legislated minimum wages, ROs), contributed to the minants of inequality by including the role of the choices rise in wage inequality at the bottom of the distribution of households and individuals, markets, and institutions. because of small and sporadic adjustments. Though the 2 Mauritius: Addressing Inequality through More Equitable Labor Markets rise in women’s participation in the labor market did not The increase in total income inequality was mainly turn out to be a large contributor in explaining changes attributable to inequality in household labor income. The in household income inequality, gender equity is certainly other components, including income from property and an important policy area that merits attention. Mauritian public and private transfers, played a relatively minor women are substantially disadvantaged in access and role. The Gini coefficient calculated on household labor remuneration in the labor market, particularly in the private income rose from 0.41 to 0.50 between 2001 and 2015. sector. This appears to be the by-product of two main fac- The government was successful in curbing the sharp tors. First, women employed in the private sector have less upward trend in inequality through redistribution poli- productive characteristics compared with men; second, the cies targeted particularly on households at the bottom pay structure seems to favor men over women. Meanwhile, of the distribution (see figure ES.2). The cornerstone of the public sector seems to be an attractive avenue for highly the redistribution effort was the social protection system. skilled women who are, on average, paid more than the This contributed considerably to reducing poverty and corresponding men, thanks to their productive character- contain inequality. However, despite important steps istics and to the milder form of unequal treatment. Finally, taken by the government toward greater coordination besides the existence of a skills shortage, which has been in social protection programs, further improvements identified as the main culprit in rising earnings inequal- are needed (World Bank 2015a). For example, the basic ity, the low quality of learning achievements is likely to retirement pension, a universal noncontributory social have generated a growing number of unemployed youth, pension paid to persons above age 60, lacks a focus who are becoming more highly educated, as well as to an on the poor because it disproportionately favors well- expanding educational mismatch among employed youth. off households, and funding is low for programs In addition to contributing to widening wage inequality, specifically intended to benefit the poor (World Bank this kind of mismatch can have negative consequences at 2015a).2 both the micro- and the macrolevel. First, it can impair the productive potential of youth and therefore influence lifetime patterns of employment and pay; second, it might hinder economic growth, productivity, and competitiveness in the Rising Inequality in Wages long term. Is the Main Contributor to Growing Inequality Household Labor Income Is in Household Incomes Driving the Rise in Inequality The typical Mauritian household is a married-couple The years between 2001 and 2015 were characterized by household composed of a husband and wife. The share of substantial economic growth, but limited shared prosperity. this type of household declined, particularly the share of Total per adult equivalent household income was growing married-couple households with children, which declined at an average annual rate of about 3 percent. However, from 61 percent in 2001 to 46 percent in 2015. By con- households initially in the lowest five percentiles of the trast, single-headed households are more common today; distribution recorded an average annual income gain of less in 2015, they made up about 20 percent of all households. than 1 percent, whereas households in the top 10 percent This is ascribable to delayed family formation decisions. posted an average annual income gain of about 3.6 percent The aging of the Mauritian population is also important. (figure ES.1).1 Such a growth pattern is clearly reflected in Single-headed households are typically headed by women the trends in inequality (figure ES.2). Household income (75 percent) in their late 50s or early 60s and without inequality expanded rapidly, notably, in the second half coresident children. of the period, between 2008 and 2015, in the aftermath of the global economic downturn and the shocks to the Changes in household types matter for the dynamics of terms of trade of Mauritius. The Gini coefficient increased inequality for at least two reasons. First, the family plays from 0.37 in 2001 to 0.42 in 2015 (+16.4 percent), while a role in providing insurance against individual risk as the P90/P10 expanded by about 37 percent, calling atten- shown by the wider inequality among single-headed house- tion to considerable changes at the tails of the household holds relative to married-couple households. Therefore, income distribution, particularly the lower tail. an increase in the share of single-headed households alone FIGURE ES.1. The Incomes of the Poorest Households Grew, but Less Rapidly Than the Incomes at the Top, 2001–15 annual growth rate of household income by percentile Growth incidence curve: 2001–2015 5 4 Annual growth rate (percent) 3 2 1 0 0 20 40 60 80 100 Percentiles Household total income (per adult equivalent) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The red line illustrates the average annual growth rate of household income. FIGURE ES.2. Inequality in Household Income Widened, Particularly in Labor Income, 2001–15 trends in the P90/P10 of different income aggregates 22 20 18 16 14 Ratio 12 10 8 6 4 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Labor + property income Labor + property + private transfer income Labor + property + total transfer income Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: Each line shows the trends in the P90/P10 of an income aggregate. 4 Mauritius: Addressing Inequality through More Equitable Labor Markets contributes to widening inequality. Second, rising female labor force participation translates into a growing share Skills Shortages Are the of married-couple households with a minimum of two Principal Culprit in the workers. This can lead to greater dispersion to the extent that earnings are strongly correlated across spouses. Widening Inequality in Individual Wages Earnings from the wage employment of household heads and spouses were the main contributor to growing inequality in Inequality in individual hourly wages expanded rapidly in the household income. The earnings of heads and spouses are second half of the period, between 2008 and 2015. Hourly the main source of household labor income. In 2015, the wage inequality, as measured by the P90/P10 ratio, increased earnings of spouses contributed 100 percent of household by almost 30 percent.3 The growth was concentrated in the labor income in 40 percent of Mauritian households and upper tail as illustrated by the trend in the P90/P50 ratio, less than 100 percent in 37 percent of the households, while, bution, the rise was limited while, at the bottom of the distri­ in 23 percent of the households, heads and spouses did not (+4 percent) (figure ES.4, panels a and b). While the rise work. Changes in household labor income inequality are in wage inequality was wider among men, inequality was a by-product of a combination of changes in household still greater among women (figure ES.4, panels c and d). demographics and the labor market attributes of house- hold members. The growing correlation of the earnings of Growing wage inequality is largely attributable to expand- husbands and wives in married-couple households, joined ing inequality between groups defined by educational with a larger increase in female labor force participation attainment. The upward trend in wage inequality can among the most affluent households, exerted upward pres- be attributed to changes in inequality across groups and sure on inequality in household labor incomes. However, inequality within groups of workers. Groups of workers are the growing inequality in individual earnings was the main defined by demographic characteristics, including gender, factor behind the rising inequality in household labor education, and age. The high to low educational attain- incomes (figure ES.3). ment ratio in the hourly wage rose by about 18 percent FIGURE ES.3. Earnings of Heads: The Main Contributor to Expanding Inequality in Household Labor Income, 2001–15 the contribution of demographic and labor market factors to changes in household labor income inequality (P90/P10) 1.0 –28.3 113.9 6.8 6.0 P90/P10 0.3 0.2 –40 –20 0 20 40 60 80 100 120 140 Percent Men’s labor income inequality Assortative mating Family characteristics Women’s participation Family mix Residual Women’s labor income inequality Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Executive Summary 5 FIGURE ES.4. Hourly Wage Inequality Widened Mostly in the Upper Tail, Especially among Men, 2001–15 a. Trends in the P50/P10 ratio b. Trends in the P90/P50 ratio 1.4 1.4 1.3 1.3 1.2 1.2 Ratio Ratio 1.1 1.1 1 1 .9 .9 .8 .8 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 c. Men: trends in the P90/P50 ratio d. Women: trends in the P90/P50 ratio 1.4 1.4 1.3 1.3 1.2 1.2 1.1 1.1 Ratio Ratio 1 1 .9 .9 .8 .8 .7 .7 .6 .6 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. among men and by less than 1 percent among women.4 In each demographic group. First, the expanding premium 2015, men with upper-secondary or higher education made for more highly educated workers is attributable to the about 87 percent more per hour worked than men with up larger increase in hourly wages relative to workers with to completed primary education. The premium was about low educational attainment, particularly among men 56 percent in 2004. The hourly wage premium associated between 2007 and 2011 (figure ES.5, panels a and b). with experience fell appreciably among both men and Second, the decline in the experience premium was driven women. For example, in 2004, an average man with 35 plus by the larger rise in hourly wages among young workers years of experience was paid about 54 percent more per relative to their older counterparts mostly between 2007 hour worked than a man with 14 years or less experience, and 2010 (figure ES.5, panels c and d). Third, the reduc- but the premium had dropped to 27 percent by 2015.5 tion in the gender gap can be attributed to the more rapid growth of hourly wages among women relative to men.6 Workers with high educational attainment posted large wage increases relative to low educated ones. The trends The shifts observed in relative hourly wages among described above are the result of the wage dynamics of workers with high or low educational attainment are 6 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE ES.5. Hourly Wages Grew More among the More Highly Educated and Young Men, 2004–15 changes in real hourly wages a. By educational attainment, women b. By educational attainment, men 35 32.0 35 31.4 30 29.3 30.5 30 25 23.5 24.0 25 22.4 21.5 20.0 20 16.8 20 18.3 16.0 16.1 Percent Percent 15 15 10 9.1 8.9 8.17.5 10.0 10 5 1.8 1.7 5 1.7 0 0 –0.7 –5 –5 –3.0 –1.8 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 Up to complete primary Up to complete primary Lower secondary Lower secondary Upper secondary/post-secondary/tertiary Upper secondary/post-secondary/tertiary c. By experience level, women d. By experience level, men 50 50 44.7 40 40 31.3 30 28.7 26.5 30 Percent 23.7 22.9 24.6 24.7 22.9 Percent 21.2 19.4 20 20 17.9 14.1 12.6 10.6 10 10 6.4 7.7 4.2 4.0 1.7 0 0 –2.3 –1.1 –1.2 –2.4 –10 –10 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 0–14 15–34 35+ 0–14 15–34 35+ Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. ascribable to structural changes. The population of change. The share of agricultural employment declined by Mauritius has become increasingly more well educated. 2 percentage points, to 6.1 percent, between 2004 and 2015. In 2001, less than 6 percent of Mauritians ages 16 plus had The share of manufacturing fell by 9 points, to 16.2 percent some postsecondary or tertiary education, but the share has in 2015. Meanwhile, the relative weight of the services sec- risen fourfold over the last 15 years. The educational level of tor expanded (see figure ES.6, panel b). Trade, hotels and the employed population also increased considerably. Between restaurants, and transport grew by 1.4, 1.8, and 1.1 points, 2004 and 2015, the share of workers with postsecondary respectively. The expansion in financial, real estate, and or tertiary education rose by about 18 percentage points, professional services was even larger (4.8 percentage while the share with completed primary education or less points). A similar transformation occurred within indus- dropped by 14 percentage points (figure ES.6, panel a). tries. A shift occurred toward high-skill occupations, such as managers and professionals, whose share rose by almost In parallel, the Mauritian economy has continued a trans- 7 percentage points to reach 23 percent in 2015 (see fig- formation away from traditional and largely low-skilled ure ES.6, panel c). Craft workers, skilled agricultural workers, sectors, possibly in line with changes in the structure of machine operators, and workers in elementary occupation product demand, increased international competition, lost importance, recording a reduction of about 13 percent- changes in trade agreements, and skill-biased technological age points, to 43.5 percent in 2015. Executive Summary 7 FIGURE ES.6. Services, High-Skilled Occupations, and Employment among the More Highly Educated Grew, 2004–15 change in the distribution of the employed population a. By educational attainment b. By sector of employment Up to Agriculture/mining −2.0 complete −14.0 Manufacturing −8.9 primary Construction −2.2 Lower −0.4 Trade 1.4 secondary Hotels/restaurants 1.8 Upper −3.5 Transport/utilities 1.1 secondary Professional activities 4.8 Post-secondary/ Public administration 0.8 17.8 tertiary Other services 3.2 −15 −10 −5 0 5 10 15 20 −12 −10 −8 −6 −4 −2 0 2 4 6 8 Percentage points change Percentage points change c. By occupation Professional, technical 6.9 and managers Clerical, sales and service 5.8 workers Craft, production and elementary −12.7 occupations −12 −10 −8 −6 −4 −2 0 2 4 6 8 Percentage points change Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. The relative supply of workers with high educational Between 2004 and 2015, there was an increase in the rela- attainment grew rapidly. The labor supply of more highly tive demand for more well educated workers, particularly educated women rose massively (+79 percent), particularly women and a decline in the relative demand for workers between 2007 and 2010 (figure ES.7, panel a). While the with low educational attainment (figure ES.8). Such demand relative supply of low- and mid-educated workers, both men shifts are largely attributable to shifts between industries and women, declined considerably (figure ES.7, panel b). and occurred, notably, between 2012 and 2015. These These trends may account for the rise in hourly wages changes may have an effect on the allocation of total labor among less well educated workers given that their relative demand across industries and can be ascribed to changes labor supply declined. Yet, changes in labor supply do not in product demand across industries or changes in the net account for the large growth in hourly wages among more international trade affecting the domestic share of output, highly educated women because this was accompanied by a such as the loss of preferential access of Mauritian sugar parallel and similarly large expansion in the relative labor and textile production to the American and European supply of these women. markets. Within-industry shifts induced a general decline in the demand for less well educated workers, both men and The rapid expansion in the relative demand for highly edu- women, and an increase in the demand for more highly cated labor outpaced the expansion in the relative supply. educated men. However, the role of within-industry changes 8 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE ES.7. Changes in the Relative Supply of Workers, by Gender, Education, and Experience, 2004–15 a. By education, women b. By education, men 90 78.9 90 70 58.1 70 50 50 28.5 30 30 Percent 14.2 Percent 10.4 10 2.4 1.6 10 1.6 0.7 0.2 2.5 –10 –0.5 –10 –5.4 –3.6 –12.6 –16.0 –15.0 –30 –22.5 –21.3 –30 –29.1 –29.1 –27.1 –50 –39.6 –50 –48.0 –70 –70 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 Up to complete primary Up to complete primary Lower secondary Lower secondary Upper secondary/post-secondary/tertiary Upper secondary/post-secondary/tertiary c. By experience, women d. By experience, men 120 120 103.8 100 100 80 80 63.1 60 49.6 60 Percent 44.2 Percent 40 28.6 40 34.7 21.1 21.6 20 8.0 5.7 11.3 14.3 20 4.2 10.4 9.6 11.3 11.4 3.0 0 0 –4.2 –3.7 –6.7 –20 –20 –13.5 –15.2 –40 –25.4 –40 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 0–14 15–34 35+ 0–14 15–34 35+ Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The numbers shown in the four panels represent percentage changes in the share of each group in total labor supply, measured in efficiency units. The latter are obtained by multiplying annual hours by the relative wage of the group, averaged over 2004–15. in labor demand was secondary. These changes affect the education and training to make sure they address the chang- relative intensity of the use of production inputs within ing needs in skills. A comprehensive strategy to reduce the industries and across occupations, and they are typically skills shortage requires that the quality of public education attributable to nonneutral technological change, changes be secured and demands an approach to providing education in the prices of nonlabor inputs, and outsourcing. that acknowledges the labor market relevance of medium skills (acquired through technical and vocational educa- The rise in wage inequality calls for structural responses. The tion) and high skills (acquired through tertiary education). increase in wage inequality was the by-product of structural Guaranteeing the relevance of education and training for adjustments of the Mauritian economy. It thus requires the labor market means there must be effective channels of long-term adjustments. Policies targeted at closing the skills communication between education and workplace actors, shortage have the potential to reduce wage inequality and as well as public-private partnerships. are also beneficial for productivity and economic growth. Key are investments in skills that are in high demand. This In the short term, fostering return migration might provide requires an accurate assessment of the current and future some relief. Mauritius has been historically characterized by needs of the country in skills, followed by adjustments in significant emigration. About 96,000 Mauritians ages 15 Executive Summary 9 FIGURE ES.8. Index of Shifts in Relative between- and within-Industry Labor Demand, by Gender and Education, 2004–15 a. Women b. Men 15 15 10 10 Percent Percent 5 5 0 0 −5 −5 ≤ Complete Lower ≥ Upper ≤ Complete Lower ≥ Upper primary secondary secondary primary secondary secondary Overall Between Within Overall Between Within Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. or above were residing abroad in 2000 (IOM 2014). Large RO-covered sectors. The ratio between legislated RO wages Mauritian diasporas have been established in Australia, and mean wages earned by covered workers declined Canada, France, Italy, South Africa, and the United Kingdom. because of two factors: intermittent adjustments of legislated In addition, every year, an increasing number of Mauritian RO wages and the shifts documented above in employment students go abroad for educational purposes: about 11,000 in in services and in high-skilled occupations. Thus, low-skilled 2015, according to Statistics Mauritius. While more evidence workers employed in traditional RO sectors recorded mod- is needed on the dimension, pattern, and characteristics est wage gains relative to high-skilled workers employed in of the Mauritian diaspora, it seems that providing incen- services that benefited from large wage gains thanks to the tives to Mauritian abroad to return to the island and simul- rampant skills shortage. For these reasons, ROs contributed, taneously incentivizing firms that operate in Mauritius to although modestly, to increasing inequality, particularly in hire returning migrants might contribute to alleviating the lower tail and up to the 30th percentile of the hourly the skills shortage. wage distribution (figure ES.9).7 The inequality-increasing effect was larger among men, especially at the bottom of the distribution (below the 20th percentile). The estimated Remuneration Orders Raised effect is in line with other studies that find that a decline in the real value of the minimum wage has been responsible Wage Inequality at the for part of the rise in inequality in Mexico and the United Bottom and Had Negative States as well as in the United Kingdom, where the falling (industry-based) minimum wage contributed to rising Employment Effects inequality in the late 1980s and early 1990s. Remuneration orders (ROs) contributed modestly to increas- ROs are estimated to have had a modest negative employ- ing inequality, particularly at the bottom of the distribution. ment effect in the covered sectors. An increase of 10 per- In the Mauritian context, a key role is played by ROs. cent in RO-legislated wages is associated with an overall Between 2004 and 2015, changes in legislated RO wages decline in employment of 0.57 percent. The effect differs were modest. The (lowest) legislated wages in about a third of in magnitude by gender. Employment is estimated to have the covered sectors declined in real terms between 2004 and declined by 0.77 percent among men and by 1.06 percent 2014, while the vast majority recorded a modest increase. among women. The larger effect among women may The change in legislated RO wages lagged behind the growth derive from the especially large rise in the legislated RO observed in actual wages earned by workers employed in wage of domestic workers, a sector in which employment 10 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE ES.9. Remuneration Orders Modestly Increased Hourly Wage Inequality, Particularly in the Lower Tail a. Effect of ROs on inequality, women b. Effect of ROs on inequality, men 1.5 1.5 1 1 Marginal effect Marginal effect .5 .5 0 0 −.5 −.5 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Percentile of log hourly wage Percentile of log hourly wage Con dence intervals in dotted lines Con dence intervals in dotted lines Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: Inequality is measured as the distance between each percentile and the 70th percentile and is a function of the gap between the legislated RO wage and the 70th percentile. The 70th percentile has been chosen as a level of earnings unaffected by legislated RO wages. A positive marginal effect indicates that an increase in legislated RO wages is associated with an increase in wage inequality in the covered sectors. among women dominates. Such estimates are within the influencing the gaps in compliance in the developing world range of a number of elasticities estimated for low-and (Bhorat, Kanbur, and Mayet 2013). First are institutional middle-income countries, ranging from -1.3 to 1.0 percent. factors such as the penalty structure for noncompliance, In terms of working hours, the estimated effect was an the complexity of the wage schedule, and the resources increase by 2.3 percent in the number of hours worked allocated to enforcement services. Second, the individual by men and a decline by 1.8 percent among women in the characteristics of inspectors, including their educational covered sectors. attainment, can influence the extent to which they are effective at achieving compliance. Third, firm character- Minimum wage policies are not the most appropriate instru- istics, such as size, distance from the enforcement agency, ment to help poor and low-income families. The main the number of previous violations, and the level of foreign argument typically offered in favor of a minimum wage ownership, will influence violations and enforcement. is that it helps poor and low-income families. A minimum Fourth, local labor market characteristics, such as the wage often brings about negative employment effects and unemployment rate, the average wage rate relative to the therefore creates winners and losers. Moreover, the policy minimum wage, and unionization, also play a role. In addi- target is frequently wrong, that is, low-wage workers instead tion, the economic environment and the implementation of of low-income families (if the two groups do not overlap). collaborative social policies that coincide with minimum Many low-income families have no workers. This is the wage policies can affect compliance, enforcement, and the case in Mauritius, where poor families are less likely to have overall economic impact of a minimum wage. working household members. In 2012, about 73 percent of the poor were unemployed or economically inactive (World A recent study has explored issues of minimum wage Bank 2015a). coverage and gaps in minimum wage compliance in 11 low- and middle-income countries (DPRU and CSDA Simple and enforceable minimum wage policies set at 2016).8 The study shows that simple national minimum meaningful levels are key to protecting low-wage workers. wage systems are typically associated with higher compli- Among the key decision areas on a minimum wage system ance rates. The foreseen introduction of a national minimum is the level of the minimum, but also the complexity of the wage might help simplify the institutional context, increase wage regime and the intensity of enforcement. Four sets compliance, and protect low-paid workers if the wage is of variables are important in understanding the factors set meaningfully. Executive Summary 11 Gender Inequality Is Declining, explained by differences in observable characteristics, including age, educational level, marital status, and but There Is Still a Long household demographic structure, including the presence Way to Go of children or older family members. However, a con- siderable portion of the observed variation is accounted for by unobservable characteristics. Among these are the Despite the progress, Mauritian women are still con- availability and cost of child and elderly care services, as siderably disadvantaged in access to the labor market. well as the cultural values and social norms that assign The household analysis shows that the disproportionate to women a traditional role as pro­ viders of children and expansion in the labor force participation of women in the most affluent households, together with the grow- elderly care and as responsible for a broad range of non- ing correlation between the earnings of husbands and market activities. wives, has contributed to the widening in household labor income inequality. If women’s employment is to In the private sector, women are also disadvantaged in pay contribute to reducing inequality in household labor largely because of unequal treatment. Between 2004 and incomes, gender gaps in the labor market must decline, 2015, women employed in the private sector were paid, on and women’s gains must be evenly distributed. Although average, about 30 percent less than men per hour worked. the female labor force participation rate rose steadily The gap is larger among low-paid women (33.5 percent at over the decade and had reached 57 percent by 2015, the 10th percentile of the wage distribution in 2015) and up the gender differential is still large, at a staggering to the median (31.4 percent in 2015), compared with women 32 percentage points (figure ES.10, panel a). The outlook earning high wages (12.7 percent at the 90th percentile in is optimistic because most of the reduction in the gen- 2015). The gender gap in the private sector appears to be a der gap is attributable to young cohorts of women, par- consequence of the combination of two factors. First, women ticularly women with secondary educational attainment. have less productive characteristics than men; they are, for In addition, educated women are more well entrenched example, disproportionately employed in traditional sectors in the labor force than other women. The labor force par- and low-skilled occupations. Second, women are subject to ticipation rate among women with postsecondary or tertiary a large negative effect because of what the literature calls educational attainment is as high as the rate among men the unexplained component (figure ES.11, panel a). This (figure ES.10, panel b). Gaps in participation are partly component captures the effect of unobservable characteristics FIGURE ES.10. Labor Market Participation, by Gender and Educational Attainment among Women, 2004–15 a. Overall, men and women b. Women, by educational attainment 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male Up to incomplete primary Upper secondary Individuals not in education Complete primary Post-secondary/ Lower secondary tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 12 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE ES.11. Women Are Paid Unequally in the Private Sector, but Are More Well Paid in the Public Sector, 2004–15 Oaxaca-Blinder decomposition of the gender wage gap a. Public sector b. Private sector 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2005 2006 2007 2005 2006 2007 2004 2008 2009 2010 2011 2012 2013 2015 2004 2008 2009 2010 2011 2012 2013 2015 2014 2014 Difference Explained Unexplained Difference Explained Unexplained Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. that would make men, on average, more productive than young, more well educated cohorts of women. Policies that women, as well as the effect of the wage structure, in short, ease the caring burden borne by women and encourage men the unequal treatment of men and women. to become more involved in home and care duties would be welcome.9 Despite the greater labor force participation, By contrast, women employed in the public sector earn, on women are likely to continue to bear most of the burden average, more than men largely because of their highly produc- in housework and family care.10 These activities compete tive characteristics. The gender premium in 2015 is estimated for women’s time and energy with the labor market and at about 7.2 percent. However, the gender differential is not may oblige women to seek less competitive and less remu- constant throughout the wage distribution. It is larger at the nerative career paths and greater employment flexibility. bottom (+10.6 percent at the 10th percentile in 2015) and Hence, subsidized child and elderly care and work-time up to the median (+15.2 percent in 2015), while it is smaller, regulations that promote flexibility and facilitate part-time not significant, and, in some years, negative at the top. In work are likely to be effective. For example, guaranteeing the public sector, women face a moderate wage advantage women the possibility to switch to part-time schedules by virtue of their favorable productive characteristics. For in the same jobs after they deliver could help reduce the example, women civil servants are, on average, more well risk of career disruptions by allowing a smooth transition educated than their men counterparts, and they are employed from maternity leave to employment. Extending paternity in high-skilled occupations. This typically compensates leave and making it more flexible are additional tools that completely for a mild effect of the unexplained component, could ease the burden of women and reduce the cost of on average (see figure ES.11, panel b), and especially in the hiring women. lower half of the distribution. The public sector appears to be absorbing the most productive women, who also benefit Economic incentives for working women and men need from a higher wage premium. to be aligned. To the extent that the gender pay gap in the private sector is the result of an unequal pay structure, To narrow the gender gap in labor force participation, labor changes need to occur through the education system that market policies need to be more woman friendly. Women’s place a stronger focus on curbing discriminatory social norms participation has the potential to increase and contribute among youth. The public sector could serve as a model of to narrowing inequality in household labor income and best practice in engaging women in the labor market and achieving the full potential of the economy. In the decades promoting more equitable treatment. Awareness campaigns ahead, it is expected to follow the trends established over the might also help shift norms and biases on the employment last 10 years because the pattern has largely been driven by of women in high-paying positions. Executive Summary 13 Unemployment among compared with the rate among individuals in the 25–29 age-group (figure ES.13, panel a). The unemployment rate Educated Youth Highlights among the younger age-group rose from about 19 percent in 2008 to 25 percent in 2015. Unemployed youth are the Education Mismatch increasingly highly educated. In 2006, around 55 percent among the Employed of unemployed youth ages 16–29 had upper-secondary educational attainment, and less than 7 percent had post- In addition to a skills shortage, the Mauritian labor market secondary or higher education (figure ES.13, panel b). In appears to be characterized by an education mismatch, 2015, in addition to a reduction in the share of unemployed especially among youth. The education mismatch cap- youth who had, at most, completed primary education, the tures the fact that the educational attainment of workers share of unemployed youth with upper-secondary educa- does not match the education required in the jobs they tion had fallen to 42.5 percent, while the share of youth perform.11 Overall, the share of mismatched workers, with postsecondary or higher education had jumped to either over- or undereducated, was roughly constant at almost 40.0 percent. about 47 percent in the last decade. Yet, the share of over- educated workers rose from 8 percent to 13 percent, on Learning achievement often does not match the type and average, notably, among youth (figure ES.12). The share quality of skills required by employers. Of special concern of overeducated workers ages 15–29 doubled between is the combination of three factors: (1) a sizable high-skills 2006 and 2015. Women spearheaded this trend. Besides shortage is driving the expansion in wage inequality; (2) the the negative effects in rising wage inequality, this type education mismatch is becoming worse, as shown by the ris- of mismatch can have negative consequences at both ing share of overeducated youth; and (3) unemployed youth the micro- and the macrolevel. First, it can impair the are increasingly highly educated. This signals a potential productive potential of youth and thus influence lifetime inefficiency in the system whereby labor demand does not patterns of employment and pay. Second, it may hinder match labor supply. Although the Mauritian population economic growth, productivity, and competitiveness in has achieved considerable progress in education, the edu- the long term. cation system is not providing workers, especially youth, with the high-quality learning required by employers. In addition, youth unemployment is on the rise, and un- This hypothesis finds corroboration in the results of the employed youth are increasingly highly educated. Unemploy- surveys of the Organisation for Economic Co-operation ment among youth ages 15–24 has consistently been three and Development’s (OECD) Program for International times the overall unemployment rate and significantly higher Student Assessment.12 According to the survey conducted FIGURE ES.12. The Share of Overeducated Youth Is on the Rise, 2006–15 a. Overeducated workers, by age-group b. Undereducated workers, by age-group 22 50 20 45 18 40 16 Percent Percent 14 35 12 30 10 25 8 6 20 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 30−44 16−24 30−44 25−29 45−64 25−29 45−64 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 14 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE ES.13. Youth Unemployment Is High, and Unemployed Youth Are Increasingly More Well Educated, 2006–15 a. Unemployment rate, by age-group 30 25 20 Percent 15 10 5 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 25−29 30−44 45−64 b. Unemployed youth, by educational attainment 100 3.1 3.7 5.0 4.7 6.1 3.7 4.2 3.9 4.3 7.1 15.8 19.9 21.4 24.5 26.9 80 8.7 14.5 14.0 13.2 55.1 12.6 53.7 59.0 53.9 60 53.4 Percent 46.2 42.5 40.3 40 42.2 42.5 11.7 14.7 17.6 15.7 15.9 20 17.2 13.5 15.2 26.4 13.1 22.3 19.8 12.9 17.4 17.5 12.1 9.6 9.2 7.0 5.0 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Up to complete primary Upper secondary Tertiary Lower secondary Post-secondary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Executive Summary 15 in 2010, 15-year-old Mauritian students lag behind cor- NOTES responding students in otherwise comparable countries in 1. For the purpose of this study, total household income is defined as the learning achievement, including in reading, mathematics, sum of income from wages and self-employment, property income, and and science literacy (Walker 2011). On the demand side, current transfers (public and private) received, excluding income from the production of household services for own consumption (the net a series of sector-specific surveys on labor shortages and value of owner-occupied dwellings). The income aggregate is before the skills gap conducted in 2011 indicates that employers taxes because the data do not allow the identification of the amount of taxes paid by each household consistently over time. face difficulty finding workers with the adequate techni- 2. To capture fully the distributional effects of government redistribution cal skills, soft skills, and past working experience for the through taxes and transfers, a fiscal incidence analysis is required. This type of analysis consists of allocating taxes, particularly personal jobs they are offering (HRDC 2012a, b, c, d). Moreover, income tax and consumption taxes, and public spending, particularly social spending, to households or individuals and comparing incomes in 2016, employers reported the educational inadequacy before taxes and transfers with incomes after taxes and transfers. of the workforce as the third most problematic factor in 3. Dispersion in individual monthly wages increased substantially between 2004 and 2015, and this was largely driven by wage dynamics rather doing business in Mauritius (Schwab 2016). than by the changes occurring in the number of hours worked. 4. High educational attainment indicates upper-secondary or higher education; low educational attainment includes any schooling up to The education mismatch on the labor market and the rising completed primary education. 5. Work experience represents potential work experience because it is calcu- share of well-educated youth among the unemployed call lated as age-years of education, less 6, which is the initial age of schooling. for a two-pronged strategy. On the one hand, the education 6. In addition to the rise in inequality among demographic groups, particularly between high- and low-skilled workers, there was an mismatch can be addressed through targeted training and expansion in within-group inequality, that is, in hourly wage inequality retraining programs for undereducated workers, who are within groups identified by gender, education, and experience. Both between- and within-group inequality is, however, largely ascribable mainly middle-age and older workers. The aging Mauritian to the effect of changes in the price of labor (hourly wages), rather population renders the adoption of a life-cycle approach to than to changes in the composition of the workforce. 7. The estimated effect is likely to be an upper bound of the true effect learning key to the success of this strategy. On the other hand, of ROs on inequality (see appendix A). the younger cohorts of workers who are increasingly over- 8. Coverage gaps represent the share of wage earners who are not covered by minimum wage legislation. Compliance gaps represent the share of educated for the jobs they perform or who are unemployed wage earners who are covered by minimum wage legislation, but still despite their high educational attainment are evidence of the make subminimum wages. 9. However, a considerable portion of the gender gap in participation remains need for improvement in the effectiveness of targeted youth unexplained. While several hypotheses may be proposed, including the employment schemes and well-functioning employment accessibility and cost of childcare, the choice of curricula that are less likely to be associated with good job outlets, and social and cultural services. The quality of learning needs to be reviewed, and norms, additional analysis is needed to provide more precise answers. technical and vocational education and training need to be 10. With financing from the World Bank under the multidonor Trust Fund for Statistical Capacity Building, Statistics Mauritius will carry made relevant to the changing needs of the labor market and out a living conditions survey that will also collect information on become more attractive to more youth, who often view such time use. This will help in comparing the time devoted by employed and nonemployed men and women to household activities, including curricula as considerably less valuable because the certifi- routine chores and family care. cates with which they are associated have not been obtained 11. This analysis relies on a measure of the match between skills and job tasks and duties, the International Standard Classification of Occupations. through an academic education. This requires promotion This normative measure is based on a division of major occupations into broad groups. It assigns a level of education to each occupational group and communication efforts, accompanied by enhanced and in accordance with the International Standard Classification of Education. continuous career guidance. The active involvement and Workers in a group who have the assigned level of education are considered well matched. Those who have a higher (lower) level of education are ownership by employers in skills development and applica- considered overeducated (undereducated). An advantage of the measure tion are key to making the response more effective. With resides in the fact that the definition of a mismatch does not change over time; the results are therefore strictly comparable. A disadvantage of a view to enhancing efficiency, the government might also the measure is the fact that formal education is only one component of review the existing range of incentives—including collective the measurement of skill level and can be subject to measurement error. 12. See PISA (Programme for International Student Assessment) (database), training funds, tax incentives, and payback clauses—and Organisation for Economic Co-operation and Development, Paris, the international evidence on what works. http://www.oecd.org/pisa/pisaproducts/.  17 INTRODUCTION S ince independence in 1968, Mauritius has posted and 16 percent, respectively. Within the secondary sector, steady progress in economic performance that is the GDP shares of manufacturing, water, electricity, and often labeled the Mauritian miracle or the success construction declined by between 13 percent and 17 per- of Africa. Svirydzenka and Petri (2014) describe how cent over the period. Ultimately, a key structural shift has Mauritius transitioned from a low-income monocrop occurred in the Mauritian economy away from primary exporter, subject to terms-of-trade and output shocks, high and secondary production to an economy characterized population growth, and ethnic tensions, to a diversified by the dominance of services. upper-middle-income economy. Mauritius is today one of the strongest economies in Africa, with a per capita Economic growth has been accompanied by a considerable income of US$9,780 in 2015, the third highest in Africa reduction in poverty. Absolute poverty, calculated using (World Bank 2017). Now, the country is aiming to achieve the 2006/07 relative poverty line as a fixed threshold, a second economic miracle and aspires to join the group declined from 8.5 percent to 6.9 percent between 2007 and of high-income economies by 2023. 2012 (World Bank 2015a).1,2 In parallel, household income inequality, as measured by the Gini coefficient, rose from About a decade ago, the Mauritian economic model began 0.34 to 0.37 (World Bank 2015a) (box I.1).3 Households encountering the first serious challenges. The loss of prefer- in the middle and top of the income distribution gained ential access of the country’s sugar and textile production more than those at the bottom. Households in the bottom to the European Union and U.S. markets, negative terms of 3 percentiles of the household income distribution saw trade, and growing international competition in low-cost their income decline by about 1.8 percent a year (World industries slowed growth and led to rising unemployment. Bank 2015a). The 2015 Mauritius, Inclusiveness of Growth The government liberalized the economy and unleashed and Shared Prosperity report also illustrates that most of resources in support of expanding sectors. The labor market the expansion in household income inequality is ascrib- was reformed; sectors were opened to foreign investment; able to income from labor, particularly to income from the business climate improved; and tax compliance was wage jobs, in which most of the Mauritian working-age simplified. Thanks to a social contract based on inclusion, population is employed (World Bank 2015a). The World redistribution, and private-public dialogue, the effort paid Bank (2015a) estimates that the reduction in absolute off in more economic growth, employment creation, foreign poverty would have been twice as large if growth had direct investment, private investment, and a declining public been more equitably shared and if inequality had not debt ratio. widened. At the same time, the country’s comprehensive social protection system helped contain the increase in the Since 2010, economic growth has fallen short of the aspira- Gini coefficient to 4 percentage points (World Bank 2015a). tions of the government and the people. Economic growth However, despite substantial effort undertaken by the slowed; job creation is weak; and inequality is widening. government to instill greater coordination across social The macroeconomic prospects are moderately positive. The protection programs, improvements are needed (World economy grew by 3.5 percent in 2015 and by approximately Bank 2015a). For example, the basic retirement pension, 3.7 percent in 2016. Between 2004 and 2014, the tertiary a universal noncontributory social pension paid to per- sector was the main driver of economic growth, posting an sons over age 60, lacks a focus on the poor because it overall increase of 11 percent to account for 73.5 percent disproportionately favors well-off households, and funding of total gross domestic product (GDP) in 2014. Within is low for programs specifically intended to benefit the poor the tertiary sector, finance is the largest and most rapidly (World Bank 2015a). growing subsector. Its contribution to GDP rose by 27 per- cent over the period. Agriculture contracted the most. Its The most recent country partnership framework document contribution to GDP shrank by half, from 6 percent in 2004 identifies three strategic focus areas for the period between to only 3 percent in 2014. The shares of GDP attributable fiscal year 2017/18 and fiscal year 2021/22: increasing to primary and secondary sectors declined by 47 percent competitiveness, fostering inclusion, and bolstering resilience 18 Mauritius: Addressing Inequality through More Equitable Labor Markets BOX I.1. Household Income: Choices, Markets, and Institutions Household equivalent disposable income is the sum of income from labor (both from wages and self-employment), income from assets, and income from transfers (both public and private), minus taxes and social security contributions. Inequality in household equivalent disposable income is the by-product of inequality in each component. Household labor income is one of the most important components. Markets, institutions, and the choices of households and individuals can have an impact on the level of inequality (figure BI.1.1). For example, the size and the composition of each household have a direct effect on equivalent household labor income and an indirect effect through the labor market choices of household members in the supply of labor on the extensive and intensive margin, that is, respectively, (1) whether to participate in the labor market or not and (2) conditional on participation in the decision, the number of working hours to offer. Labor market forces, captured by the interaction of labor supply and demand, affect the earnings of individuals because they determine the price of labor, the wage. Labor market institutions, including the minimum wage and labor unions, can alter what would otherwise be the price of labor determined by market forces alone. FIGURE BI.1.1. Choices, Markets, and Institutions Affect Labor Income Choices Household Labor Market Working HOUSEHOLD LABOR INCOME (family; individual) Demographics Participation Hours FACTORS AFFECTING Household Labor Labor Market Demand Supply Income Institutions Minimum Wage Unions Within this context, this study presents the patterns of total household income inequality and the main components of this inequality. Chapter 2 explores the role of household demographics and market factors in household labor income inequality. Chapter 3 takes a deep dive into the role of women in the labor market. Chapter 4 illustrates detailed trends in wage inequality as a driver of inequality in household labor income and investigates the main source of rising wage inequality, particularly a skills shortage. and sustainability (World Bank 2017). Building on the factors is important for identifying effective policy levers to findings of the World Bank (2015a) and on the country mitigate and possibly reverse the upward trends, consolidate partnership framework’s focus on fostering inclusion, recent progress, and ensure Mauritius enters a sustainable this study analyzes the dynamics of inequality over the track toward high-income-country status. longer period 2001–15 with a view to understanding the driving forces behind the growth in inequality and identify The report is structured as follows. Chapter 1 sets the stage possible policy levers that could mitigate and, in the long by presenting stylized facts on the trends in household income run, reverse the upward trend in inequality. This objective inequality between 2001 and 2015, comparing these trends becomes more relevant given the recent global economic with trends in consumption inequality, and identifying the slowdown and terms-of-trade shocks because narrowing main culprit behind the rapidly rising inequality in house- inequality is key to fostering economic growth and achieving hold incomes, that is, household labor income. Chapter 2 the twin goals of eliminating extreme poverty and boost- supplies a set of descriptive trends of the two groups of ing shared prosperity.4 factors, namely, household demographics and labor mar- ket forces, that contribute to changes in household labor Unravelling the conundrum of the growing income inequality income and follows up with a decomposition exercise on in Mauritius by understanding the main sources and driving changes in household labor income between 2001 and 2015. Introduction 19 Because the analysis indicates that an unequal increase in candidates include the inter­ action of changes in labor female labor force participation and rising inequality in supply and labor demand, giving rise to skills shortages or individual earnings are among the main contributors to the surpluses, and changes in labor market institutions, namely, expanding inequality in household labor income, Chapter 3 remuneration orders (ROs). The chapter concludes with an takes a deep dive into the issue of gender inequality in the analysis of an additional source of skills mismatches among labor market. The chapter illustrates the gender gap in the employed population, namely, education mismatches, labor market participation, describes the differences in the and advances potential explanations for the coexistence activities of working women in the labor market relative to of a substantial skills shortage, overeducation, particularly men, and concludes with a detailed analysis of gender gaps among youth, and a large share of highly educated youth in wages separately in the public and private sectors. Chap- among the unemployed. ter 4 resumes the main analysis of the drivers of increasing inequality in individual earnings. The chapter first presents The analysis briefly described above has required the use of stylized facts about overall inequality in wages and then a number of different empirical approaches. Box I.2 defines separates out changes in inequality between and within income, and table I.1 summarizes the main methodologies groups defined by demographic characteristics. The chapter adopted in each chapter, together with the references in the distinguishes the role of changes in prices (or wages) and economic literature, and a short, yet exhaustive description the role of changes in the composition of the workforce in of the approach taken each time. Appendix B illustrates a rising earnings inequality. The second part of the chapter description of the major data sources used throughout the is devoted to the analysis of the role of the main potential report and the definitions adopted to define labor market drivers of expanding earnings inequality. The possible variables. BOX I.2. The Definition of Income The conceptual definition of household income established at the 17th International Conference of Labour Statisticians and adopted by the Canberra Group in the second edition of the handbook published in 2011, is the following:a Household income consists of all receipts whether monetary or in kind (goods and services) that are received by the household or by individual members of the household at annual or more frequent intervals, but excludes windfall gains and other such irregular and typically one-time receipts. Household income receipts are available for current consumption and do not reduce the net worth of the household through a reduction of its cash, the disposal of its other financial or nonfinancial assets, or an increase in its liabilities. Household income may be defined to cover: (1) income from employment (both paid and self-employment), (2) property income, (3) income from the production of household services for own consumption, and (4) current transfers received. Data of the Continuous Multi-Purpose Household Survey (CMPHS) allow the identification of all the components of household income listed above with the exception of income from the production of household services for own consumption, namely, the net value of owner-occupied dwellings.b Income from employment is derived from the income module of the survey in which each household member is asked to report the last monthly payment from any wage job and income from self-employment. The income section also provides information about other sources of income, including goods produced for own-consumption or barter (and including vegetables, fruits, eggs, fish, and so on), income from assets, and transfer income. Income from property includes rent from land, buildings, machinery, equipment, and so on. Income from financial assets includes, for example, interest and dividends. Income from transfers is composed of both private and public transfers. Among the first are regular transfers from parents or relatives, regular allowances from social or religious organizations, maintenance allowance or alimony, pension from employers (privately funded and voluntary), and other regular incomes. Public transfers include social security benefits (old-age pensions and others), pension from the National Pension Fund, and widow and children pensions. The most comprehensive household income measure adopted in this study is total income, which is the sum of income from employment, assets income, and transfer income (for example, see figure BI.2.1). Unfortunately, the information available from the survey does not allow a consistent identification of the amount of taxes paid over time, and therefore a measure of disposable income cannot be constructed. (continued) 20 Mauritius: Addressing Inequality through More Equitable Labor Markets BOX I.2. The Definition of Income (continued) FIGURE BI.2.1. Density of Log-Household Income with and without Housing Rental Value, 2015 .8 .6 .4 .2 0 6 8 10 12 14 Log PAE HH total income Log PAE HH total income (including imputed rent) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Including the value of the flow of services derived from owner-occupied dwellings would decrease inequality, particularly in the lower tail of the distribution. Using information from the 2015 CMPHS about owner estimates to account for the value of housing services, the Gini coefficient would decline from 42.2 to 39.8 (-5.8 percent) and the income ratio of the 90th percentile of the distribution to the 10th percentile (P90/P10) by 12 percent. The reduction in inequality would be larger in the lower tail. The P50/P10 ratio would fall by 8.4 percent compared with a 3.8 percent reduction in the upper tail (the P90/P50 ratio). a. See “Resolution I: Resolution Concerning Household Income and Expenditure Statistics” International Labour Organization, Geneva, http://www.ilo. org/wcmsp5/groups/public/—-dgreports/—-stat/documents/normativeinstrument/wcms_087503.pdf; UNECE (2011). b. See CMPHS (Continuous Multi-Purpose Household Survey) (database), Statistics Mauritius, Port Louis, Mauritius, http://statsmauritius.govmu.org/ English/CensusandSurveys/Pages/Continous-Multipurpose-survey-Lists.aspx. TABLE I.1. List of Methodologies Adopted in the Analysis Reference Chapters literature Description of the methods adopted in the study Chapter 1 Decomposition: Azevedo, This method implements a Shapley decomposition of changes in a welfare indicator, the Gini changes Nguyen, and index in this case, by constructing counterfactual distributions for period t + 1 by substituting in income Sanfelice the observed level of the indicators in period t, one at a time. For each counterfactual inequality (2012); distribution, inequality measures can be calculated, and the counterfactuals are interpreted Azevedo, as the inequality level that would have prevailed in the absence of a change in that Sanfelice, and indicator. The decomposition is path-independent because the methodology calculates the Nguyen (2012) decomposition across all possible paths and then takes the average among them. Chapter 2 Decomposition: Fortin and Schirle This decomposition method is used to gauge the relative contribution to changes in the density household (2006) of (log equivalent) household labor income of six factors: men’s labor income, female labor labor income force participation, women’s labor income, assortative mating, household mix, and household characteristics. The approach consists of a conditional reweighting decomposition carried out in a series of sequential steps. After each step, the counterfactual densities of household labor income and the corresponding inequality measures are compared with those based on observed household labor income. In the first step, log individual earnings of men in the initial year (2001) are regressed on total hours worked, dummy variables for three education categories (up to completed primary, lower secondary, and upper secondary and above), a quartic in experience and interactions of the experience quartic with education categories, and district fixed effects. Estimated coefficients are applied to the characteristics of a sample of men in the final year (2015) so as to generate a counterfactual men’s earnings distribution for (continued) Introduction 21 TABLE I.1. List of Methodologies Adopted in the Analysis (continued) Reference Chapters literature Description of the methods adopted in the study 2015 that holds men’s earnings structure at the 2001 level.a The counterfactual is derived by replacing actual men’s earnings with the predicted ones in the 2015 observed household labor income distribution. This exercise allows an appraisal of the contribution of changes in men’s earnings to household labor income inequality. Building on the previous phase, the second step adjusts the household labor income distribution obtained in the previous step by keeping female labor force participation at the level observed in 2001. First, the estimated coefficients from a participation regression run on a sample of women (heads or spouses) in 2001 are used to generate counterfactual fitted participation rates of a corresponding sample of women in 2015, and participation rates are also predicted for 2015. Second, a weight is constructed that is a function of the ratio between counterfactual and fitted participation rates. The weight, multiplied by the sampling weight, is employed to reweigh the household labor income distribution obtained at the end of step 1 and derive a new adjusted distribution that additionally incorporates the impact of changes in female labor market participation. Step 3 repeats the routine performed in step 1, but on a sample of women and also after an adjustment for selection bias associated with participation. Similarly, the remaining phases involve carrying out ad hoc versions of step 2 that model different outcomes, but each ultimately produces an adjustment weight combining counterfactual and actual predicted outcomes. Specifically, in step 4 (assortative mating), the outcome being considered is spouse labor income correlation among married-couple households with and without children.b In step 5 (family mix), the outcome to be corrected is the probability of membership in a given family type. In the last step (household characteristics), it is the probability, calculated separately for married- and unmarried-couple households, that heads and spouses have certain characteristics, such as age, educational attainment, and district of residence. Once all steps have been performed, the resulting 2015 adjusted log equivalent household labor income distribution can be brought into comparison with the 2001 and 2015 observed distributions. However, being a progressive routine, the final outcome is contingent on the order in which phases are carried out. A reverse order decomposition is therefore performed to check the consistency of the results. The decomposition is implemented on a sample of households restricted to those with nonzero income from labor, with household heads (and spouses in case of married-couple households) of working age (16–64) and either employed with nonzero individual labor income and nonzero working hours or unemployed or inactive. Chapter 3 Blinder-Oaxaca Blinder (1973); The Blinder-Oaxaca decomposition is used to gauge the extent to which differentials in decomposition Oaxaca (1973) observed outcomes (in our case, labor force participation and hourly wages) between two by gender: comparison groups (men and women in the analysis) are ascribable to differences in the labor force observed and unobserved characteristics of the two groups. The effect associated with participation the first difference constitutes the explained component of the differential, also known differential; as characteristics, composition, or endowment effect, in that it reflects the portion of hourly wage the differential associated with group differences in individual observable attributes (for differential by example, education, experience, main sector of activity, industry, occupation). The effect gender related to the second difference is referred to as the unexplained component. This embodies the portion of the outcome gap stemming from the differential valuation of women’s and men’s characteristics in the labor market that arises because of differences in unobservable characteristics or unequal pay structures between the two groups. Chapter 4 Compositional Autor, Katz, and The compositional adjustment of hourly wages serves the purpose of depurating observed adjustment of Kearney (2008) wages from the effect of shifts in the gender, experience, education composition of the wages workforce over a given period, thus allowing wage comparisons over time that are not mechanically affected by shifts in the sample composition. To derive composition-adjusted wages, the data are sorted into 18 sex-education-experience cells based on the combinations of two sexes, three education categories (up to completed primary, lower secondary, and upper secondary and above), and three potential experience categories (0–14, 15–34, and 35+ years). Log hourly wages of workers ages 16–64 not in education are then regressed for each year and separately by sex, on dummy variables for three education categories, a quartic in experience, and interactions of the experience quartic with education categories. The composition-adjusted mean log hourly wage for each of the 18 cells in a given year is the predicted log hourly wage from these regressions evaluated at the relevant experience level (7, 25, and 40 years depending on the experience group) and educational level. (continued) 22 Mauritius: Addressing Inequality through More Equitable Labor Markets TABLE I.1. List of Methodologies Adopted in the Analysis (continued) Reference Chapters literature Description of the methods adopted in the study Composition-adjusted log hourly wages for broader demographic groups (defined by gender, educational level, or potential experience) in each year are calculated as weighted averages of the relevant composition-adjusted cell means in that year, using a fixed set of weights equal to the cell shares of total hours worked in each macrogroup over 2004–15. By proceeding in this way, the relative employment shares of given demographic groups are held constant across the whole period. For example, in a given year, the composition- adjusted log hourly wage for the macrogroup of women is given by the weighted average of that year’s mean predicted wages in each of the nine underlying education-experience cells for women, where weights are the constant ratios of the hours worked by individuals in each cell to the total number of hours worked by all women, computed over the entire period. Relative supply Katz and Murphy For the construction of relative supply measures, a quantity sample is constructed that consists shifts (1992) of total yearly hours worked by all employed workers (including those in self-employment) ages 16–64 in each of 18 sex-education-experience cells resulting from combinations of two sexes, three education categories (up to completed primary, lower secondary, and upper secondary and above), and three potential experience categories (0–14, 15–34, and 35+ years). The quantity sample is then matched to a price sample containing real mean hourly wages for each cell in any given year. The wages are normalized to a relative wage measure obtained by dividing them by the wages of a reference earnings group (consisting of men holding upper-secondary education and above and 10 years of experience) in the contemporaneous year. An efficiency unit measure for each sex-education-experience cell is then derived as the arithmetic mean of the relative wage measure in that cell over the entire 2004–15 period. By multiplying mean relative wages by a cell’s quantity of labor supply in year t, as expressed in terms of annual hours worked, it is possible to obtain that group’s labor supply measure in efficiency units for that year. By so doing, yearly quantities of labor supply are expressed in terms of the average marginal productivity they are paid on the market over the entire period. To compute labor supply measures in efficiency units for broader demographic groups, it is sufficient to sum over supply measures in each of the underlying cells in a given year. Finally, for each year, relative supply measures in efficiency units are derived by dividing each cell or aggregate group’s supply quantities by the total supply of all groups. Relative demand Katz and Murphy Relative demand shifts can be thought of as arising from between- and within-industry changes. shifts (1992) Between-industry changes are shifts that change the allocation of total labor demand between industries at fixed relative wages. Within-industry changes are shifts that change relative factor intensities within industries at fixed relative wages. To measure the role of changes in relative demand, the following overall demand shift indicator is constructed: d ∆X k = ∆Dk = ∑α j jk ∆E j (TI.1.1) Ek Ek This index measures demand changes within 27 industry-occupation cells separately for each sex-education demographic group, and it is relative because it is deflated by total employment of the same group in a reference period (2004–15). Decomposition DiNardo, Fortin, This reweighting approach is employed to assess the extent to which changes in inequality are of hourly wage and Lemieux ascribable to price effects resulting from the interaction of labor demand and labor supply and inequality: (1996) to compositional effects that mechanically introduce changes in inequality by altering the shares price of demographic groups that have more or less dispersion in wages. The approach consists effect and of decomposing the observed density of wages in two time periods, say t and t′, into a price composition function that provides the conditional distribution of wages for given characteristics and time effect and a composition function that provides the density of characteristics over that time period. It is then possible to construct a counterfactual wage density and counterfactual inequality measures by combining the price function from a period t with the composition function from a different period t′. To calculate the counterfactual, it is necessary to reweigh the price function at time t by the ratio of the density of characteristics at time t′ to the density of the characteristics at time t. In practice, the reweighting function can be estimated using a logit or probit model applied to the pooled data from times t and t′. (continued) Introduction 23 TABLE I.1. List of Methodologies Adopted in the Analysis (continued) Reference Chapters literature Description of the methods adopted in the study To assess the contribution of shifts in composition and prices to observed changes in overall and residual inequality, the workforce composition data in each year between 2004 and 2015 are applied to the price function from the years 2004, 2008, and 2015. This allows the simulation of a set of hypothetical scenarios wherein workforce composition changes as it actually did over time while prices are held constant at their 2004, 2008, and 2015 level. In the calculation of the reweighting function, a set of demographics characteristics, including dummies for education, a quartic in experience, interactions of the experience quartic with education categories, and dummies for district of residence, are controlled for in regressions run separately by gender. The outlined procedure is applied to the construction of counterfactuals for overall inequality. In the case of residual inequality, the price function is replaced by a residual pricing function obtained by regressing the logarithm of hourly wages in each year on the full set of characteristics described above and replacing the wage observations with corresponding residuals from the ordinary least squares regression. The residual price function is then used to calculate counterfactual residual densities. Remuneration Autor, Manning, See appendix A. orders: effect and Smith on inequality (2016); Bosch and Manacorda (2010) Remuneration Gindling and See appendix A. orders: Terrell (2007) effect on employment and working hours a. To enhance precision in the construction of the counterfactual distribution, fitted values are augmented by adding the predicted residuals obtained from an analogous regression carried out on the 2015 subsample. b. The reweighting function is set to 1 for single individuals and for couples with only one working spouse. NOTES expected to have a negative effect on growth, and inequalities arising because of differences in effort that might go in the opposite direc- 1. Over the same period, relative poverty, measured against a poverty line tion. Recent estimates for countries of the Organisation for Economic equal to 50 percent of median per adult equivalent household income, Co-operation and Development (OECD) show that inequality is harmful expanded from 8.5 to 9.8 percent. to medium-term growth: a rise in inequality of 3 Gini points would 2. The income aggregates derived from CMPHS and HBS data are not reduce economic growth by 0.35 percentage points a year for 25 years strictly comparable as the survey instruments differ between the two (OECD 2015). Some studies have proposed that this ambiguity may surveys. arise because income inequality has different and offsetting effects on 3. In the United States, the Gini coefficient increased by 4.5 percentage growth among population subgroups. Marrero and Rodríguez (2013) points between 1977 and 1992, the period during which income find that it is not overall inequality, but inequality of opportunity dispersion has been most often studied (Atkinson 2015). that has a negative effect on growth because it favors human capital 4. While, in theory, the effect might go either way, and various channels accumulation by individuals with better socioeconomic backgrounds, exist through which inequality can be predicted to influence growth, rather than by the most talented individuals. Van der Weide and the empirical literature attempting to establish the direction of the Milanovic´ (2014) find that income inequality is bad for the growth relationship does not reach a consensus. Recent research summarized prospects of the poor, but good for the rich. Marrero, Rodríguez, by the World Bank (2016) shows that narrower inequality in disposable and van der Weide (2016) estimate that inequality of opportunity is income is correlated with more rapid and durable growth for a given level bad for growth of the poor and that, after controlling for inequality of redistribution. However, these results have been challenged. Other of opportunity, the effect of overall inequality on growth is drastically recent studies distinguish between inequality of opportunity, which is reduced.  25 CHAPTER 1 Inequality in Mauritius: Stylized Facts A s documented in World Bank (2015a), the years economic downturn (2008–09) and in the aftermath of between 2007 and 2012 were marked by steady the loss of the preferential trade agreements of Mauritius economic growth, but limited shared prosperity, (the Sugar Protocol and the Multi Fibre Arrangement) and whereby affluent households benefited more from eco- continued to expand thereafter, markedly over the course nomic growth compared with households at the bottom of of 2010–15. Third, while business cycle fluctuations can the distribution of both consumption and income. A similar affect the consumption and income distribution, household pattern emerges if the period of analysis is expanded to income is more severely hit by macroeconomic trends the years between 2001 and 2015 (figure 1.1). Households relative to household consumption. Households at the initially in the lowest 5 percentiles of the distribution bottom of the income distribution experienced a modest recorded an average annual income gain of less than decline in income in 2008 relative to the level in 2001, 1 percent, whereas households in the top 10 percent posted while their consumption level never declined below the an average annual income gain of about 3.6 percent. This level at the beginning of the period.1 compares with an annual average growth rate of per adult equivalent income of about 3 percent (figure 1.1, Income inequality measured by the Gini coefficient increased panel b). from 0.36 in 2001 to 0.42 in 2015 (+16.4 percent), whereas the P90/P10 ratio rose by 36.9 percent over the same period The inequality gap in Mauritius is moderate compared (figure 1.4, panels a and b).2 The differences in the trends with countries at a similar level of economic develop- of the Gini coefficient and of the P90/P10 ratio suggest ment and narrower than the inequality gap in South substantial changes occurred at the tails of the distribu- Africa (figure 1.2). Among 216 countries on which data tion. Income dispersion expanded more in the lower tail of are available, 88 countries (40 percent of the total) are the distribution: the P50/P10 ratio rose by 18.5 percent, more unequal than Mauritius. However, the inequality whereas the P90/P50 ratio rose by 15.5 percent (figure 1.4, in Mauritius has widened substantially over the last panels c and d). 15 years. The patterns of household income growth are reflected in Total household income is composed of income from the trends in inequality. Figure 1.3 shows clearly the fanning employment, income from assets, and income from public out of the distribution of both consumption and income and private transfers (see Introduction, box I.2 for details). over time. First, the median of the distribution has mir- Figure 1.4 shows clearly that each dispersion metric varies rored a rise in household consumption and income in real significantly for each component and that the increase in terms by 45 percent between 2001 and 2015. Households total income dispersion is largely ascribable to the increase above the median and, notably, above the 90th and 95th in labor income inequality. Government redistribution, percentiles, recorded a 62 percent increase in consumption captured by the public transfer income component, serves and in income. By contrast, households at the bottom, as a stabilizing force against cyclical fluctuations and plays particularly those in the 5th and 10th percentiles, experi- an important inequality-compressing role, notably, in the enced consumption growth of around 22 percent or less. lower tail of the distribution both in static and dynamic The incomes of the most disadvantaged households have terms. First, inequality in income before government transfers increased by as little as 14 percent over the entire period. is always wider than inequality after public transfers. For Second, changes in inequality were not constant through- example, in 2001, the P90/P10 ratio was at 5.7 based on out the period. Both income and consumption inequality household incomes before government transfers compared started to rise significantly at the beginning of the global with 4.6 in the same year, but calculated based on household 26 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 1.1. Growth Incidence Curve, Household Consumption and Income, 2001–15 a. Consumption b. Income Growth incidence curve: 2001–2015 Growth incidence curve: 2001–2015 4 5 Annual growth rate (percent) Annual growth rate (percent) 4 3 3 2 2 1 1 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Percentiles Percentiles Household consumption expenditure (per adult equivalent) Household total income (per adult equivalent) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 1.2. Inequality in Mauritius and in the Rest of the World, 2000–12 70 60 50 Gini index 40 30 20 4 6 8 10 12 Log per capita GDP (constant 2010US$) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius; WDI. Note: Red dots indicate Mauritius; orange dots indicate South Africa. Inequality in Mauritius: Stylized Facts 27 FIGURE 1.3. Trends of Selected Percentiles of Household Consumption and Income, 2001–15 a. Consumption b. Income 70 70 60 60 50 50 40 40 Percent Percent 30 30 20 20 10 10 0 0 –10 –10 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p5 p25 p75 p95 p5 p25 p75 p95 p10 p50 p90 p10 p50 p90 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius; WDI. Note: Each line shows the trend in selected percentiles of the consumption or income distribution (normalized to zero in 2001). income after transfers. The income differences post- and rise in the P50/P10 ratio on household income including pre-government transfers are larger in the lower tail com- private transfers. By contrast, the difference in income pared with the upper tail of the distribution: the gap in inequality with and without private transfers is negligible the P50/P10 ratio is 0.36 compared with the gap of 0.14 in in the upper tail (figure 1.4, panel d). the P90/P50 ratio. Second, public transfers count in dynamic terms because they helped mitigate the increase in inequality The role of property income, that is, income deriving from over time. While income inequality after government trans- asset ownership, was limited. The Gini coefficient of house- fers (the P90/P10 ratio) rose by 37 percent between 2001 hold labor income increased by over 19 percent; yet, the and 2015, income inequality before government transfers growth in inequality was even stronger at the bottom of the rose by over twice as much (78 percent). The inequality- distribution. The P50/P10 ratio has gone up by 167 percent compressing effect of public transfers in the lower tail of over the last 15 years. the distribution over the period of analysis is impressive. While the P50/P10 ratio of post-transfer income grew by The expansion in labor income inequality is concentrated 18.5 percent, the same indicator calculated on pre-transfer in the second half of the period, between 2008 and 2015. income skyrocketed, recording a 47 percent increase. Thus, It started in the aftermath of the global economic down- in the absence of any form of public transfer, households in turn and trade shocks that hit Mauritius and has contin- the lower tail of the distribution would have experienced ued since. Labor income has had an even more greater an inequality gap 2.5 times larger than the one observed influence on inequality. To put things in perspective, the between 2001 and 2015. average household labor income in the 10th percentile in 2015 was MUR 1,200 (MUR 2,400 in 2001) com- Private transfers also had an inequality-containing role, pared with MUR 27,300 (MUR 16,900 in 2001) in the notably in the lower tail. Private transfers include private 90th percentile. pensions, alimony, regular transfers from parents and other relatives, regular transfers from social or religious organi- More formally, an increase in income inequality is zations, and other forms of regular income.3 Figure 1.4, the by-product of any combination of three factors: panel c illustrates that the rise in inequality calculated on (1) the increasingly unequal distribution of any particu- an income measure that excludes private transfers was as lar source of income, (2) a rising share of any particular high as 136 percent, which compares with a 47 percent unequally distributed source of income, and (3) the allocation 28 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 1.4. Measures of Household Income Inequality, 2001–15 a. Gini coef cient b. P90/P10 ratio .5 22 .48 20 18 .46 16 .44 14 Ratio Gini .42 12 .4 10 8 .38 6 .36 4 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Labor income Labor + property income Labor + property income Labor + property + private transfer income Labor + property + private transfer income Labor + property + total transfer income Labor + property + total transfer income c. P50/P10 ratio d. P90/P50 ratio 8 3.2 3.1 7 3 6 2.9 2.8 Ratio Ratio 5 2.7 4 2.6 2.5 3 2.4 2 2.3 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Labor income Labor + property income Labor + property income Labor + property + private transfer income Labor + property + private transfer income Labor + property + total transfer income Labor + property + total transfer income Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: All income measures are expressed in per adult equivalent terms. of a particular income source in a way that disproportion- (figure 1.5). Income from public transfers is less unequally ately favors the most affluent households. distributed than income from private transfers, and the dispersion of both sources of income declined between 2001 First, the contribution of one particular income source and 2015. Labor income is not as unequally distributed as to total income inequality depends not only on the size income from property; yet, its dispersion rose substantially of the contribution relative to total income, but also on over time. the dispersion. Income from property is more unequally distributed than income from any other source, and, yet, Second, changes in the relative importance of each the associated Gini coefficient did not change over time income component affect inequality. For example, the Inequality in Mauritius: Stylized Facts 29 FIGURE 1.5. Gini Coefficient, by Income Source, 2001 and 2015 1.000 0.900 0.800 0.700 0.600 0.500 0.400 0.300 0.200 2001 0.100 2015 0.000 Labor income Property income Public transfers Private transfers Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. increase in the share of an income component that is automatically translate into a higher risk of experiencing extremely unequally distributed may alone lead to falling income because an effective public transfer system is widening inequality. Wages and salaries, together with in place that helps protect the most vulnerable households. income from self-employment activities, make up the The shares of income from property and of income from largest share of total household income (figure 1.6, private transfers declined slightly between 2001 and 2015, panel a). In 2001, 86 percent of total household income when they contributed 0.1 percent and 5.2 percent of total was income from labor. The share of labor income declined, income, respectively. By contrast, the richest 20 percent of and, in 2015, it contributed, on average, around 80 percent the total income distribution received an average of around to total household income, a relative reduction of almost 89 percent of their total income from labor in 2001 (com- 6 percentage points (figure 1.6, panel b). Thus, the most pared with 74.4 percent among the poorest 20 percent). unequally distributed income component lost its relative Also, among the richest households, the contribution of importance. labor income shrank. However, the reduction was smaller: –4.4 percentage points, compared with –9 percentage points Income from property contributes around 2 percent of between 2001 and 2015. All the remaining income compo- total household income. Its relative importance declined nents play a minor role in total household income among modestly, from 2.1 percent to 1.9 percent, between 2001 the richest households, with the exception of income and 2015. By contrast, the share of income from public from private transfers, which rose substantially, from and private transfers rose. Private transfers went up by 3.6 percent to 7.2 percent. 2.8 percentage points and reached 6.8 percent in 2015. Government transfers climbed even more, on average: Breaking down inequality in total household income the share of income from public transfers increased from into the four different sources, Figure 1.8, panel a illus- 7.8 percent to 11.1 percent. trates that income from labor contributed the most to inequality in total income. In 2001, around 92 per- These patterns and trends in the various components cent of inequality was attributable to labor income, fol- of total household income have been different across lowed by property income (4 percent), private transfers the distribution, and notably so at the bottom quintile (3 percent), and public transfers (1 percent). Similarly, in and at the top quintile of the total income distribution 2015, almost 88 percent of total inequality was ascrib- (figure 1.7). In 2001, the poorest 20 percent of the distri- able to inequality in labor income, 6.6 percent to private bution received an average of about 74 percent of total transfers, 4 percent to property income, and 1.4 per- household income from labor. This share had declined cent to public transfers. These findings are in line with to less than two-thirds by 2015. At the bottom of the results of the World Bank (2015a), which breaks down income distribution, unemployment and inactivity do not total household income inequality by income source 30 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 1.6. Total Household Income, Changes and Shares of Components, 2001–15 a. Composition, by year 100 7.8 7.6 7.2 7.7 7.3 8.4 8.8 9.8 7.9 8.6 8.8 9.4 9.0 11.1 90 4.1 4.9 5.1 5.6 5.4 6.3 5.8 5.9 6.2 6.2 6.4 6.5 2.1 1.3 1.4 1.4 1.4 6.8 1.2 1.3 1.5 1.1 2.0 6.8 2.4 2.2 1.9 80 1.9 70 60 Percent 50 86.0 86.2 86.3 85.3 85.9 84.6 84.0 84.2 84.2 81.0 82.9 82.0 82.6 80.1 40 30 20 10 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Property income Private transfers Public transfers b. Changes in composition, 2001–15 4 3.29 2.79 2 0 −0.18 Percent −2 −4 Labor income Property income Private transfers Public transfers −6 −5.90 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: All income measures are expressed as per adult equivalents. Inequality in Mauritius: Stylized Facts 31 FIGURE 1.7. Total Household Income, Changes and Shares of Components, by Quintile, 2001–15 a. Composition, bottom quintile, by year 100 90 19.9 19.7 20.2 20.6 21.2 22.5 22.3 22.0 23.2 25.3 26.4 27.2 29.5 29.5 80 5.5 6.3 6.7 0.2 7.5 7.9 6.0 6.9 7.6 0.3 0.7 7.1 0.2 0.3 0.2 0.3 6.9 6.5 70 0.2 0.3 6.4 0.3 0.3 7.2 5.2 0.2 0.1 0.2 60 Percent 50 40 74.4 73.8 72.4 71.7 70.6 69.5 71.3 70.5 70.0 67.6 66.8 66.2 65.2 63.1 30 20 10 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Property income Private transfers Public transfers b. Composition, top quintile, by year 100 3.2 2.9 3.2 2.8 3.1 3.6 3.4 3.4 5.2 3.3 3.4 3.8 3.3 4.7 3.6 4.8 5.1 5.3 4.7 6.0 5.9 5.8 6.1 5.9 5.5 5.9 7.2 90 3.7 2.2 2.3 2.5 2.5 6.7 2.3 2.4 2.6 1.6 3.6 4.1 3.4 4.5 3.4 80 70 60 Percent 50 89.1 89.9 89.6 89.0 90.1 88.3 88.3 88.4 88.9 87.1 87.7 83.6 86.2 84.8 40 30 20 10 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Labor income Property income Private transfers Public transfers (continued) 32 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 1.7. Total Household Income, Changes and Shares of Components, by Quintile, 2001–15 (continued) c. Changes in composition, bottom quintile 9.63 10 5 Percent 0 −0.08 −0.38 −5 Labor income Property income Private transfers Public transfers −10 −9.17 d. Changes in composition, top quintile 4 3.63 2 1.07 Percent 0 −0.33 −2 Labor income Property income Private transfers Public transfers −4 −4.36 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: All income measures are expressed in per adult equivalents. Inequality in Mauritius: Stylized Facts 33 FIGURE 1.8. Decomposition of Total Household Income Inequality, by Income Source, 2001–15 a. Decomposition of inequality levels, by year b. Decomposition of changes in inequality 4.0 2015 87.9 6.6 1.4 –28.8 98.3 10.2 20.3 3.0 2001 91.8 4.4 0.8 0 10 20 30 40 50 60 70 80 90 100 –40 –20 0 20 40 60 80 100 120 140 Percent Percent Labor income Private transfers Labor income Private transfers Property income Public transfers Property income Public transfers Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. using data from the 2007 and the 2012 Household as observed in Latin America (Azevedo, Inchauste, and Budget Surveys. Sanfelice 2013). A decomposition of the changes in total household income By contrast, government redistribution was key to mitigat- inequality between 2001 and 2015, as measured by the ing the increase in inequality in total household income, Gini index, formally corroborates the patterns illustrated particularly among households in the lower tail of the in figure 1.4: household labor income was the main cul- distribution. The social protection system is comprehensive prit behind rising inequality (figure 1.8, panel b). Labor and contributed considerably to reducing poverty and income contributed over 98 percent to the change in total containing inequality. Despite substantial steps taken by household income inequality, followed by private transfers the government to promote greater coordination in social and property income (20.3 and 10.2 percent, respectively), protection programs, improvements are needed (World whereas public transfers had an inequality-narrowing effect Bank 2015a). For example, the basic retirement pension, (–28.8 percent). a universal noncontributory social pension paid to persons above age 60, lacks a focus on the poor because it dispro- This chapter illustrates the main patterns of house- portionately favors well-off households, and funding is hold income inequality and its components over the low for programs specifically intended to benefit the poor last 15 years. While total household income inequality (World Bank 2015a). However, to capture fully the distri- expanded substantially, especially in the aftermath of butional effects of government redistribution through taxes the global economic downturn and terms-of-trade shock and transfers, a fiscal incidence analysis is required. This that hit Mauritius between 2008 and 2015, inequality in analysis consists of allocating taxes, particularly personal household labor income skyrocketed; the Gini index rose income tax and consumption taxes, and public spending, from 41.7 in 2001 to 50.0 in 2015. Although it is not the particularly social spending, to households or individuals most unequally distributed source of household income, and comparing incomes before taxes and transfers with dispersion in household labor income alone explains incomes after taxes and transfers. (The issue of missing most of the rising inequality in total household income top incomes is described in box 1.1.) BOX 1.1. Do Household Surveys Underestimate Inequality? The Challenge of Top Incomes The literature on income inequality is torn regarding the sources of data that one should use in the analysis of trends and levels. A large number of studies, ranging from official statistics to academic papers, have made use of income measures from household survey data and constructed several inequality indicators. On the other hand, the literature on top incomes has used administrative record data on personal income tax returns that report estimates of the shares of total income retained by the richest portion of the population. Estimates in studies reflecting the views of the two groups have recently been at odds. Estimates from tax return data show a larger expansion in equality over the last two decades in the United Kingdom and the United States. The divergence in the findings might be partly derived from the different inequality indices and income definitions. Yet, the most important source of discrepancy lies in the fact that household surveys do not do a good job in capturing top incomes, that is, the incomes of the top 1.0 or 0.1 percent of the population. Jenkins (2016) explains that the undercoverage of top incomes in survey data is attributable to multiple factors. One is underreport- ing among high-income respondents. (One might think of this as the survey counterpart of tax evasion.) An aspect of this might be top-coding applied by survey administrators to limit the effects of measurement error on aggregates. This possible cause of un- dercoverage implies that data are right-censored. A second source of undercoverage is the sampling of high-income respondents. The respondent community may provide sparse coverage of the top income ranges, the survey might not include any high-income respondents by design, or the survey team may be unable to reach top earners, who may also refuse to participate. (This can be thought as a second survey counterpart of tax evasion.) This additional source of undercoverage translates into a right-truncation of a true distribution. Regardless of the cause, undercoverage introduces a downward bias in survey estimates of inequality for a given year because there is insufficient income observed in the top income ranges. While the top 0.1 percent or 1.0 percent of the income distribution is barely identifiable on the horizontal axis of a Lorenz curve plot, Atkinson (2007, 19–20) points out that, “if we treat the very top group as infinitesimal in numbers, but with a finite share, S*, of total income, then the Gini coefficient can be approximated by S* + (1 – S*)G, where G is the Gini coefficient for the rest of the population.” A number of approaches have been used to estimate inequality, while addressing undercoverage problems. They include tech- niques entirely based on survey data and deriving an inequality estimate from the richest households by fitting a Pareto type I distribution to the top income observations in the survey data and then estimating total inequality as the sum of inequality within the top group, within the non-top group, and between-group inequality. Others use tax return data and replace the highest income in the survey with cell-mean imputations based on corresponding observations from tax data, or they combine estimates from the two data sources instead of combining data. The latter is the most promising avenue for overcoming the undercoverage issue of survey data. Recently, a team effort led by Anthony Atkinson and Thomas Piketty produced the World Wealth and Income Database, which pro- vides estimates of top income and wealth shares worldwide by combining fiscal, survey, and national accounts data. In Mauritius, the estimated share of fiscal income held by the top 1 percent increased from around 5 percent in 2001 to over 7 percent in 2011 (figure B1.1.1). In comparison with highly unequal countries such as South Africa and the United States, the share is still relatively low. Yet, the magnitude of the change observed over the last 10 years is comparable. This study does not attempt to combine esti- mates from CMPHS data and tax return data, which were not available to the authors at the time of writing, to derive a more precise measure of income inequality. However, for the reasons illustrated so far, it is important to acknowledge that estimates presented in this study are likely to be a lower bound of the true level of income inequality in Mauritius. FIGURE B1.1.1. Top 1 Percent of the Fiscal Income Share, 1999–2011 Top 1% scal income share, Mauritius, 1999–2011 8 7 Share of total (%) 6 5 4 2000 2002 2004 2006 2008 2010 Fiscal income | Top 1% | share | adults | tax unit Graph provided by www.wid.world Source: Data of WID (World Wealth and Income Database), Paris School of Economics, Paris, http://www.parisschoolofeconomics.eu/en/research/ the-world-wealth-income-database/. Inequality in Mauritius: Stylized Facts 35 NOTES 3. Private pensions make up the largest component of private transfers among the richest households: in 2015 over 90 percent of private transfers were private pensions among households in the richest 1. To characterize fully the trends in income dispersion, a set of different 20 percent of the income distribution. Among the poorest households, metrics are used because measures of inequality that emphasize the besides income derived from private pensions (about 40 percent lower tail of the distribution can evolve differently from measures that of total private transfers), regular allowances from parents and/or zoom in on the upper tail of the distribution. relative were the largest component, followed by other sources of 2. The Gini coefficient is a dispersion metric more sensitive to transfers at regular income, alimonies, and regular allowance from social/religious the center of the distribution than to transfers occurring at the tails. organizations.  37 CHAPTER 2 Drivers of Growing Inequality in Household Labor Income C Trends in Household 2.1  hapter 1 illustrates the main patterns of inequality in total household income and its components. Household labor income is, on average, the largest Demographics component of total household income: in 2015, it con- tributed, on average, some 80 percent to total household The typical Mauritian household is a married-couple house- income. Despite not being the most unequally distributed hold, that is, composed of a husband who is virtually always source of household income, dispersion in household the household head in these households and a wife (spouse) labor income increased sizably between 2001 and 2015 with or without children and other family members (fig- and is the income source that alone explains most of ure 2.1, panel a). In 2001, about 84 percent of households the observed increase in inequality in total household were married-couple households. Their share declined income. over the last 15 years and reached 80 percent in 2015. The share of married-couple households with children declined This chapter focuses on the drivers of expanding inequality from 61 percent in 2001 to 46 percent in 2015, whereas in equivalized household labor income. First, equivalized married-couple households with no children increased their measures incorporate household size and composition, and relative importance from 23 percent to 35 percent over the this implies that any change in household size or composi- same period. Single-headed households are today more tion is mechanically reflected in the measure, along with common than in the past (20 percent of all households in any related measures of the consequence on inequality. 2015). These trends are the by-product of two main phe- Second, household demographics have an indirect effect on nomena: (1) family formation decisions are increasingly equivalized household welfare measures because household delayed, and this leads to a delay in the birth of children, and demographic structure, household mix (that is, the distri- (2) population aging is accompanied by an increase in the bution of different types of households), and household share of single-headed households, typically headed by characteristics impact the distribution of income within and women (75 percent of these households) in their late 50s across households. Moreover, labor market factors affect or early 60s and without coresident children. inequality in household labor income. Following a strand of research including Burtless (1999), Daly and Valletta The changes observed in the patterns of household types (2006), Devereux (2004), Fortin and Schirle (2006), Juhn were to some extent different across quintiles of total and Murphy (1997), and Pencavel (2006), the analysis household income. The increase in the share of single- assumes that inequality in household labor income is headed households, with and without children, was affected by two main groups of factors: (1) demographics, remarkable among the poorest households. In the bottom including household composition, household mix, household quintile, they increased by 15 percentage points, whereas characteristics, and assortative mating, that is, the degree to in the top quintile their share did not change. In parallel, which individuals marry within their own income group, married-couple households without children increased and (2) labor market factors, including dispersion of labor their weight particularly among the richest households income among men and women and female labor force from 34 to 47 percent (figure 2.1, panel c). participation. The chapter first describes changes observed in household demographics and labor market factors; it The average household size declined across family types: then assesses the relative contribution of these factors to in 2001, the average household was composed of 4.5 mem- changes in household labor income inequality. bers; in 2015 this number was down to 4.2. The reduction FIGURE 2.1. Household Size and Composition, 2001–15 a. By type and year 6 5.6 5.3 5 5.2 5.2 5.2 5.1 5.1 5.1 5.0 5.0 5.0 5.0 4.9 4.9 4.8 4.9 4.9 4.8 4.9 4.9 4.9 4.8 4.8 4.8 4.8 4.8 4.8 4.8 4.7 4.8 4 Average household size 3.9 3.9 3.8 3.7 3.7 3.7 3.7 3.6 3.6 3.6 3.6 3.6 3.6 3.6 3.5 3 3.0 2.9 2.9 2.8 2.7 2.7 2.7 2.6 2.6 2.6 2.6 2.6 2.6 2.6 2.6 2 1 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Married couples with children Single householders with children Married couples without children Single householders without children b. Distribution, by type and year 100 9 9 9 10 10 10 10 10 11 11 11 11 11 12 12 7 7 8 7 7 7 7 7 7 7 8 8 8 8 8 80 23 24 22 24 24 26 27 27 27 29 30 31 32 33 35 60 Percent 40 61 60 61 59 60 57 56 56 55 54 52 50 49 47 46 20 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Married couples with children Single householders with children Married couples without children Single householders without children (continued) Drivers of Growing Inequality in Household Labor Income  39 FIGURE 2.1. Household Size and Composition, 2001–15 (continued) c. Distribution, by type and quintile 100 9 9 11 8 8 10 9 11 9 11 9 13 13 13 12 9 7 8 4 4 4 9 10 7 6 6 12 8 8 8 80 13 16 22 16 23 26 30 34 19 40 29 32 47 60 37 41 Percent 40 68 65 68 63 61 58 57 54 50 53 48 45 20 41 41 39 0 01 08 15 01 08 15 15 01 08 15 01 08 15 01 08 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 1 2 3 4 5 Quintile of household total income Married couples with children Single householders with children Married couples without children Single householders without children Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. in size of married-couple households with children from rising fraction of married-couple households have at least 6.0 to 4.6 members, on average, was considerable and two workers, and this reduces dispersion to the extent that largely ascribable to the decline in the number of children labor income is not perfectly correlated across spouses (figure 2.1, panel b). Married-couple households without (figure 2.2, panel b). children recorded a modest decline in size from 3.9 to 3.6  members between 2001 and 2015. By contrast, the To study the extent of assortative mating, that is, the increas- change in size of single-headed households with and without ing correlation in terms of educational, occupational, or children was negligible. income characteristics among spouses, the analysis looks at the degree of correlations of labor income between The increase observed in the share of single-headed house- spouses among double earners couples.1 Following Fortin holds and the parallel decline in the share of married-couple and Schirle (2006), assortative mating is defined by the households are both relevant in terms of household income likelihood of a person in labor income decile i to be mar- labor inequality. First, family plays a role in providing ried to a spouse in the same labor income decile, according insurance against individual risk, a role that increased to their respective labor income distribution. Table 2.1 over time thanks to the rapid expansion in female labor shows the percentage of double earners couples sorted by market participation, which will be discussed extensively the husband’s and the wife’s labor income deciles in 2001 in the second part of the chapter. Therefore, an increase in and 2015. The degree of assortative mating is captured the percentage of single-headed households, which show a by the percentage of couples along the main diagonal: an higher dispersion in labor income, contributes to the over- increase in the sum of these shares implies an increase in all rise in income inequality (figure 2.2, panel a). Second, assortative mating. For example, in 2001 relative to 2015, because of increased female labor force participation, a wives in the bottom two and in the top three deciles were 40 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 2.2. Double Earners and Dispersion of Household Labor Income, by Family Type, 2001–15 a. Variance of log household labor income, b. Double-earner households in married-couple by family type households 1 38 37 .9 36 .8 35 Log points 34 Percent .7 33 .6 32 31 .5 30 .4 29 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 28 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Married households labor income Single households labor income Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. TABLE 2.1. Assortative Mating: Double-Earner Couples, by the Husband and Wife’s Labor Income Decile, 2001 and 2015 Spouse labor income decile 2001 1 2 3 4 5 6 7 8 9 10 Total 1 2.7 1.5 2.1 1.9 0.8 0.5 0.6 0.3 0.2 0.4 10.8 2 1.1 1.9 1.6 1.3 1.4 1.9 0.5 0.2 0.2 0.0 10.0 Head labor income decile 3 1.6 1.8 1.7 1.8 1.6 1.3 0.9 0.4 0.1 0.2 11.3 4 1.6 1.8 1.6 2.0 1.4 1.5 1.2 0.7 0.4 0.0 12.1 5 0.9 0.3 0.6 1.5 0.6 1.0 0.8 0.6 0.2 0.2 6.6 6 1.2 2.2 1.2 2.1 0.7 1.4 1.3 0.7 0.5 0.3 11.6 7 0.8 0.7 0.7 1.5 0.6 1.3 1.8 1.3 1.9 0.4 10.9 8 0.2 0.4 0.2 0.4 0.4 0.7 1.1 1.3 1.8 0.8 7.3 9 0.3 0.4 0.1 0.4 0.4 0.7 0.9 1.7 2.2 3.1 10.1 10 0.0 0.1 0.1 0.3 0.1 0.4 0.6 1.1 2.3 4.3 9.3 Total 10.5 11.1 9.7 13.2 8.0 10.8 9.6 8.2 9.7 9.4 100 Spouse labor income decile 2015 1 2 3 4 5 6 7 8 9 10 Total 1 3.2 2.4 1.4 1.3 0.8 0.8 0.6 0.6 0.2 0.1 11.3 2 2.2 2.1 1.7 2.0 2.0 0.8 0.6 0.5 0.3 0.0 12.2 Head labor income decile 3 1.0 1.4 0.6 1.3 1.2 0.8 0.4 0.3 0.2 0.0 7.2 4 1.7 1.9 1.3 1.8 1.9 1.4 1.0 0.7 0.3 0.1 12.2 5 0.9 1.2 1.2 1.4 1.7 1.2 1.2 0.4 0.3 0.3 9.8 6 1.0 0.9 0.7 1.0 0.9 0.7 1.0 0.8 0.7 0.2 7.9 7 0.5 0.7 0.7 0.6 1.4 0.9 1.3 2.3 1.1 1.9 10.4 8 0.6 0.8 0.6 0.5 0.8 0.6 1.5 1.9 1.7 1.1 10.1 9 0.4 0.2 0.2 0.5 0.6 0.6 0.9 1.8 2.6 2.0 9.7 10 0.2 0.0 0.1 0.2 0.3 0.4 0.3 0.8 2.7 4.1 9.3 Total 11.6 11.6 8.5 10.7 11.5 8.2 8.9 10.1 10.0 8.9 100 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Drivers of Growing Inequality in Household Labor Income  41 less likely to be married to husbands with labor income income of the household. According to figures for the in their same deciles. Because assortative mating increased most recent year, heads and spouses’ labor income con- over time, the positive effect of increased female labor force tributed 100 percent of total household labor income participation might be mitigated or even reversed by the in about 4 households in 10 as opposed to 16 percent higher between-spouse correlation of labor income and of the households where labor income contributed between therefore contribute to the expansion in household labor 99 and 50 percent, and 18 percent of households where income inequality. the contribution was less than 50 percent. Some 23 percent of households did not have any labor income because the head and spouse were not employed. Trends in Labor 2.2  Market Factors In the case of single-headed households, which are typi- cally headed by women, fewer than one head in two is inactive (figure 2.4, panel a). This is partly explained by The second set of factors that might have a large impact the age distribution of single-household heads. In 2001, on household labor income inequality involves the labor about 40 percent of them were above 60 years of age; this market, more precisely, dispersion in men and women’s share had increased to 50 percent by 2015. This implies labor income and female labor force participation. This that half of them are likely to be retirees or beneficiaries of subsection start with a description of trends in the labor public transfers. Among heads of working age (16–64) or force status of household members separately for married- below 60 years of age, about one in four in 2015 was inactive couple and single-headed households. It then briefly and this has not changed since 2001. About 65 percent of describes trends in female labor force participation, which coresident children are employed, with a modest increase are analyzed in the second part of the chapter, and earnings over time (figure 2.4, panel b). The remaining household dispersion of men and women, which are studied in detail members, son or daughter-in-law, coresident parents, and in chapters 3 and 4. grandchildren, are largely inactive (54 percent) (figure 2.4, panel c). The head of married-couple households is virtually always a man. About 80 percent of men heads of married-couple One important change in the labor market that might affect households are employed, and this percentage modestly household labor inequality was the considerable increase declined over time (figure 2.3, panel a). Their wives, instead, in female labor force participation. Between 2001 and do not participate by and large in the labor market. This 2015, Mauritian women’s participation increased from is one of the pivotal issues that characterize the Mauritian about 43 percent to 57 percent. These changes are largely labor market and that is likely to be partly ascribable to ascribable to young and highly educated Mauritian women the role that women have traditionally had in Mauritian (figure 2.5). In terms of family type, the largest expansion society. Married women carry most of the family burden, was observed among married women, whose participation, including child and elderly care and home management, however, still lags behind that of single women, by about and this housework competes for women’s time and energy 7 percentage points. with work on the market. While the trend has generally been in the right direction, there is still a long way to go. In addition to rising female labor force participation, In 2001, fewer than 40 percent of wives participated in women also had large gains in labor income relative to men. the labor market; by 2015 this share had reached close Figure 2.6 illustrates the considerable gains in women’s to 50 percent (figure 2.3, panel b). As to the remaining earnings, both monthly and hourly, between 2001 and household members, about half of coresident children are 2015. The 2015 monthly and hourly earnings distribution employed, while the rest are inactive (and in education) or among women was shifted to the right relative to the 2001 unemployed (10 percent). The rest of household members distribution, whereas, in the case of men, the rightward are prevalently inactive, and no significant changes occurred shift was smaller. over time to their labor market status. Thanks to increases in women’s earnings, women’s mean In the majority of households, the head and spouse’s hourly earnings partially gained on men’s earnings. The income from labor contributes to virtually all the labor woman/man mean log hourly earnings ratio fell from FIGURE 2.3. Labor Force Status of Household Members of Married-Couple Households, 2001–15 a. Head 100 15.3 16.1 13.6 16.1 16.2 16.7 17.0 17.8 17.3 17.7 18.9 20.0 20.4 20.9 21.6 0.7 1.5 1.3 1.0 1.2 1.3 1.1 1.1 1.1 0.8 80 1.2 1.2 1.2 1.5 1.0 60 Percent 83.2 85.6 83.0 82.6 82.6 82.0 82.0 81.1 81.6 81.5 80.0 40 78.8 78.4 77.6 77.4 20 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF b. Spouse 100 80 53.4 53.4 51.2 51.8 50.8 55.9 55.3 54.1 53.8 56.7 59.2 56.7 57.1 62.6 60.5 60 Percent 4.1 3.5 3.0 5.0 5.1 4.4 3.8 3.1 6.6 5.1 40 5.9 5.1 3.0 4.1 2.8 44.7 44.7 46.2 40.2 40.9 41.5 41.8 42.8 20 37.5 37.5 37.8 39.6 34.6 36.5 36.7 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF (continued) Drivers of Growing Inequality in Household Labor Income  43 FIGURE 2.3. Labor Force Status of Household Members of Married-Couple Households, 2001–15 (continued) c. Children 100 26.8 28.6 28.0 28.7 31.5 32.0 33.9 34.3 33.7 33.1 33.0 31.5 35.8 37.8 36.0 80 12.0 10.8 10.7 10.2 9.9 8.9 7.8 9.0 9.1 9.2 8.3 8.8 8.8 9.2 60 8.5 9.1 13.7 11.6 10.4 9.2 9.9 9.4 Percent 10.5 9.3 6.5 7.7 8.5 8.5 8.1 7.6 40 52.8 50.7 49.9 50.8 47.5 49.0 48.1 47.8 48.8 47.2 48.9 48.8 49.3 49.3 45.5 20 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF not in education Non-LF currently in education d. Other members 100 80 65.1 66.0 64.9 63.9 63.1 64.3 67.1 66.9 69.0 71.9 70.2 70.4 71.0 70.2 73.5 60 Percent 40 2.9 4.1 4.1 4.5 3.8 4.2 3.0 5.2 4.1 5.2 3.9 3.4 2.3 3.8 2.5 20 30.4 30.9 33.2 32.8 31.6 29.9 27.9 27.5 30.2 24.2 24.0 25.8 25.9 25.7 25.6 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 44 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 2.4. Labor Force Status of Household Members of Single-Headed Households, 2001–15 a. Head 100 80 49.3 50.6 52.8 54.2 52.3 53.2 53.4 55.2 54.7 54.4 52.9 55.7 58.2 56.4 55.7 60 Percent 1.4 3.0 1.4 3.2 2.6 2.8 1.8 1.9 2.4 2.3 2.5 2.2 2.1 2.5 40 2.4 49.3 46.4 45.7 43.9 44.5 44.2 43.8 43.0 43.1 45.3 20 42.4 42.2 41.4 41.9 39.4 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF b. Children 100 8.5 9.3 8.3 9.0 9.8 9.4 10.7 10.8 10.6 10.6 9.6 10.7 10.8 10.0 12.1 17.1 15.5 16.2 15.3 17.3 13.7 16.7 14.5 14.4 17.6 16.0 14.8 13.0 16.5 14.6 80 7.2 6.9 7.9 8.2 7.1 6.5 9.5 9.9 8.5 7.8 12.1 10.1 10.3 8.9 7.9 60 Percent 40 68.3 66.0 68.0 67.8 65.9 66.7 67.5 69.1 65.7 65.7 62.5 63.0 64.5 65.0 62.3 20 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF not in education Non-LF currently in education (continued) Drivers of Growing Inequality in Household Labor Income  45 FIGURE 2.4. Labor Force Status of Household Members of Single-Headed Households, 2001–15 (continued) c. Other members 100 80 53.2 53.3 54.0 52.9 52.8 53.1 51.1 55.8 54.5 55.8 56.7 55.6 58.1 59.5 61.3 60 Percent 4.7 4.3 4.3 6.5 7.4 5.5 6.2 5.6 6.0 5.4 4.4 5.8 40 5.6 4.0 6.0 41.9 40.6 40.9 42.9 42.6 42.5 20 38.6 39.6 39.4 38.8 38.9 38.6 34.7 36.3 34.5 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Employed Unemployed Non-LF Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 2.5. Female Participation Rate, by Age-Group and Family Type, 2001–15 a. By age-group b. By family type 80 65 70 60 60 55 Percent Percent 50 50 40 45 30 40 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 30−44 Married w/child Nonmarried w/child 25−29 45−64 Married no child Nonmarried no child Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 46 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 2.6. Monthly and Hourly Labor Income, by Gender, 2001 and 2015 a. Density of women’s log hourly earning b. Density of women’s log monthly earning .8 .8 .6 .6 Density Density .4 .4 .2 .2 0 0 2 4 6 8 10 2 4 6 8 10 2001 2015 2001 2015 c. Density of men’s log hourly earning d. Density of men’s log monthly earning .8 .8 .6 .6 Density Density .4 .4 .2 .2 0 0 2 4 6 8 10 2 4 6 8 10 2001 2015 2001 2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. -34 percent in 2001 to -26 percent in 2015 (figure 2.7, panel a). Women continue to work fewer hours than men, 2.3 Explaining Changes and the overall decline in hours worked was considerably in Equivalent Household larger among women compared with men (figure 2.7, panel b). Labor Income Inequality The substantial gains among working women were accom- Figure 2.9 displays trends in selected percentiles of house- panied by rapid increases in labor income inequality. Fig- hold labor income by family type: married couples with ure 2.8 shows trends in three indicators of inequality that children, married couples without children, single-headed are sensitive to different shifts in the earnings distribution households with children, and single-headed household (P90/P10, P50/P10, and P90/P50). For both men and women, without children. Overall, despite an initial period of stagna- labor income inequality is observed to rise rapidly in the tion until 2004, median household labor income increased upper tail, particularly among women, whereas inequality by 43 percent between 2001 and 2015. For married-couple at the bottom changed modestly. Given these patterns in households, median labor income grew by 34 percent women’s labor income inequality, rising female participation and 37 percent (households without and with children, might, ex ante, have either a moderating or exacerbating respectively), whereas single-headed households with no effect on household labor income inequality. children experienced a median increase of 41.5 percent, and Drivers of Growing Inequality in Household Labor Income  47 FIGURE 2.7. The Gender Gap in Hourly Earnings and Trends in Hours Worked, by Gender and Education, 2001–15 a. Woman/man log hourly earnings ratio, by year b. Average hours worked per week −24 Female Male −25 45 45 −26 −27 −28 40 40 Percent −29 −30 Hours −31 35 35 −32 −33 −34 30 30 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No edu up to complete primary Lower secondary Upper secondary/post-secondary/tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. those with children had a 24 percent gain. The expansion gains in percentage terms compared with the poorest ones. of inequality took place largely during the second half of Married-couple households with children and single-headed the period (2007–15), and this pattern has been observed households with no children at the bottom of the distribu- to a different extent across all family types. tion saw their labor income declining by 11.0 percent and 2.4 percent, respectively, between 2001 and 2015. The gains Overall, the percentage change observed in labor income in labor income were extraordinarily large among families is monotonically increasing as one moves up along the dis- above the median beginning in 2007, particularly among tribution. This means the richest households posted larger households in the 90th and 95th percentiles (+60 percent and FIGURE 2.8. Trends in Individual Labor Income Inequality, by Gender, 2001–15 a. Women b. Men 160 160 140 140 120 120 100 100 Percent Percent 80 80 60 60 40 40 20 20 0 0 −20 −20 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 P50/P10 P90/P50 P90/P10 P50/P10 P90/P50 P90/P10 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 48 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 2.9. Trends in Selected Percentiles of Household Labor Income by Family Type, 2001–15 a. Married couples with children b. Married couples with no children 20,000 20,000 15,000 15,000 2015 Rupees 2015 Rupees 10,000 10,000 5,000 5,000 0 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p5 p25 p75 p95 p5 p25 p75 p95 p10 p50 p90 p10 p50 p90 c. Single-headed households with children d. Single-headed households with no children 20,000 20,000 15,000 15,000 2015 Rupees 2015 Rupees 10,000 10,000 5,000 5,000 0 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p5 p25 p75 p95 p5 p25 p75 p95 p10 p50 p90 p10 p50 p90 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. +65 percent growth between 2007 and 2015). By contrast, mating, household mix, and household characteristics—to the poorest households (5th and 10th percentiles) recorded a changes in the density of log equivalent household labor modest increase in their labor income, about 0.3 percent and income.2 The contribution of each factor is evaluated by 10.9 percent, respectively. As a result, inequality measured comparing the unadjusted density of 2015 log equivalent by the P90/P10 ratio rose considerably across all family household labor income with the counterfactual obtained types and, notably, among married-couple households with by holding sequentially each factor at the 2001 level (fig- children and single-headed households with no children. The ure 2.10, panel h). P90/P50 ratio expanded considerably, too, whereas inequality at the bottom of the distribution measured by the P50/P10 Figure 2.10 displays the estimated counterfactual densities ratio did not grow as much. This emerges clearly through that are the result of the decomposition exercise. The the time trends of the 5th, 10th, 25th, and 50th percentiles original (unadjusted) density of log equivalent household that closely mirror over the entire period (see figure 2.9). labor income is presented as a solid line in panel a. The counterfactual (adjusted) densities are shown as a dashed A decomposition method is used to assess the relative con- line, which becomes the solid line to be adjusted in the tribution of six factors—men’s labor income, female labor next panel. For example, panel a shows the original 2015 force participation, women’s labor income, assortative density (solid line) and the 2015 density adjusted for men’s Drivers of Growing Inequality in Household Labor Income  49 FIGURE 2.10. Step-Wise Decomposition of Household Labor Income, 2001 and 2015 a. Effect of men’s labor income b. Effect of female participation .8 .8 .6 .6 Density Density .4 .4 .2 .2 0 0 4 6 8 10 12 4 6 8 10 12 Log PAE HH labor income − 2015 Log PAE HH labor income − 2015 Unadjusted Male labor income Male labor income Female LFP c. Effect of women’s labor income d. Effect of assortative mating .8 .8 .6 .6 Density Density .4 .4 .2 .2 0 0 4 6 8 10 12 4 6 8 10 12 Log PAE HH labor income − 2015 Log PAE HH labor income − 2015 Female LFP Female labor income Female labor income Assortative mating e. Effect of family mix f. Effect of family characteristics .8 .8 .6 .6 Density Density .4 .4 .2 .2 0 0 4 6 8 10 12 4 6 8 10 12 Log PAE HH labor income − 2015 Log PAE HH labor income − 2015 Assortative mating Family mix Family mix Family characteristics (continued) 50 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 2.10. Step-Wise Decomposition of Household Labor Income, 2001 and 2015 (continued) g. Original and adjusted distribution h. Demographic and labor market factors in inequality .8 1.0 6.8 .6 P90/P10 6.0 Density .4 –28.3 113.9 .2 0.3 0.2 0 –40 –20 0 20 40 60 80 100 120 140 4 6 8 10 12 Log PAE HH labor income − 2001 Percent Log PAE HH labor income − 2015 Men's labor income inequality Log PAE HH labor income − 2015 fully adjusted Women's participation Women's labor income inequality Assortative mating Family mix Family characteristics Residual Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. labor income (dashed line). Panel b displays the 2015 density modestly lower (figure 2.10, panel b). While this factor (solid line), which is the one adjusted in panel a, and the might appear counterintuitive, a careful look at changes adjusted 2015 density adjusted for men’s labor income and in female labor force participation by family type and by female labor force participation (dashed line). quintile of household labor income reveals the conundrum. The rise in female participation was not constant along Holding the structure of men’s labor income at the 2001 the distribution of household labor income (figure 2.11). level leads to fewer families at the tails and more families In the case of married-couple households, which make at the center (around the mode) of the distribution. This a up the largest share of households in Mauritius, female consequence of the expansion in men’s wage inequality (see participation remained roughly stable in the lowest quintile, chapter 4). This is largely attributable to the differential grew by 17 percentage points in the second quintile, by trends in the skilled labor demand and skilled labor supply. 11 percentage points in the third quintile, by 18 percentage Overall, changes in the structure of men’s labor income points in the fourth quintile, and by 13 percentage points can explain about 140 percent of the increase in the Gini in the top quintile. Among single-headed households, coefficient of equivalized household labor income and which are largely headed by women, the participation of about 165 and 95 percent of the rise in upper and lower heads declined across all quintiles except for the bottom, tail inequality, respectively. This is in line with the location where it modestly increased. Had women’s participation of the bulk of the expansion in men’s wage inequality in been more evenly shared among households along the the upper tail (chapter 3). labor income distribution, inequality could have declined. In parallel and for single-headed households, the receipt of The increase in female labor force participation contrib- substantial (mostly public) transfers might play against uted to increasing inequality among families between labor market participation. 2001 and 2015. By adjusting the 2015 density to hold female labor force participation at the level observed in Meanwhile, changes in the structure of women’s labor income 2001, one sees that inequality measures generally decline. explain a sizable portion of the increase in household labor Thus, had women not entered in increasing proportions income inequality between 2001 and 2015. Had women’s into the labor force, inequality in 2015 would have been labor income inequality remained as it was in 2001, the Drivers of Growing Inequality in Household Labor Income  51 FIGURE 2.11. Labor Market Status of Spouses, by Quintile of Household Labor Income, 2001–15 100 80 39.1 37.1 46.9 50.4 49.2 58.5 59.1 54.9 56.6 54.6 62.6 61.9 69.3 66.4 74.7 0.8 60 0.9 Percent 0.9 2.3 1.8 3.9 2.1 40 5.0 6.4 7.1 1.9 1.8 3.8 60.0 62.2 9.2 52.2 5.2 47.3 48.9 41.2 43.3 20 35.0 33.8 35.6 38.4 36.1 29.8 20.1 21.5 0 01 08 15 01 08 15 01 08 15 01 08 15 01 08 15 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 1 2 3 4 5 Quintile of household total income Employed Unemployed Non-LF Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. change in overall inequality (P90/P10) would have been Changes in family mix account for shifts in the distribution 7 percent lower, and the change in upper-tail inequality of family types: married-couple households with and without (P90/P50) would have been around 36 percent lower than children and single-headed households with and without the change that was actually observed. children. Between 2001 and 2015, there was an increase in the share of single-headed households. This is reflected in Assortative mating measures the extent of labor income figure 2.10, panel e, which shows that families moved from correlations among husbands and wives.3 Rising cor- the middle and upper part of the distribution toward the relation is expected to increase labor income inequality lower end in the adjusted density plot. This translated into across households. Between 2001 and 2015, the percent- an increase in inequality compared with what would have age of married-couple households with children in which been observed had the family mix stayed constant at the husbands and wives fall in the same decile increased by 2001 level. Changes in family mix accounted for 1.3 percent about 2 percentage points, while, among corresponding of the changes in the Gini coefficient. households with no children, the rise was almost 3 points. The fact that the likelihood that husbands and wives had Figure 2.10, panel f displays the density of log household somewhat more similar income in 2015 compared with labor income that accounted for changes in household 2001 implied a movement of some married households characteristics. Between 2001 and 2015, men and women from the middle to the upper tail of the distribution. This heads or spouses grew older and were increasingly more is reflected in the slight decline (-2.6 percent) in upper- well educated, on average. Therefore, applying 2001 tail inequality if assortative mating had stayed constant household characteristics to 2015 household labor income at the 2001 level. would shift the distribution to the left, generating an increase 52 Mauritius: Addressing Inequality through More Equitable Labor Markets in overall inequality, particularly in the lower tail. In terms the density of the counterfactual log equivalent household of inequality, family characteristics accounted for 6 percent labor income, is much closer and has a shape more similar of the change in the P90/P10 ratio and for a remarkable to the observed 2001 density (orange solid line) compared 14 percent of the change in lower-tail inequality (P50/P10) with the observed 2015 density (solid blue line).4 Increas- between 2001 and 2015. ing inequality in men’s labor income was certainly the major contributor to expanding inequality in equivalized These findings are in line with some exceptions in the obser- household labor income. However, other factors, including vations on 23 countries of the Organisation for Economic the disproportionate increase in labor force participation Co-operation and Development (OECD) from the mid-1980s among women in the most affluent households, the relative to the mid-2000s (OECD 2011). First, the increase in men’s expansion of single-headed households, and inequality in earnings inequality is the main factor driving household women’s labor income also played a role. labor income inequality. Second, the contribution of assorta- tive mating and household structural changes to increased household labor income inequality was positive, but much NOTES more modest. Third, unlike in the case of Mauritius, the 1. The extent of assortative mating observed in labor income might reflect increase in women’s employment had an equalizing effect a general pattern of educational (or occupational) sorting. in all countries. 2. The approach consists of a conditional reweighting decomposition procedure introduced by DiNardo, Fortin, and Lemiuex (1996) and applied to the case of household earnings by many scholars, including All the demographic and labor market changes that occurred Daly and Valletta (2010) and Fortin and Schirle (2006). The focus of between 2001 and 2015 explain a significant portion of the decomposition exercise is equivalent household labor income. In the implementation of the analysis, the household sample was restricted the shift in the distribution of log equivalent household to households with nonzero income from labor and with household labor income, particularly in the lower tail. Figure 2.10 heads (and spouses in case of married-couple households) of working age (16–64) and employed with nonzero individual labor income and clearly illustrates that the changes observed in the six working hours or unemployed/inactive. For more on the decomposition, components explain a considerable part of the shift and refer to Fortin and Schirle (2006) and the Introduction, table I.1. 3. The reweighing function is set to 1 for single individuals and for couples the change in the shape of the distribution and therefore in with only one working spouse. inequality measures. The dashed blue line, which represents 4. The reverse order decomposition leads to similar results.  53 CHAPTER 3 The Role of Gender Inequality C hapter 2 looks at the role of household demographics The chapter starts with a description of the main stylized and labor market factors in rising household labor facts concerning women’s role in the labor market. Mauritian income inequality and identifies individual earn- women appear to be disadvantaged in terms of access to ings as a major contributor.However, additional factors the labor market according to labor market participation turned out to be a considerable source of rising inequality: statistics. However, there are important caveats as young the more rapid increase in labor market participation of generations are increasingly gaining space, particularly the the women living in the most affluent households and the most well educated. The second part of the chapter takes growing degree of assortative mating. Nonetheless, despite a deep dive into the issue of gender pay gaps by going the rise in female participation, women still suffer from beyond the description of average wage differentials and relatively low labor force participation, particularly the comparing the wages of men and women with similar least educated and older cohorts of women, and, in the characteristics in a multivariate framework. A decomposi- private sector, women also appear to be paid less compared tion exercise helps separate gender pay differences into a with men. component that is ascribable to different endowments of working men and working women and a component that This chapter takes a deep dive into the issue of gender gaps is a combination of discrimination effects and unobserved in labor market access and outcomes. Gender equality in characteristics. the labor market is important on equity grounds, and it is also smart in economic terms: it can enhance economic efficiency and productivity (World Bank 2011). Gender 3.1 Women’s Labor equality in the labor market has implications in terms of household income. The growth of married women’s Market Participation participation and the rise in women’s earnings mean that a larger number of wives are contributing to an increasing On average, labor market participation in Mauritius has share of family labor income. The simple expansion in been around 70 percent over the last decade, a figure in the number of employed wives could reduce household line with participation rates among OECD countries labor income inequality to the extent that such a change is (71.3 percent in 2015). Yet, women are severely disadvan- evenly distributed across households. Yet, rising earnings taged in access in the Mauritian labor market. Restricting inequality among individuals and an increasing degree of the sample to individuals of working age (16–64) who are assortative mating can also pull in the opposite direction. currently not in education, figure 31, panel a shows that only 47 percent of working-age women were active in Female participation in the labor force is crucial to the the labor market in 2004. Female participation increased functioning of labor markets for both efficiency and steadily over the past decade to 57 percent in 2015. Despite equity reasons. Unleashing additional and valuable human the progress, a stark disparity continues to persist between capital resources into the economy contributes to making men and women. The average participation gender gap in the economy more productive, thus helping it attain its full 2015 was still at a staggering 32 percentage points notwith- potential. As higher female labor market participation leads standing a significant narrowing by about 12 percentage to more members working within a given household, this can points since 2004. contribute to reducing dispersion in household labor income. The income pooling of heads and spouses reduces inequality Figure 3.1, panel b compares participation rates among as family plays a role in insuring against individual risk in women ages 16 years and above across several countries the labor market. ranked by per capita GDP. First, the plot illustrates that 54 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.1. Labor Market Participation Rates, Mauritius and the Rest of the World, 2004–15 a. Mauritius, by gender (ages 16–64) b. Female rates, Mauritius and world (ages 16+) 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 6 7 8 9 10 11 12 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Log GDP per capita, PPP (constant 2011 international $) All Female Male Mauritius Lower-middle income Individuals not in education Upper−middle income High income OECD average Low income Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius; WDI data. the labor force participation rate of Mauritian women is structural expansion of the female labor force across new around the average for upper-middle-income countries. generations. Second, the participation rate rose steadily throughout the period, from 40.1 percent in 2004 to 46.2 percent The participation gender gap is at a record low among young in 2015, though still below the OECD average, which, individuals ages 16–24 and 25–29 (figure 3.3). The former in 2015, registered a rate of 51.4 percent. Mauritius’s group is associated with the smallest gap, which decreased female participation rate is also catching up with some moderately during the last decade. The latter group appears other upper-middle-income countries in the region, such to be the most affected by the structural change as they as South Africa, which featured a female participation rate experienced the most rapid growth in the female participa- of 47.6 percent in 2015, but it is still distant from those of tion rate and, consequently, the most rapid reduction in other African countries featuring rates above 60 percent the gender gap over the period. However, older individuals such as Botswana (63.2 in 2013) and Seychelles (67.3 per- exhibit a modest reduction in the gender gap and progres- cent in 2015).1 sively lower participation rates among both women and men. Overall, this indicates a general improvement in equality in Average figures hide significant variation across cohorts labor market access, concentrated especially among younger throughout the life cycle and over time (figure 3.2). cohorts of women. Men tend to reach high participation rates well above 90 percent. Younger (older) cohorts increase (decrease) For the purpose of gaining a better understanding of their participation as they enter (leave) the labor market, the barriers obstructing women’s entrance in the labor and the participation of middle cohorts remain stable market, figures 3.4 and 3.5 show participation rates by throughout the mid-life cycle (figure 3.2, panel a). This marital status and educational level, a set of sociodemo- process appears to unfold over time in a rather homo- graphic characteristics that typically affect participation. geneous pattern across cohorts (figure 3.2, panel b). By Despite having increased their participation slightly more contrast, female participation trends are characterized than single women between 2004 and 2015 (by 12 and by more volatility and, after peaking between 25 and 9 percentage points among married and single women, 35 years of age, begin gradually fading out much earlier respectively), married women’s labor force participation than men’s. There was also a significant upward shift in is still considerably lower compared with single women, participation rates among younger cohorts, which is evi- on average, 20 percentage points lower. This gap is mag- dent in figure 3.2, panels c and d, indicating a progressive nified during the early stages of the life cycle, reaching a The Role of Gender Inequality 55 FIGURE 3.2. Labor Market Participation Rates, by Gender and Cohort, 2004–15 a. Men cohorts, over the life cycle b. Men cohorts, over time 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 15 20 25 30 35 40 45 50 55 60 65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 ≤1950 1971−1975 1951−1955 1976−1980 ≤1950 1971−1975 1956−1960 1981−1985 1951−1955 1976−1980 1961−1965 1986−1990 1956−1960 1981−1985 1966−1970 1990−1999 1961−1965 1986−1990 1966−1970 1990−1999 c. Women cohorts, over the life cycle d. Women cohorts, over time 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 15 20 25 30 35 40 45 50 55 60 65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 ≤1950 1971−1975 1951−1955 1976−1980 ≤1950 1971−1975 1956−1960 1981−1985 1951−1955 1976−1980 1961−1965 1986−1990 1956−1960 1981−1985 1966−1970 1990−1999 1961−1965 1986−1990 1966−1970 1990−1999 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. peak among single women in their early 20s when their level. A particularly strong relationship can be observed participation rates are similar to those of men. This seems among women with postsecondary or tertiary education, to suggest that marriage, particularly during the school- who participate in the labor market almost as much and to-work transition years, may represent a deterrent to as long as men do, reaching rates of around 90 percent. women’s entrance into the labor market, while life-cycle The association, however, is not as strong at lower levels events, including pregnancy, impact similarly on the of education, where women are outperformed by men age-profile of participation rates among both single and by at least 30 percentage points, suggesting that higher married women. education is one of the key factors that can help close the participation gender gap. However, young women with Conversely, education emerges as a major driver of women’s upper-secondary education reach a relatively high partici- participation in the labor force. Figure 3.5, panel a illus- pation rate (up to 75 percent), which then rapidly declines trates that female participation increases with educational and reaches a plateau at about 50 percent among women 56 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.3. Labor Market Participation Rates, by Gender and Age-Group, 2004–15 16−24 25−29 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 30−44 45−64 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male Individuals not in education Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius; WDI data. FIGURE 3.4. Female Labor Force Participation Rates, by Marital Status, 2004–15 a. Over time a. Over the life cycle 100 70 90 60 80 70 50 60 Percent Percent 40 50 30 40 30 20 20 10 10 0 0 15 20 25 30 35 40 45 50 55 60 65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Married Single All Married Single Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. The Role of Gender Inequality 57 FIGURE 3.5. Female Labor Force Participation Rates, by Educational Attainment, 2004–15 a. Over time b. Over the life cycle 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 15 20 25 30 35 40 45 50 55 60 65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Up to incomplete primary Upper secondary Up to incomplete primary Upper secondary Complete primary Post-secondary/ Complete primary Post-secondary/ Lower secondary Tertiary Lower secondary Tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. in their 40s, and, from that point on, it is only modestly to a smaller extent, also with upper-secondary schooling higher than the participation rate of women with lower (7.6 percent), while lower levels of education appear to levels of education. be mostly insignificant or even to exert a mildly negative impact (up to -5 percent). The results of a multivariate analysis of labor force partici- pation confirm the bivariate correlations illustrated so far.2 Overall, all the individual and household characteristics Women’s participation increases with age at a decreasing controlled for in the analysis are not able to account for rate, that is, it rises rapidly at young ages to then progres- most of the variation observed in female labor force partici- sively slow at older ages. Married women are, on average, pation. Even if women had exactly the same characteristics 16 percent less likely to enter the labor market relative to observed among men in age, educational level, marital status, single women, while divorced or separated women are rela- and so on, for example, the participation rate predicted by tively more likely (about 10 percent) to do so. Family age means of the multivariate regression would not be as high composition also plays a key role: the presence of children as that of men (figure 3.6, panels a and b).3 this finding is up to 5 years of age can significantly restrain participation corroborated by a Blinder-Oaxaca decomposition of the rates (-6.5 percent), and the same is observed, although to participation differential on a sample of men and women, a lesser degree (-1 percent), with the presence of elderly as displayed in panel c. people (ages 65+). In this respect, the supply of day-care centers and preprimary schools can help attenuate women’s Nearly all the gender difference in participation rates is driven family burdens (box 3.1). This means that caring for chil- by the unexplained component. While the regression does dren or older family members can represent a remarkable not control for differences in the supply and cost of child and obstacle to women’s access to the labor market. By contrast, elderly care services within districts, which could account the presence of children ages 6–15 has a modest positive for some of the unexplained component, the relevance of the effect (4.3 percent) on the probability of participating component may be interpreted as suggestive that women’s in the labor market, which might be partially ascribable access to the labor market critically hinges on factors other to the role that children in that age-group can have in than socio­ economic and demographic characteristics. Cultural complementing the housework of mothers. A major role values and social norms that assign to women a traditional in shaping women’s participation decisions is also played role as the main providers of child and elderly care, house- by educational attainment. In particular, a strong difference hold chores, and other nonmarket activities, dominate the in the probability of entering the labor market is associated empowering effect of education among women with less with postsecondary or tertiary education (31.4 percent) and, than postsecondary educational attainment. 58 Mauritius: Addressing Inequality through More Equitable Labor Markets BOX 3.1. Day-Care Centers and Preprimary Schools Unpaid care work encompasses three aspects: direct care of persons, housework, and unpaid community work (Esquivel 2014). Balancing work and childcare may be particularly difficult for low-income women who have access to a limited range of childcare services. In this respect, policy guidelines, the National Early Childhood Development policy paper (0- to 3-year-olds), was approved by Parliament in 1998 and has been implemented to improve children’s overall development through the introduction and adoption of integrated and holistic approaches to early child development. The Institutions for Welfare and Protection of Children Regulations 2000, under the Child Protection Act, with established norms and standards, was enacted in December 2000 to regulate childcare services, including home-based facilities. It is mandatory for all day-care centers to be registered with the Ministry of Gender Equality, Child Development and Family Welfare. Publicly funded day-care centers have been set up by the ministry to provide free childcare facilities to vulnerable families in deprived regions. Thus, the Mauritius Family Planning and Welfare Association has been offering day-care services at La Tour Koenig, Surinam, and Rivière du Rempart since 1988, 1989, and 1991, respectively. It has provided child day-care and family planning services in depressed areas for more than 20 years. The La Tour Koenig day-care center has adopted an integrated approach to providing day care, babysitting facilities, and sexual and reproductive health services to mothers working in the industrial sector. It aims to offer services in a conducive and child-friendly environment for the optimum development of the child. Services offered at the day-care center include baby care (infants from 3 months to 3 years), babysitting (children over age 3) before and after school hours and during school vacation, medical and pediatric care, and recreational activities and excursions. Services offered to parents include educational sessions on children’s issues (nutrition, development, health, and so on), counseling sessions (children’s behavior, individual welfare, couple relationships, sexual and reproductive health issues), contraceptive supply, specialized sexual and reproductive health services (pap smears, echography, electrocardiography, gynecology, and so on), laboratory services, and HIV counseling and testing. The center has the capacity to cater for 50 infants on a full-time basis. Babysitting facilities are offered for some 40 primary-school children. There is a need for day-care services to help mothers effectively manage their different roles at home, at work, and in the community. Childcare facilities for children ages 0–3 are predominantly offered in Mauritius by for-fee private day-care centers. The number of registered private day-care centers stands at 105 in 2017. The highest share of day-care centers are in the region of Plaines-Wilhems (29 percent), followed by Port-Louis (17 percent) and Flacq (14 percent). Children ages 3–6 also have free access to public preprimary schooling. Most preprimary schools are administered by private individuals, though there are still preprimary schools under municipalities or village councils that are under the purview of local governments. These are also run by the Early Childhood Care and Education Authority, which operates under the aegis of the Ministry of Education and Human Resources, Tertiary Education, and Scientific Research. The authority seeks to provide equal access for all children to quality preschooling, including children at risk of delayed development and disabilities and children living in conditions of vulnerability, through a child-centered and play-based approach with the involvement of the parents. Over the years, there has been a slight decline in the total number of preprimary schools across the various districts. Financial support for childcare is also provided under the National Pensions Act and the Social Aid Act targeted at needy families. In 2013, the government introduced the child allowance (MUR 750 per child), a conditional cash transfer scheme aiming at reducing drop-out rates in primary school. However, there are no data available to assess the impact of this scheme. For example, a child allowance is payable to the children of beneficiaries of a basic widow’s pension or basic invalid’s pension and beneficiaries of social aid, unemployment hardship relief, and income support. The child should normally be under ages 15 or 20 if in full-time education. 3.2 Working Women: at 33.2 percentage points as of 2015. As women find it difficult to get a job, their unemployment rates are notably What Do They Do? larger than men’s figure 3.7, panel b). Yet, between 2005 and 2015, the unemployment gender gap dropped from The trends observed in access to the labor market are largely 3.3 to 1.5 percentage points. reflected in the condition of women who actively participate. The employment-to-population ratio of women is system- This is most markedly so at the middle of the life cycle, atically lower compared with men (figure 3.7, panel a). In where unemployment gender differentials are, on average, 2004, the female employment ratio was at 40.6 percent, 3.2 percentage points and increasing among individuals whereas, among men, it was as high as 86.1 percent, trans- ages  25–29, while the average among individuals ages lating to a gap of 45.5 percentage points, though there 30–44 is 4.2 percentage points and slowly decreasing (fig- was a reduction over time: the employment gap was still ure 3.8, panel b). To a certain extent, earlier stages of the FIGURE 3.6. Counterfactual Participation Rate, by Gender, and Oaxaca-Blinder Decomposition of the Gap, 2004–15 a. Counterfactual female participation rates b. Counterfactual female and actual male participation rates .6 .9 .8 .55 Percent Percent .7 .5 .6 .5 .45 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Actual male Female with male characteristics Female with male characteristics Female with female characteristics Female with female characteristics c. Oaxaca-Blinder decomposition of the participation gap 0 −5 Percentage points −10 −15 −20 −25 −30 −35 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Difference Explained Unexplained Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 3.7. Employment Ratio and Unemployment Rate, by Gender, 2004–15 a. Employment-to-population ratio b. Unemployment rate 100 10 90 9 80 70 8 60 Percent Percent 7 50 40 6 30 5 20 4 10 0 3 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male All Female Male Individuals not in education Individuals not in education Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 3.8. Employment Ratio and Unemployment Rate, by Gender and Age-Group, 2004–15 a. Employment to population ratio 16−24 25−29 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 30−44 45−64 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male Individuals not in education b. Unemployment rate 16−24 25−29 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 25 25 20 20 15 15 Percent Percent 10 10 5 5 0 0 30−44 45−64 25 25 20 20 15 15 Percent Percent 10 10 5 5 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male Individuals not in education Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. The Role of Gender Inequality 61 life cycle have also been characterized by a widening of autonomous professions that favors men. However, without the unemployment gender gap, particularly in recent years, loss of generality, the rest of the analysis is restricted to which have witnessed a sustained rise in the differential, men and women wage workers. from 0.6 percentage points in 2010 to 4.3 percentage points in 2015. Among wage workers, a stable share of about 20 percent work in the public sector (see figure 3.9, panel b). Through- While these stylized facts clearly indicate the presence out the period, the share of men was an average 8 percent of significant labor market entry barriers among young larger than the share of women among wage workers women, the last stage of the life cycle (45–64), when employed in the public sector. This gap shrank during the women move out of the labor market, is associated with last decade, but only modestly, by around 1.5 percent. negligible unemployment gender differentials (on average, 0.5 percentage points). Figure 3.10 offers a more comprehensive view on the secto- rial distribution of women and men employees.4 Overall, Figure 3.9, panel a illustrates the employment category during the last decade, agriculture shrank by approximately distribution among working women and men. The over- 3 percentage points, while textile manufacturing underwent whelming majority of workers are employed for a wage a major contraction of around 9 percentage points. The (85 percent), while the remaining 15 percent are composed reduction in the agricultural sector was of similar magnitude almost exclusively of the self-employed and employers. across genders. The drop in the employment share of textile The share of the employer and self-employed is 7 percent- manufacturing was instead much more pronounced among age points larger among men than women, which might women (down by 20 percentage points) than men (decreas- indicate a gap in the ability to access entrepreneurial and ing by 5 percentage points). Conversely, the shares of other FIGURE 3.9. Employment Category Distribution and Share of Wage Workers in the Public Sector, by Gender, 2004–15 a. Employment category 100 80 60 Percent 40 20 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Female Male All Wage worker Employer/self-employed Contributing family worker Apprentice/intern/other (continued) 62 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.9. Employment Category Distribution and Share of Wage Workers in the Public Sector, by Gender, 2004–15 (continued) b. Wage workers 30 25 20 Percent 15 10 5 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 All Female Male Waged workers aged 16−64 not in education Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. manufacturing and secondary sectors remained roughly (23 percent, on average), and, together with technicians constant, at a 10 percentage point greater concentration of (30 percent), contribute over 50 percent of all employed men. The tertiary sector, including trade, transport, hotels and women in the public sector. The shares of women managers restaurants, information and communication, professional and professionals increased slightly during the last decade, activities, and other services, recorded a moderate increase while the shares of women technicians and clerks steadily (8 percentage points), with shares generally rising relatively declined. In comparison, a much smaller—about 20 percent more among women, with the exception of information cumulatively—and rather stable share of women in the and communication services, which increased by 1.4 per- public sector are employed in services, sales, and elemen- centage point among men and only 0.3 percentage points tary occupations. Compared with men in the public sector, among women. An even more marked and persistent gender women are overrepresented in the top three occupations; difference arose with respect to the relative employment by contrast, men have a larger share in services, sales, and shares in household activities, where a stable 8–9 percent elementary occupations. of women were employed throughout the period, relative to a meagre 1 percent of men. A quite different scenario arises if one looks at the pri- vate sector. Here, about one-quarter of women perform Figure 3.11 illustrates the occupational distribution of wage elementary occupations, and a sizable portion are machine workers separately by gender in the public and private sec- operators (figure 3.11, panel c). The latter, however, sharply tors. About one woman in three employed in the public declined, from 23.5 percent in 2004 to 6.8 percent in 2015, sector is either a manager (2.7 percent) or a professional which is largely ascribable to the reduction in the relative FIGURE 3.10. Sectoral Distribution of Wage Workers, by Gender, 2004–15 a. Women 2004 4.7 31.6 8.7 1.0 12.0 6.0 1.9 6.3 8.5 14.6 4.9 2005 5.1 28.9 8.2 0.8 12.8 6.6 2.3 7.2 7.2 15.5 5.3 2006 4.2 26.5 8.8 0.8 13.7 7.8 2.6 7.3 7.9 15.5 4.9 2007 3.8 24.3 8.2 1.2 14.2 8.0 3.2 8.1 8.3 16.3 4.4 2008 3.6 23.9 8.0 1.6 13.6 7.2 2.2 10.6 7.9 16.3 5.0 2009 4.2 20.9 7.9 1.3 13.2 8.5 1.9 12.6 7.8 17.0 4.9 2010 3.5 18.8 6.9 1.3 14.2 8.9 2.4 12.8 8.1 18.2 4.8 2011 3.2 16.7 8.3 1.3 17.0 8.2 1.6 12.6 8.2 17.6 5.2 2012 2.8 15.1 8.9 1.3 19.0 8.7 2.0 12.8 8.0 16.0 5.6 2013 2.4 14.1 7.7 1.3 17.7 10.2 3.7 10.9 8.3 18.0 5.7 2014 2.2 13.4 7.5 1.9 17.4 8.8 2.1 12.4 9.7 19.3 5.4 2015 2.0 11.6 8.3 1.3 17.0 9.1 2.2 14.6 8.6 19.9 5.2 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Other secondary Information and communication Other services Textile manufacturing Trade and transports Professional activities Public administration Other manufacturing Hotels and restaurants Household activities b. Men 2004 8.1 8.2 12.9 16.0 17.4 8.1 1.5 9.1 0.8 7.5 10.3 2005 7.4 6.8 13.1 17.4 16.4 8.8 1.6 9.8 0.8 7.3 10.6 2006 7.3 7.3 13.5 16.1 16.3 9.1 1.9 8.6 0.9 8.4 10.7 2007 7.6 7.0 13.5 17.2 16.3 8.8 2.1 8.4 0.6 8.4 10.2 2008 6.8 6.3 13.1 17.5 16.9 8.9 1.9 10.0 0.6 8.5 9.3 2009 6.2 5.5 12.2 17.6 18.0 9.0 1.8 10.6 1.0 8.2 9.9 2010 6.2 4.8 12.1 17.6 17.7 8.8 1.8 11.1 0.8 8.5 10.7 2011 6.4 4.5 12.6 15.4 18.4 8.9 2.0 10.9 0.8 8.7 11.3 2012 5.5 3.9 12.3 16.4 19.7 8.9 2.1 10.6 0.7 8.5 11.4 2013 4.9 3.8 11.1 16.7 21.1 9.7 2.4 9.4 1.0 8.3 11.4 2014 5.1 3.9 10.2 15.8 20.5 9.6 2.6 10.8 0.8 8.5 12.3 2015 5.8 3.3 11.3 13.9 19.6 9.5 2.9 12.7 0.9 8.3 11.9 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Other secondary Information and communication Other services Textile manufacturing Trade and transports Professional activities Public administration Other manufacturing Hotels and restaurants Household activities Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 64 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.11. Occupational Distribution of Wage Workers, by Gender and Main Sector, 2004–15 a. Women wage workers, public sector 2004 2.8 22.6 24.6 27.8 11.4 0.9 0.1 0.0 9.8 2005 1.7 20.8 29.8 26.8 9.6 1.1 0.6 0.3 9.4 2006 1.7 20.7 31.6 24.0 10.5 0.0 1.2 0.0 10.2 2007 2.2 19.5 31.6 24.1 12.8 0.5 0.2 0.4 8.8 2008 4.6 19.4 30.9 23.4 12.1 0.2 0.2 0.3 8.9 2009 3.5 18.9 31.5 24.5 9.9 0.3 0.7 0.0 10.7 2010 3.2 20.3 32.5 22.9 9.5 0.3 0.2 0.7 10.5 2011 2.7 27.0 28.3 24.4 0.6 0.0 7.5 0.0 9.5 2012 2.2 21.8 32.0 24.0 8.7 1.1 0.4 0.1 9.6 2013 1.9 28.5 29.1 20.4 8.6 0.1 0.3 0.8 10.2 2014 3.0 26.7 28.8 21.1 10.5 0.3 0.0 0.1 9.4 2015 2.7 25.4 30.1 19.8 11.3 0.9 0.1 0.0 9.7 0 10 20 30 40 50 60 70 80 90 100 Percent Managers Clerks Craft workers Professionals Service and sales Machine operators Technicians Skilled agricultural Elementary occupations b. Men wage workers, public sector 2004 2.5 7.4 15.6 8.0 27.3 2.1 7.3 7.7 22.1 2005 2.7 6.6 15.0 6.9 31.3 1.6 6.9 7.6 21.5 2006 2.7 7.5 14.9 7.3 30.7 2.0 6.9 6.7 21.2 2007 2.6 8.4 14.9 8.2 30.5 2.0 5.9 6.9 20.7 2008 3.2 8.7 14.9 7.4 29.7 1.9 5.4 8.1 20.8 2009 2.8 8.0 16.8 6.1 27.0 2.1 6.5 7.8 22.9 2010 3.2 7.2 17.6 8.5 27.4 2.2 7.0 6.0 21.0 2011 2.2 8.9 15.9 7.3 29.4 2.2 9.3 5.9 19.0 2012 3.3 8.9 17.4 7.3 28.6 2.2 7.8 6.8 17.6 2013 2.8 9.4 21.2 5.2 27.5 1.3 6.6 9.5 16.6 2014 3.3 9.8 19.8 6.1 27.8 1.2 7.9 8.1 15.9 2015 3.5 9.7 17.5 6.7 33.2 1.5 5.9 6.1 15.9 0 10 20 30 40 50 60 70 80 90 100 Percent Managers Clerks Craft workers Professionals Service and sales Machine operators Technicians Skilled agricultural Elementary occupations (continued) FIGURE 3.11. Occupational Distribution of Wage Workers, by Gender and Main Sector, 2004–15 (continued) c. Women wage workers, private sector 2004 1.3 4.6 11.7 9.5 16.8 0.5 5.0 23.5 27.1 2005 1.7 5.1 11.4 11.5 18.5 0.6 5.6 20.9 24.8 2006 0.7 5.0 10.8 12.9 20.0 0.4 5.7 19.7 24.9 2007 1.4 5.4 12.7 12.5 20.9 0.3 5.5 16.3 24.9 2008 1.0 5.1 11.8 13.2 22.0 0.4 5.1 16.2 25.0 2009 1.3 5.2 10.4 13.6 23.8 0.5 5.0 15.8 24.5 2010 1.8 6.4 11.4 12.3 24.1 0.84.3 14.1 24.7 2011 1.8 6.3 6.9 16.5 24.5 0.7 6.1 11.9 25.1 2012 1.9 5.6 8.0 16.2 24.8 0.8 5.7 10.1 26.8 2013 3.2 7.7 9.6 15.6 24.3 0.8 6.0 8.4 24.3 2014 2.6 7.4 10.4 14.9 25.8 0.9 5.1 8.0 24.9 2015 2.9 9.1 9.2 16.2 25.4 0.9 5.6 6.8 23.8 0 10 20 30 40 50 60 70 80 90 100 Percent Managers Clerks Craft workers Professionals Service and sales Machine operators Technicians Skilled agricultural Elementary occupations d. Men wage workers, private sector 2004 2.9 3.1 5.9 5.1 18.7 2.8 25.7 15.7 20.1 2005 2.9 3.3 5.2 4.8 20.0 2.4 25.8 16.1 19.5 2006 3.5 3.8 5.0 5.2 20.8 2.3 25.7 15.2 18.6 2007 4.0 3.4 4.3 5.3 18.6 2.3 27.3 16.0 18.7 2008 3.7 3.1 5.1 6.3 18.7 2.5 26.4 16.0 18.1 2009 3.9 4.1 5.8 6.3 19.5 2.9 25.8 14.5 17.2 2010 3.9 4.8 5.9 6.3 18.8 3.0 25.1 15.4 16.8 2011 4.3 4.3 7.4 6.6 20.1 3.1 25.7 12.9 15.7 2012 3.5 3.8 8.1 5.9 21.0 3.0 26.7 12.4 15.6 2013 4.2 4.7 9.5 6.1 19.9 3.5 25.6 13.3 13.3 2014 5.0 5.1 10.0 5.8 20.1 3.3 25.1 12.3 13.3 2015 5.5 5.7 8.7 6.7 20.2 3.7 23.3 11.9 14.3 0 10 20 30 40 50 60 70 80 90 100 Percent Managers Clerks Craft workers Professionals Service and sales Machine operators Technicians Skilled agricultural Elementary occupations Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 66 Mauritius: Addressing Inequality through More Equitable Labor Markets size of the textile sector following the loss of preferential of the wage distribution to about -10/-15 percent in the access to the American and European markets. An increas- upper tail in 2015. ing number of women are employed in services and sales occupations and as clerks. Moreover, rather differently The unconditional gender wage gap, that is, the gender dif- relative to their counterparts in the public sector, only ference in hourly wages without accounting for employed about 15–20 percent, instead of 50–55 percent of women characteristics, is not necessarily a good indicator. This is private sector wage workers are professionals or technicians. because women and men working in either the public or The average share of women managers over the period the private sector might well be endowed with a set of dif- was also smaller, albeit only slightly (1.8 percent against ferent characteristics, some of which were described earlier 2.6 percent). Similar trends characterize men’s employment in this section, that make them more or less productive. The in the private sector, with the main differences lying in the conditional gender wage differentials, that is the wage gaps larger shares of craft workers and machine operators and obtained after controlling for a set of worker characteris- the significantly lower shares of service and sales work- tics and estimated through a standard wage equation, are ers, professionals, and technicians. In addition, the overall reported in figure 3.14.5 The results show that, all else being employment share in machine-intensive and elementary equal, women in the private sector are paid hourly wages occupations in the private sector is still higher among significantly lower compared with men. The gap is estimated women, though the gap shrank from 16.3 to 5.5 percent- at about 25–30 percent, on average, and has been roughly age points between 2004 and 2015. constant over time. In the public sector, women received an hourly wage premium of about 7 percent in 2015. Trends in educational attainment—shown in appendix D, figure D.1—reveal a considerable increase in the share of Estimates from a Blinder-Oaxaca decomposition indicate wage workers holding a postsecondary or tertiary degree the extent to which the differences observed in hourly wages and a marked decrease in the share among the less well between men and women are ascribable to differences in educated. However, digging deeper, one may uncover impor- the observable characteristics of the two groups or the tant differences by gender in the private and public sectors. explained component, to different treatments of men and As illustrated in figure 3.12, public sector employees are, women, or to unobserved characteristics (or the unexplained on average, more well educated than their private sector component) (see figure 3.14). counterparts. More specifically, on average, 83 percent of public sector workers hold at least some upper-secondary In the public sector, the explained and unexplained compo- education, against only about 50 percent in the private nents work in opposite directions. Differences in observable sector. Moreover, while no substantial educational differ- characteristics (or the explained component) exert a positive ences emerge between genders within the private sector action on the gender wage premium that tends to shore it (figure 3.12, panels c and d), the opposite appears to be true up in favor of women. Among observable characteristics, in the public sector, where a sizable gender discrepancy can occupation and education are capable of narrowing the be observed: a striking 94 percent of women have at least gender wage gap, while others such as demographics, some upper-secondary or higher education, as opposed to industrial sector, and job characteristics appear to dis- 73 percent of men. advantage women. This is in line with the stylized facts presented above, whereby women in the public sector are Gender Wage Gap in the 3.3  relatively more concentrated in high-end occupations and have, on average, a higher educational level. By contrast, Public and Private Sector the unexplained component drags the wage differential down toward negative territory. This component is asso- The socioeconomic differences introduced in the sub­ ciated with a different wage structure or to unobserved section above are, as one would expect, reflected by the characteristics that would, on average, make men more unconditional wage differentials illustrated in figure 3.13. A productive than women. positive wage differential of up to 22 percent is observed between women and men employed in the public sector In the private sector, the decomposition results unveil a in 2004. Conversely, a negative and larger gap is found sizable and negative wage differential at around 30 per- among private sector employees, which ranged from as cent. The two components (explained and unexplained) much as -25/-30 percent in the low and middle portions run in the same direction, and almost all the hourly wage The Role of Gender Inequality 67 FIGURE 3.12. Educational Distribution of Wage Workers, by Gender and Main Sector, 2004–15 a. Women wage workers, public sector 2004 7.8 1.5 62.9 12.5 15.3 2005 6.6 1.0 64.2 13.5 14.8 2006 6.8 1.9 60.2 15.5 15.6 2007 4.9 1.4 62.5 16.0 15.2 2008 3.21.3 62.6 14.2 18.7 2009 5.2 2.2 61.0 10.7 20.9 2010 5.8 3.3 54.4 16.2 20.3 2011 2.82.2 42.5 14.9 37.5 2012 4.8 1.9 39.9 17.0 36.4 2013 4.7 1.0 37.0 15.4 41.9 0.6 2014 3.2 38.0 12.5 45.7 2015 2.42.3 33.5 15.5 46.3 0 10 20 30 40 50 60 70 80 90 100 Percent Up to complete primary Upper secondary Tertiary Lower secondary Post-secondary b. Men wage workers, public sector 2004 44.7 11.9 38.6 2.62.2 2005 43.4 10.4 39.8 3.5 3.0 2006 43.5 10.0 40.5 3.0 3.0 2007 40.7 8.6 43.4 3.7 3.6 2008 39.8 10.5 41.6 3.9 4.2 2009 39.5 10.6 41.8 3.4 4.7 2010 37.5 11.2 42.4 3.5 5.3 2011 33.8 12.3 38.7 5.9 9.3 2012 33.9 11.3 36.2 8.0 10.6 2013 32.0 10.3 36.8 9.5 11.4 2014 31.3 10.4 35.7 9.9 12.6 2015 27.3 10.3 38.6 9.1 14.7 0 10 20 30 40 50 60 70 80 90 100 Percent Up to complete primary Upper secondary Tertiary Lower secondary Post-secondary (continued) 68 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.12. Educational Distribution of Wage Workers, by Gender and Main Sector, 2004–15 (continued) c. Women wage workers, private sector 2004 24.6 8.4 52.6 6.6 7.9 2005 25.3 7.3 53.0 6.7 7.7 2006 24.9 6.2 52.5 8.5 8.0 2007 22.3 7.1 54.7 7.2 8.7 2008 22.8 6.9 52.8 6.2 11.2 2009 23.1 6.7 51.8 6.6 11.8 2010 22.0 6.0 51.5 8.4 12.0 2011 21.2 6.7 46.0 7.4 18.6 2012 19.0 6.3 43.1 10.5 21.2 2013 17.3 5.0 42.1 12.2 23.5 2014 14.4 5.4 43.4 11.6 25.2 2015 13.4 5.3 46.5 10.6 24.2 0 10 20 30 40 50 60 70 80 90 100 Percent Up to complete primary Upper secondary Tertiary Lower secondary Post-secondary d. Men wage workers, private sector 2004 44.8 13.6 36.7 2.32.6 2005 43.8 13.9 36.4 2.9 3.0 2006 41.9 14.5 37.7 2.5 3.5 2007 41.2 14.9 37.3 2.9 3.8 2008 40.5 14.1 38.7 3.0 3.7 2009 38.7 14.4 38.8 3.3 4.8 2010 36.9 14.2 40.2 2.9 5.8 2011 34.9 13.8 36.1 5.3 9.8 2012 34.6 13.5 34.2 8.9 8.7 2013 34.3 13.5 32.5 10.4 9.3 2014 32.1 13.1 33.2 10.2 11.5 2015 30.6 14.1 33.3 9.4 12.6 0 10 20 30 40 50 60 70 80 90 100 Percent Up to complete primary Upper secondary Tertiary Lower secondary Post-secondary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. The Role of Gender Inequality 69 FIGURE 3.13. Unconditional Gender Differentials in Hourly Wages, by Quantile and Sector, 2004–15 a. Public sector 2004 2015 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 0 20 40 60 80 100 0 20 40 60 80 100 Quantile ∈ [Q5,Q95] Quantile ∈ [Q5,Q95] b. Private sector 2004 2015 0 0 −10 −10 −20 −20 Percent Percent −30 −30 −40 −40 −50 −50 0 20 40 60 80 100 0 20 40 60 80 100 Quantile ∈ [Q5,Q95] Quantile ∈ [Q5,Q95] Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The figure shows quantiles of differences between men and women’s hourly wages. difference is ascribable to the unexplained component larger differences in the pay structure and are unable to (about 86 percent in 2015). Observable characteristics bear counteract the effects owing to their relatively exiguous in comparison only limited explanatory power. In particular, productive endowments. the effect of demographics, education, and occupation lay flat around zero, while job characteristics and the indus- The unconditional gender wage differential does not appear trial sector exert a mildly negative effect upon the wage to be constant across the distribution.6 It becomes clear differential. now that the great bulk of the wage difference in the public sector is concentrated in the lower half of the distribution Overall, these results suggest that, on average in the public (figure 3.15, panels a, b, and c). A positive wage difference sector, women have a moderate wage premium by virtue ranging between 10 percent and 15 percent is estimated at of their favorable productive endowments, which some- the 10th and 50th percentiles, while, at the 75th percentile, what compensate for mild forms of unequal treatment. the premium fell to about 8 percent in 2015 and was not Meanwhile in the private sector, women are subjected to significant at the 90th percentile. The Oaxaca-Blinder 70 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.14. Oaxaca-Blinder Decomposition of the Gender Wage Differential at the Mean, by Sector, 2004–15 a. Public sector b. Private sector 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Difference Explained Unexplained Difference Explained Unexplained Group decomposition of explained component Group decomposition of explained component 30 30 20 20 Percent Percent 10 10 0 0 −10 −10 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Demographics Occupation Demographics Occupation Education Industry Education Industry Job characteristics Job characteristics Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. decomposition shows that, in the upper part of the distri- In the private sector, a large and pervasive negative gender bution, the two components diverge, and the unexplained wage differential is estimated along the distribution and component is more negative (figure 3.15, panel c). Thus, across all years (figure 3.15, panels d, e, and f). In this while, in the bottom and middle of the distribution, the nega- case, the gender wage gap becomes larger as one moves tive effect exerted upon the differential by the unexplained down the wage distribution, where it wavers at around component is outdone by a positive composition effect, the 35 percent (as opposed to 10-20 percent at the 75th and opposite is true in the upper tail. This highlights that the 90th percentiles). The wage gap estimated in the private mitigating effect of factors advantaging women, such as sector narrowed over time, particularly in the upper tail. occupation, education and, to some extent, demographic Between 2004 and 2015, the gap declined at the 90th per- characteristics, becomes progressively weaker relative to the centile from 20 percent to 13 percent; it dropped at the different wage structure or unobservables. This ultimately median from 55percent to 36 percent; and it shrank results in a worsening of the gender wage differential toward modestly at the 10th percentile from 43 percent to 39 per- the higher end of the public sector wage distribution. cent. The decomposition results on this sector show that The Role of Gender Inequality 71 FIGURE 3.15. Oaxaca-Blinder Decomposition of the Gender Wage Differential at Selected Percentiles, by Sector, 2004–15 a. P10, public sector b. P50, public sector Group decomposition of explained component 40 40 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 demographics industry difference explained unexplained occupation job characteristics education 40 Group decomposition of explained component 30 40 30 20 20 10 Percent 10 Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 demographics industry difference explained unexplained occupation job characteristics education (continued) the explained and unexplained components are always Despite the considerable progress made over the last decade, negative, and the wage structure (or unexplained) effect Mauritian women are still disadvantaged in the labor market. was larger than the composition effect at all times. This At 57 percent in 2015, female labor force participation was translates into broad gender pay gaps at the bottom and in still 32 percentage points below male participation. Young the middle of the distribution largely because of different women with secondary education contributed the most to wage structures. At the 90th percentile, there was both a narrowing the average gap in participation; single women slight improvement in the composition effect stemming showed a substantially higher participation rate relative to from women’s occupational and industry characteristics married women. In addition to low participation, working and, in some years, also education, and there was a marked women were also penalized in the returns to work in the reduction in the wage structure effect, especially in more private sector. Although, in the private sector, the gender recent years. This led to progressively more moderate trends wage gap was to some extent ascribable to systematic dif- in gender wage differentials in the upper tail of the wage ferences in the observed productive endowments men and distribution of private sector employees. women bring to the labor market, the main factor behind 72 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 3.15. Oaxaca-Blinder Decomposition of the Gender Wage Differential at Selected Percentiles, by Sector, 2004–15 (continued) c. P90, public sector d. P10, private sector 40 40 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 difference explained unexplained difference explained unexplained Group decomposition of explained component Group decomposition of explained component 40 40 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 demographics industry demographics industry occupation job characteristics occupation job characteristics education education (continued) the wage gap appears to have been related to differences young and more well educated cohorts of women. However, in unobservable characteristics or in the pay structure, that a considerable portion of the gender differential remains is, an unequal pay structure to the disadvantage of women. unexplained. While some hypotheses can be posited, includ- By contrast, women in the public sector received a modest ing access to and the cost of childcare, choice of curricula wage premium relative to men thanks to their consider- that are less likely to have more successful job outlets, ably more productive endowments and even though the social and cultural norms, and so on, additional analysis unexplained component operates in favor of men. is needed to provide more fitting answers. Women’s participation has the potential to increase further The public sector appears to be absorbing the most produc- and contribute to narrowing inequality in household labor tive women who benefit from a wage premium with respect income and achieving the full potential of the economy to to their men counterparts. For both equity and efficiency the extent that the income will be more evenly shared across reasons, the substantial gender gap in the private sector, households. In the decades ahead, female participation is notably in the lower tail of the distribution, cannot remain expected to continue following the trends observed over unaddressed. Despite their increased labor force participa- the last 10 years because the pattern was largely driven by tion, Mauritian women are likely to continue to bear most The Role of Gender Inequality 73 FIGURE 3.15. Oaxaca-Blinder Decomposition of the Gender Wage Differential at Selected Percentiles, by Sector, 2004–15 (continued) e. P50, private sector f. P90, private sector 40 40 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 difference explained unexplained difference explained unexplained Group decomposition of explained component Group decomposition of explained component 40 40 30 30 20 20 10 10 Percent Percent 0 0 −10 −10 −20 −20 −30 −30 −40 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 demographics industry demographics industry occupation job characteristics occupation job characteristics education education Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. of the household burden in terms of housework and family for women to switch to a part-time schedule in the same care.7 These activities compete for women’s time and energy job after they have given birth might help reduce the risk of with work on the labor market and might force women career interruptions by allowing a smooth transition from to look for less competitive and less remunerative career maternity leave to employment. Extending paternity leave paths and greater flexibility at work. This might prevent and making it more flexible is an additional instrument women from obtaining access to jobs that reward long aimed at easing the burden borne by women and reducing and inflexible working hours. Policies aimed at easing the the cost of hiring women. caring burden borne by women and encourage men to get more actively involved in housework are certainly welcome. To the extent that the gender pay gap in the private sector Subsidized child and elderly care and work-time regulations is the result of an unequal pay structure, the change needs that promote flexibility and facilitate part-time work may to pass through the education system, which should place be effective. This is particularly important because married a strong emphasis on curbing discriminatory social norms women are less likely to participate in the labor market after among youth. In this respect, the public sector could be pregnancy. Thus, for example, guaranteeing the possibility an example of the best practice in encouraging women’s 74 Mauritius: Addressing Inequality through More Equitable Labor Markets engagement in the labor market and more equitable treat- had men’s characteristics, that is, counterfactual female participation rates were predicted by applying coefficients estimated for women ment. Awareness campaigns might also help shift norms onto the distribution of men’s characteristics. regarding the employment of women in high-pay positions. 4. Figure 3.10 is a close approximation of the sectorial distributions within the private sector because the public sector covers a more restricted number of industries that exhibit negligible gender differences (see appendix C, figure C.1). NOTES 5. Regressions control for second-degree polynomials in age and tenure and individual dummies for each year-of-birth cohort, educational 1. These estimates are based on models of the International Labour level, district of residence, occupation, industry and sectoral category, Organization. that is, the domain of employment in the public sector (central or local 2. A linear probability model was fit by regressing a dummy for participa- government, publicly owned or controlled enterprises) and in the private tion—taking value of 1 if a woman participates in the labor market and sector (export- or nonexport-oriented privately owned businesses, 0 otherwise—onto a set of individual and household characteristics, private household services, cooperative enterprises). including a second-degree polynomial in age, dummies for year-of- 6. To estimate the size of this gap and its effect along the distribution, an birth cohort, dummies for residing in each of the Mauritius districts, unconditional quantile regression is estimated at selected percentiles marital status, educational level, family age composition, and quintiles using the rifreg command in Stata. of household consumption. 7. With financing from the World Bank under the Multi Donor Trust Fund 3. A probit regression was estimated by regressing the participation for Statistical Capacity Building, Statistics Mauritius will carry out a dummy upon the same set of controls employed in the linear probability living conditions survey that will also collect information on time use. model. Estimated coefficients were then used to generate a prediction This will help compare the time devoted by employed and unemployed of the probability of women participating in the labor market as if they men and women to household activities, including chores and family care.  75 CHAPTER 4 Rising Inequality in Wages among Individuals C hapter 1 shows that the observed expansion in in 2001) derived from the employment of household mem­ total household income inequality is largely bers in wage jobs. ascribable to rising inequality in household labor income, while government redistribution helped contain the adverse effects of labor market forces. Chapter 2 4.1  Stylized Facts looks at the role of household demographics and labor market factors in rising labor income inequality among 4.1.1  TRENDS IN OVERALL WAGE INEQUALITY households and identifies the earnings of individuals as Figure 4.2, panel a displays trends in selected percen­ the major contributor. tiles of the logarithm of real monthly earnings for all wage workers between 2004 and 2015.3 It shows a steady This chapter first examines wage inequality among individu­ expansion of monthly earnings inequality. Thus, the earn­ als by (1) presenting stylized facts on the wage structure, ings of 90th percentile earners rose by 46 percent, while (2) pinpointing the principal sources of rising wage dis­ 10th percentile earners experienced an increase of a scant persion, and (3) separating out the effect of the price of 7 percent. The dynamic of hourly wages is mechanically the labor and of workforce composition on wage inequality.1 by-product of the dynamics observed in monthly earnings The second part of the chapter investigates the role of changes in labor supply and labor demand in explaining and in working time.4 the expansion in wage inequality. Beyond labor market forces, institutions can be linked to the dynamics of the There was a reduction in the number of working hours wage structure. The chapter then investigates the effect (measured by the number of hours worked during the of the complex system of ROs on wage inequality and on week preceding the interview) at both the bottom and the employment. The last part of the chapter looks beyond the top of the distribution of log monthly earnings, notably skills shortage, whereby demand (or supply) for a particular between 2010 and 2015 (figure 4.2, panel b).Workers in the type of skills exceeds the supply (or demand) of people with 10th percentile suffered the largest reduction in hours those skills, to address another side of the skills problem: worked. Coupled with the modest increase in their monthly the mismatch between worker educational endowments wages, this led to a rise in their hourly wages moderately and job skill requirements, that is, overeducation and larger than the increase in their monthly wages. The dynam­ undereducation.2 ics observed in monthly earnings and weekly working hours suggests that hourly earnings (or wages for the sake In high- and middle-income countries, labor markets are of simplicity) are expected to show a pattern similar to typically characterized by a large share of wage employ­ that of monthly earnings or slightly less unequal because ment. Mauritius is no exception. Wage workers contributed of the large decline in working hours at the bottom of over 80 percent of total employment over the last decade; the distribution of monthly earnings.5 Figure 4.2, panel c self-employment (including employers) accounts for less displays trends among the percentiles of hourly wages over than 20 percent; and the rest of the employed popula­ the same period ranked according to the distribution of tion is largely composed of contri­ buting family workers monthly wages. As expected, the gradual fanning out (2 percent) and apprentices (0.2 percent) (figure 4.1, pane a). of the distribution over time is more limited compared Wage employment is not only the principal employment with what occurred with monthly wages. For example, category in the labor market, it is also the main source of 10th percentile earners recorded a 19 percent rise in hourly household income from labor (figure 4.1, panel b). About wages relative to a 65 percent increase among earners in 85 percent of household labor income in 2015 (81 percent the 90th percentile. FIGURE 4.1. Composition of Employment, by Category, and of Household Labor Income, by Source, 2001–15 a. Distribution of employment, by type and year 0.5 0.3 0.2 0.1 0.1 0.1 0.1 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0 100 0.1 0.2 0.2 0.1 0.1 0.1 0.1 0.1 0.1 0.2 0.1 0.1 0.1 0.9 0.5 1.6 1.6 2.3 1.9 1.9 1.9 2.0 2.0 2.0 2.3 1.9 2.2 2.1 2.1 2.2 15.9 14.7 14.6 14.3 14.8 14.2 13.9 13.7 13.4 14.6 15.0 14.0 14.2 16.3 17.3 2.9 2.6 2.9 3.2 3.3 3.9 4.0 3.9 3.7 80 2.2 2.1 3.6 4.4 3.9 2.6 60 Percent 40 79.8 79.5 80.3 80.8 80.8 79.8 80.5 80.1 79.7 80.7 79.4 79.2 78.6 79.1 77.4 20 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Waged worker Employer Self-employed Contributing family worker Apprentice/intern Other b. Distribution of labor income, by source and year 0.0 0.2 0.3 0.2 0.1 0.1 0.1 0.1 0.1 0.1 0.0 0.0 0.1 0.0 0.0 100 90 14.7 14.8 15.3 15.2 15.6 15.2 15.3 15.2 17.9 17.2 18.3 17.4 19.3 19.2 21.3 80 70 60 Percent 50 40 30 20 10 80.7 78.4 81.7 80.6 82.8 81.6 82.6 85.2 84.6 84.7 85.1 84.4 84.7 84.6 84.8 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Paid employment Self-employment/enterprise Own production Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: All income measures are expressed in per adult equivalents. Rising Inequality in Wages among Individuals 77 FIGURE 4.2. Trends in Monthly Earnings, Weekly Hours Worked, and Hourly Wages, Selected Percentiles, 2004–15 a. Real monthly wage b. Weekly working hours 80 80 60 60 40 40 Percent Percent 20 20 0 0 –20 –20 –40 –40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p10 p50 p90 p10 p50 p90 p30 p70 p30 p70 c. Real hourly wage 80 60 40 20 0 –20 –40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 p10 p50 p90 p30 p70 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The percentiles of the three variables displayed are ranked according to the distribution of log real monthly wages. In hourly wages, overall inequality can be broken down Overall wage inequality widened more rapidly among into components, namely, between-group inequality and men than women; yet, the level of inequality remained within-group (or residual) inequality. Overall inequality, higher among the latter (figure 4.4). The Gini coefficient measured by the Gini coefficient, expanded by 6 percent increased by 5 percent among women (0.48 in 2015) and in 2004–15, while the P90/P10 ratio increased by almost by 7 percent among men, reaching 0.44 in 2015. The 29 percent, notably between 2008 and 2015 (figure 4.3, P90/P10 ratio rose more sizably, by 42 percent among men panels a and b). The rise in inequality was concentrated in and by half as much among women. Different trends are the upper tail of the distribution. The P90/P50 ratio rose by uncovered in the bottom and top halves of the distribution. about 24 percent, whereas, at the bottom of the distribu­ Among women, inequality grew more rapidly in the lower tion, wage inequality widened by less than 4 percent over tail: the P50/P10 ratio rose by almost 13 percent relative to the entire period (figure 4.3, panels c and d). a growth of 7.3 percent in the P90/P50 ratio (figure 4.4, 78 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.3. Hourly Wage Inequality, by Year, 2001–15 a. Gini coef cient b. P90/P10 ratio .5 2 .475 1.9 .45 1.8 .425 Ratio Gini .4 1.7 .375 1.6 .35 1.5 .325 .3 1.4 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 c. P50/P10 ratio 2015 d. P90/P50 ratio 1.4 1.4 1.3 1.3 1.2 1.2 Ratio Ratio 1.1 1.1 1 1 .9 .9 .8 .8 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. panels a and b). The inequality patterns among men con­ in 2004, men workers with upper-secondary or higher edu­ trast with those among women. Among men, the P50/P10 cation made, on average, about 56 percent more per hour ratio recorded a growth of 10 percent relative to a growth of work than men workers with up to completed primary of almost 30 percent in the P90/P50 ratio (figure 4.4, education. This premium expanded to almost 87 percent panels c and d). in 2015. Conversely, highly educated women workers were paid substantially more than their low-educated counter­ An understanding of the sources of the rise in wage inequal­ parts (about 1.3 times in 2015), and the premium did not ity fully requires an analysis of the role of key observable change much between 2004 and 2015. demographic factors, such as educational attainments, age, and gender. Figure 4.5 displays the education, experience, The experience premium, calculated as the ratio between and gender hourly wage premium.6 The education wage the average hourly wage of workers with 35 or more premium has a distinct gender pattern (figure 4.5, panel a). years of potential experience and the average hourly wage Among men workers, the premium rose by 18 percent of workers with up to 14 years of potential experience, between 2004 and 2015, whereas, among women, it did declined substantially among both men and women (fig­ not increase sizably, though it was considerably larger than ure 4.5, panel b). Nonetheless, in 2015, an average man the corresponding premium among men. To add perspective, worker with 35 plus years of experience still made about Rising Inequality in Wages among Individuals 79 FIGURE 4.4. Hourly Wage Inequality, by Year and Gender, 2001–15 a. P50/P10 ratio, women b. P90/P50 ratio, women 1.4 1.4 1.3 1.3 1.2 1.2 1.1 1.1 Ratio Ratio 1 1 .9 .9 .8 .8 .7 .7 .6 .6 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 c. P50/P10 ratio, men d. P90/P50 ratio, men 1.4 1.4 1.3 1.3 1.2 1.2 1.1 1.1 Ratio Ratio 1 1 .9 .9 .8 .8 .7 .7 .6 .6 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 27 percent more per hour relative to a man with up to experience over 2004–15 and three subperiods, namely, 14 years of experience. By contrast, more experienced women 2004–06, 2007–10, and 2011–15. experienced a wage penalty relative to young women, and the penalty rose from about 20 percent in 2004 to about Three key facts emerge about trends in hourly wages by 33 percent in 2015. The average women’s hourly wage gap group from figures 4.6 and 4.7. First, hourly wages rose more in both the public and private sectors declined modestly, rapidly among women (29.5 percent over 2004–15) com­ from -28 percent in 2004 to -24 percent in 2015 (figure 4.5, pared with men (22.6 percent). The increase among women panel c). was remarkable during the more recent years (2011–15): 25.5 percent compared with 19.9 percent among men. In the first period, 2004–06, men recorded a small decline (-1.8 percent), while women experienced a modest growth INEQUALITY BETWEEN AND 4.1.2  in hourly wages (+1 percent). WITHIN DEMOGRAPHIC GROUPS To investigate wage inequality dynamics across demographic Second, the rise in the wage premium among workers with groups, figures 4.6 and 4.7 display changes in hourly wages upper-secondary or higher education is attributable to the across groups defined by gender, education, and working large increase in hourly wages among these workers relative 80 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.5. Education, Experience, and Gender Hourly Wage Differentials, 2004–15 a. Upper-secondary education premium, by year b. Experience premium, by year 160 80 140 60 120 40 Percent 20 Percent 100 80 0 60 −20 40 −40 20 −60 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Males Females Males Females c. Women’s wage gap −15 −20 −25 Percent −30 −35 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 4.6. Changes in Relative Hourly Wages, by Gender and Education, 2004–15 a. Changes, overall and by gender b. Changes, by educational level, all 35 35 31.1 29.5 30 30 25.5 24.9 25 25 22.8 22.2 21.8 22.6 21.0 19.9 20 19.1 20 17.5 Percent 15.8 Percent 15 13.3 13.1 13.4 15 10 8.4 7.9 10 5 1.7 5 1.0 0 0 –0.9 –1.8 –2.2 –0.6 –5 –5 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 Up to complete primary All Female Male Lower secondary Upper secondary/post-secondary/tertiary (continued) Rising Inequality in Wages among Individuals 81 FIGURE 4.6. Changes in Relative Hourly Wages, by Gender and Education, 2004–15 (continued) c. Changes, by educational level, women d. Changes, by educational level, men 35 32.0 35 31.4 29.3 30.5 30 30 25 23.5 24.0 25 22.4 21.5 20.0 20 20 18.3 16.8 16.1 Percent 16.0 Percent 15 15 10.0 10 9.1 8.9 10 8.1 7.5 5 5 1.7 1.8 1.7 0 0 –0.7 –5 –3.0 –1.8 –5 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 Up to complete primary Up to complete primary Lower secondary Lower secondary Upper secondary/post-secondary/tertiary Upper secondary/post-secondary/tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 4.7. Changes in Real Hourly Wages, by Gender and Experience, 2004–15 a. Changes by experience, all b. Changes by experience, women 40 36.1 50 44.7 35 40 30 25 24.1 24.1 30 28.7 20.0 21.4 23.7 22.9 26.5 22.9 Percent Percent 20 18.3 21.2 20 15 12.9 11.5 10.6 10 10 4.2 5.0 4.0 1.7 5 0.8 0.3 0 0 –2.3 –5 –2.4 –10 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 0–14 15–34 35+ 0–14 15–34 35+ c. Changes by experience, men 50 40 31.3 30 24.6 24.7 Percent 17.9 19.4 20 14.1 12.6 10 6.4 7.7 0 –1.1 –1.2 –10 –2.4 2004–2006 2007–2010 2011–2015 2004–2015 0–14 15–34 35+ Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 82 Mauritius: Addressing Inequality through More Equitable Labor Markets to low-educated workers. Over 2004–15, hourly wages larger among men. Residual inequality among men rose by among highly educated workers increased by 31 percent, almost 53.0 percent, compared with 38.7 percent among compared with 22.2 percent and 15.8 percent among women. Upper-tail residual inequality went up by 29.0 percent workers with lower-secondary and up to completed primary among men and 19.6 percent among women. Such increases education, respectively (figure 4.6, panel b). The gap in the in hourly wage inequality within groups implies that there rate of change was, in this case, larger during 2007–11 was a wage loss among the less well educated and the less relative to more recent years. Among men, the overall experienced and also among the least well educated and the period change in hourly wages among highly educated least experienced within each category. workers was as high as 31.4 percent, which compares with a 10.0 percent and 21.5 percent change among men with lower-secondary and up to completed primary education, Effects of Changes in Wages 4.2  respectively (figure 4.6, panel d). Among women, the gap in hourly wage growth was much smaller: between 2004 and Workforce Composition and 2015, highly educated women posted an hourly wage on Rising Inequality increase of 30.5 percent; women with up to completed primary education experienced a rise of 29.3 percent, and Changes in workforce composition may help explain the women with lower-secondary education had, on average, changes documented in wage inequality. Changes in the a growth of 24 percent (figure 4.6, panel c). composition of the workforce that induce an increase in the share of workers with more unequally distributed wages Third, young workers received larger wage increases can increase wage inequality even if wages (the price of with respect to middle-age and older workers (figure 4.7). labor) are kept constant. Workers with a larger number of Overall, young workers gained 36 percent between 2004 years of working experience typically show less wage dis­ and 2015, which compares with an increase of 24.1 percent persion relative to younger workers; hourly wage dispersion and 11.4 percent among workers ages 15–34 and 35 and is typically higher among highly educated workers relative above years of experience. The large wage growth gap to less well educated workers. These are two examples of the is mainly attributable to the dynamics observed between reason changes in the education or experience composition 2007 and 2010, whereas, during the first period and the of the employed population could lead to changes in wage most recent period, the differences in the growth rate dispersion. These compositional effects are separate from were smaller. Young men and women both experienced a price effects that are the result of shifts in labor demand wage increase considerably larger relative to older workers and labor supply (market forces), plus institutional factors. (figure 4.7, panels b and c). However, while young women If wages are held constant, such compositional effects gained relative to both middle-age and older workers, young can mechanically raise or reduce overall and residual men had a large gain relative to old workers and a wage wage inequality through the effects on the distribution of growth similar to the wage growth among workers with workers with different characteristics that have more or less 15–34 years of experience. dispersed wages. Demographics account for up to one-third of the differences Price effects, holding workforce composition constant, are in hourly wages across workers. There is thus room for rela­ measured by the vertical distance across series in a given year tive hourly wage changes within these groups. Dispersion (figure 4.9; box 4.1).8 The effects of changes in composition within gender, education, and experience cells can be used (holding prices at their 2004, 2010, or 2015 levels) capture as a measure of within-group or residual inequality.7 Overall changes in the level of each series (figure 4.9).9 residual wage inequality expanded sizably in 2004–15, and the expansion accelerated over the second half of the Overall inequality, measured by the P90/P10 ratio, widened period so that the gap between total and residual inequality by about 29 percent between 2004 and 2015. If workforce at the end of the period (28 percent) was smaller than at composition is held constant at the level of 2004, 2010, or the beginning (46 percent) (figure 4.8, panel a).Upper-tail 2015, the rise in overall inequality would still have been residual inequality expanded by about 27.0 percent, com­ remarkable and at least 60 percent as large as the growth pared with a rise of 15.6 percent in lower-tail inequality in unadjusted inequality. Similarly, growth in upper-tail (figure 4.8, panels b and c). The pattern was similar among inequality would have been moderately lower, but still at men and women, although the expansion was considerably least 80 percent as large as the observed upper-tail inequality. Rising Inequality in Wages among Individuals 83 FIGURE 4.8. Overall and Residual Wage Inequality, by Year, 2004–15 a. P90/P10 ratio b. P50/P10 ratio 2 .85 .8 1.8 Ratio Ratio .75 1.6 .7 1.4 .65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Overall actual Residual actual Overall actual Residual actual C. P90/P50 ratio 1.2 1.1 1 Ratio .9 .8 .7 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Overall actual Residual actual Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. By contrast, compositional effects played a large role in observed inequalities had the workforce composition been lower-tail inequality. Had workforce composition stayed held constant at the level of 2004, 2010, or 2015. In the at the level of 2004 or 2010, P50/P10 would have risen by case of women, lower-tail inequality would have widened less than 1 percent or stayed constant. If the characteristics by only about 4 percent relative to the actual 12 percent of the employed population had remained at the level of increase observed between 2004 and 2015. the end of the period the whole time, lower-tail inequality would have declined by 2.5 percent. Residual hourly wage inequality is responsible for a sub­ stantial part of the rise in total inequality. For this reason, Breaking down the analysis by gender indicates that most one might investigate the extent to which the effect is of the compositional effect drives the change observed attributable to changes in the composition of the workforce in inequality among women (figure 4.10, panels a and b). as opposed to changes in prices. Figures 4.11 displays Yet, it fails to explain a large part of the changes observed observed and counterfactual residual wage inequality for in wage inequality among men. Figure 4.10, panels c all workers. Overall residual inequality, measured by the and d show that, in the case of men, upper- and lower-tail P90/P10 ratio, rose by about 47 percent between 2004 inequality would have been 65 percent as large as the and 2015. If workforce composition is held constant at the 84 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.9. Actual and Counterfactual Wage Inequality, 2004–15 a. P90/P10 b. P50/P10 2.1 .9 2 .85 Ratio Ratio 1.9 .8 1.8 1.7 .75 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Counterfactual (2004) Counterfactual (2015) c. P90/P50 1.2 1.15 1.1 Ratio 1.05 1 .95 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. BOX 4.1. The Reweighting Approach A reweighting approach is employed to assess the extent to which changes in equality are ascribable to price effects resulting from the interaction of labor demand and labor supply or to compositional effects that mechanically introduce changes in inequality by altering the share of demographic groups that experience more or less dispersion in wages. The approach was introduced by DiNardo, Fortin, and Lemieux (1996). It consists of decomposing the observed density of wages in two time periods, say t and t′, into a price function that provides the conditional distribution of wages for given characteristics and time period and a composition function that provides the density of characteristics in the same time period. A counterfactual wage density and counterfactual inequality measures may be constructed using this decomposition by combining the price function from a period t with the composition function from a different period t ′. To calculate the counterfactual, the price function at time t must be reweighted by the ratio of the density of characteristics at time t ′ and time t . In practice, such a reweighting function can be estimated using a logit/probit model applied to the pooled data from times t and t ′. The validity of the exercise rests on the assumption of partial equilibrium; thus, prices and quantities can be viewed as independent, (continued) Rising Inequality in Wages among Individuals 85 BOX 4.1. The Reweighting Approach (continued) and changes in labor market quantities do not affect labor market prices. The assumption is not appealing given the changes in labor supply; yet, it might be viewed as an informative exercise. To assess the contribution of shifts in composition and prices to observed changes in overall and residual inequality, the workforce composition data in each year between 2004 and 2015 are applied to the price functions from the years 2004, 2008, and 2015. This allows a set of hypothetical scenarios to be simulated whereby workforce composition changes as it actually did over time, while prices are held constant at the levels in 2004, 2008, and 2015. In the calculation of the reweighting function, a set of demographic characteristics, including dummies for education, a quartic in experience, interactions of the experience quartic with education categories, and dummies for district of residence, are controlled for in regressions run separately by gender. The procedure outlined above is applied to the construction of counterfactuals for overall inequality. In the case of residual inequality, the price function is replaced by a residual pricing function obtained by regressing the logarithm of hourly wages in each year on the full set of characteristics described above and replacing the wage observations with corresponding residuals from the ordinary least squares regression. The residual price function is then used to calculate counterfactual residual densities. FIGURE 4.10. Actual and Counterfactual Wage Inequality, by Gender, 2004–15 a. P90/P50, women b. P50/P10, women 1.4 .8 1.35 .75 1.3 Ratio Ratio 1.25 .7 1.2 .65 1.15 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Counterfactual (2004) Counterfactual (2015) c. P90/P50, men d. P50/P10, men 1.2 .8 1.1 .75 Ratio Ratio 1 .7 .9 .65 .8 .6 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Counterfactual (2004) Counterfactual (2015) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 86 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.11. Actual and Counterfactual Residual Wage Inequality, 2004–15 a. P90/P10 b. P50/P10 1.8 .8 1.7 1.6 .75 Ratio Ratio 1.5 .7 1.4 1.3 .65 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Counterfactual (2004) Counterfactual (2015) c. P90/P50 .95 .9 .85 Ratio .8 .75 .7 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Observed Counterfactual (2010) Counterfactual (2004) Counterfactual (2015) Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. level of 2004, 2010, or 2015, the rise in overall inequality of the rise in overall wage inequality in 2004–15. Despite would still have been remarkable and at least 70 percent large shifts in workforce composition, the evidence pin­ as large as the growth in unadjusted residual inequality. points price effects that are a by-product of the interaction Similarly, growth in upper- and lower-tail inequality would of labor supply and labor demand as the principal culprit have been moderately lower, but still at least 70 percent as behind widening between-group inequality. By contrast, large as the observed lower-tail inequality. compositional changes had only second-order effects. This chapter examines the skills mismatch or skills shortage The Role of Labor 4.3  whereby the demand for or supply of a particular type of Market Forces skill exceeds the supply of or demand for people with that skill. It examines the extent to which this contributed to Chapter 3 shows that income from wage employment is the rising between-group wage inequality. In addition, labor largest contributor to labor income among households. It market institutions can be linked to the dynamics of the also identifies between-group inequality as the main source wage structure. The chapter’s analysis therefore investigates Rising Inequality in Wages among Individuals 87 the effect of the complex system of ROs currently in place in to 15 percent in 2015. Construction remained stable at Mauritius on wage inequality and on employment. Lastly, about 10 percent, while the service sector grew in relative the chapter addresses another side of the skills mismatch importance. The share of trade, hotels and restaurants, that appears to have had an impact: the mismatch between and transport expanded, and the expansion of financial, job market requirements and worker educational endow­ real estate, and professional services (from 5.8 percent to ments, that is, the education mismatch that is given by the 11.2 percent) was remarkable. sum of over- and undereducation.10 A similar transformation occurred within industries in terms Mauritius is an export-oriented and highly diversified of occupations (figure 4.12, panel b). The share of managers economy producing textiles, tourism, and financial and and professionals almost doubled between 2001 and 2015 information and communication technology (ICT) services (from 8.4 percent to 14.4 percent), and technicians and and in which agriculture accounts for less than 3 percent of clerks gained in importance, together with service workers. GDP. Despite the recent slower growth and rising unemploy­ By contrast, low-end occupations, including craftworkers, ment following reductions in trade opportunities, including skilled agricultural workers, and machine operators, and the end of the Sugar Protocol and the Multi Fibre Arrange­ elementary occupations, recorded a reduction in relative ment, and the lower prices for sugar and textile exports, share from 58.7 percent to 45.7 percent. Thus, over the last structural transformation has continued. In employment, the 15 years, the economy continued a transformation away agricultural share that was slightly above 10.0 percent of from agriculture, other traditional sectors, and low-end total employment in 2001 declined and was at 7.5 percent occupations toward modern sectors, including services, in 2015 (figure 4.12, panel a). Manufacturing recorded a particularly professional services, ICT, and tourism with a larger reduction in relative terms, from 26 percent in 2001 parallel increase in the share of high-skilled occupations. FIGURE 4.12. Distribution of the Employed Population, by Industry, Occupation, and Year, 2001–15 a. By sector 2001 11.5 14.6 11.1 1.8 9.6 12.4 5.9 5.7 1.4 5.8 6.5 13.7 2002 10.5 14.3 10.8 1.6 10.4 12.0 6.1 5.5 1.5 6.1 7.1 14.1 2003 11.8 12.5 10.3 1.4 10.6 13.3 5.4 6.2 1.4 6.2 6.8 14.1 2004 9.8 11.7 10.6 1.4 10.6 13.4 5.7 6.3 1.4 6.7 7.5 14.8 2005 9.4 10.8 10.4 1.6 10.5 13.0 5.7 7.3 1.6 7.5 7.5 14.7 2006 9.2 10.3 10.7 1.4 10.3 12.9 5.8 7.4 1.9 7.1 7.7 15.2 2007 8.8 9.7 10.6 1.5 11.1 13.5 6.0 7.2 2.1 6.9 7.1 15.5 2008 8.6 9.4 10.2 1.5 11.2 13.1 6.0 7.3 1.9 8.5 6.7 15.6 2009 8.4 8.5 9.7 1.7 10.7 13.3 6.2 7.6 1.7 9.4 7.0 16.0 2010 7.9 7.6 9.3 1.3 11.1 13.1 6.1 8.0 2.0 10.0 7.2 16.5 2011 7.8 7.0 9.8 1.3 10.2 15.4 5.8 6.8 1.6 9.9 8.1 16.4 2012 7.6 6.2 9.6 1.6 10.5 15.6 6.5 7.1 1.8 9.8 7.8 15.8 2013 7.1 5.9 9.1 1.6 10.4 15.1 6.8 7.9 2.4 8.8 7.6 17.3 2014 7.1 5.6 8.4 1.7 10.6 14.5 6.6 7.5 2.1 9.6 8.0 18.1 2015 7.4 5.0 9.4 1.4 9.6 14.3 6.2 7.5 2.4 11.2 7.5 18.2 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture, shing, mining Textile manufacturing Other manufacturing Utilities Construction Trade Transports Hotels and restaurants Information and communication Professional activities Public administration Other services (continued) 88 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.12. Distribution of the Employed Population, by Industry, Occupation, and Year, 2001–15 (continued) b. By occupation 2001 2.8 5.6 9.2 6.6 17.1 3.7 19.1 13.8 22.2 2002 2.9 6.3 9.3 6.2 17.5 4.0 18.8 14.0 21.0 2003 3.0 6.5 8.5 6.7 18.5 5.6 18.8 12.2 20.2 2004 3.1 6.6 8.5 7.0 18.3 4.0 18.9 12.8 20.6 2005 3.0 6.8 8.7 7.4 19.7 3.9 18.3 12.7 19.6 2006 3.0 6.8 8.4 7.7 20.2 3.8 18.1 12.2 19.7 2007 3.2 6.9 8.5 7.6 20.1 3.6 18.8 12.0 19.2 2008 3.1 6.9 8.6 8.1 20.2 4.1 18.3 11.8 18.9 2009 3.2 7.3 8.9 8.2 20.4 4.2 17.9 11.3 18.6 2010 3.4 7.6 9.5 8.2 20.7 4.3 17.4 10.7 18.1 2011 3.5 7.8 8.7 9.5 21.1 4.2 18.2 9.3 17.8 2012 3.1 7.4 9.8 9.1 21.0 4.5 18.5 9.0 17.5 2013 4.3 9.4 10.5 8.4 19.7 4.5 18.4 9.1 15.9 2014 4.6 9.5 10.8 8.5 20.0 4.6 18.0 8.1 16.0 2015 4.6 9.8 10.4 9.0 20.6 4.8 16.8 7.6 16.4 0 10 20 30 40 50 60 70 80 90 100 Percent Managers and senior of cials Professionals Technicians Clerks Service and sales workers Skilled agricultural Craft workers Machine operators Elementary occupations Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. In parallel, the population became increasingly more well educated. In 2015, less than 5 percent of the population 4.4 Changes in Workforce ages 16 or above had no schooling or only preprimary edu­ Composition cation; around 7 percent had some education, but less than completed primary. The increase in the share of workers The trends in overall and between-group inequality are with secondary, postsecondary, or tertiary education was certainly affected by the relative supply of workers with remarkable. In 2001, less than 6 percent of Mauritians different characteristics. Over the course of the past decade, ages 16 or above had postsecondary or tertiary education. Mauritius witnessed important changes in the supply of Thus, the share had risen almost fourfold over the course labor. Overall, there was a sizable rise in the relative supply of only 15 years (figure 4.13, panel a). The changes in the of women, which increased by 58 percent, and a decline in the educational attainment of the employed population were relative supply of men by almost 11 percent. The expansion even more striking, corroborating the theory that shifts (reduction) of women’s (men’s) relative labor supply has in labor demand and labor supply translated into a net steadily grown over time. The largest change was observed positive shift toward more skilled workers at the expense between 2011 and 2015. of less skilled workers, particularly workers with less than upper-secondary education (figure 4.13, panel b; box 4.2). One of the important factors affecting wage inequality The share of workers with postsecondary or tertiary educa­ is, without doubt, the relative supply of labor at different tion rose from about 7 percent in 2001 to almost 30 percent levels of education and experience. In Mauritius, education in 2015, while the share of workers with less than upper- opportunities improved notably over the years, and primary- secondary and those with upper-secondary education declined and secondary-school enrollments are now comparable by 15.0 and 5.3 percentage points, respectively. with those in upper-middle- and high-income countries FIGURE 4.13. Distribution of Total (Ages 16+) and Employed Population, by Educational Level and Year, 2001–15 a. Population ages 16+ 2001 7.9 12.1 29.2 9.9 35.3 2.4 3.3 2002 7.7 12.0 28.5 10.3 34.7 2.8 4.0 2003 7.1 13.2 27.9 9.9 34.8 2.9 4.1 2004 7.2 11.1 28.5 10.0 35.9 2.8 4.4 2005 7.0 10.8 28.1 9.9 35.9 3.4 4.9 2006 7.1 9.4 28.9 9.9 36.3 3.3 5.1 2007 6.7 8.9 28.0 10.3 37.1 3.4 5.6 2008 6.5 8.5 28.4 10.6 36.4 3.6 6.0 2009 6.2 7.8 28.2 10.5 36.8 3.7 6.7 2010 5.8 8.1 27.2 10.3 37.3 4.2 7.2 2011 6.3 7.2 26.1 10.5 34.0 4.8 11.2 2012 5.6 7.7 25.4 10.1 32.1 6.8 12.2 2013 5.3 7.7 25.1 9.8 31.0 7.8 13.3 2014 4.8 7.4 24.2 9.8 31.7 7.3 14.7 2015 4.5 7.1 23.3 10.0 33.1 6.9 15.1 0 10 20 30 40 50 60 70 80 90 100 Percent No education/pre-primary Incomplete primary Complete primary Lower secondary Upper secondary Post-secondary Tertiary b. Employment population 2001 3.3 9.0 31.2 10.7 38.4 3.4 3.9 2002 2.6 9.1 29.8 11.7 38.1 4.0 4.8 2003 2.9 10.8 29.1 10.8 37.5 4.0 4.8 2004 2.3 8.0 30.1 11.1 39.2 4.0 5.2 2005 2.3 8.0 29.0 11.1 38.8 5.0 5.8 2006 2.1 6.7 30.3 11.0 39.0 5.1 5.9 2007 2.2 5.9 29.3 11.4 40.2 4.9 6.3 2008 1.8 5.6 30.1 11.4 39.5 5.1 6.6 2009 1.9 5.1 29.5 11.1 39.8 5.0 7.5 2010 1.7 5.1 28.2 11.3 39.8 5.6 8.2 2011 1.8 4.8 26.3 11.4 35.6 6.3 13.9 2012 1.5 5.1 25.6 10.8 33.3 8.9 14.7 2013 1.4 4.8 25.1 10.6 31.7 10.1 16.3 2014 1.3 4.3 23.9 10.7 32.1 9.4 18.3 2015 1.14.2 22.8 11.1 33.2 9.2 18.4 0 10 20 30 40 50 60 70 80 90 100 Percent No education/pre-primary Incomplete primary Complete primary Lower secondary Upper secondary Post-secondary Tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 90 Mauritius: Addressing Inequality through More Equitable Labor Markets BOX 4.2. Relative Labor Supply and Relative Labor Demand: A Simplified Framework A pure supply and demand approach to the analysis of changes in the wage structure is adopted in this study. This approach assumes that changes in the wage structure are largely driven by changes in competitive forces (Freeman 1975; Katz and Murphy 1992; Murphy and Welch 1992). It thus does not account for the role of changes in institutional factors. Workers belong to one of two skill groups—skilled (sk) and unskilled (uk)—that are considered two separate labor inputs. The relative wages of these groups are generated by the interaction of the relative supplies of the labor of the groups and an aggregate produc- tion function with associated demand schedules. The framework is partial equilibrium because it does not model the determinants of relative factor supplies. The key assumption is that observed factor prices and quantities are on the demand curve. Assume that there are two periods of time and that the relative wage and relative employment of the skilled group expand over time. Under the assumption of inelastic short-run relative supplies, the increase in the relative employment of skilled workers is reflected by a rightward shift in the relative supply of skilled workers (from S0 to S1 in figure B4.2.1). If the relative demand were stable, the relative wages of skilled workers would decline (from the initial point A to point D). If, instead, the relative wages are observed to go up (by assumption), an outward shift in the relative demand for skilled workers (from D0 to D1) must have determined the rise in the relative wage (from point A to point B). FIGURE B4.2.1. The Relative Supply and Relative Demand Model ln(Wsk/Wuk) S0 S1 D1 D0 C B W1 W0 A D N0 N1 ln(Nsk/Nuk) To test the role of relative labor supply within this simplified framework with two labor inputs and two time periods, t and τ, under the assumption of stable relative factor demand, an increase in the relative supply of a group must lead to a reduction in the relative wage of the same group. (W − W )′ ( X − X ) ≤ 0 t τ t τ (B4.2.1) Time periods in which the above inequality is satisfied could, in theory, be explained by a pure supply shift scenario. Positive inner prod- ucts, however, reject a stable factor demand hypothesis and require an investigation into changes in relative labor demand. In practice, a combination of shifts in relative supply and relative demand is likely to be at play and to be responsible for changes in relative wages. Relative labor demand shifts can be thought of as arising from two types of changes, as follows: • Shifts between industries change the allocation of total labor demand across industries at fixed relative wages (for example, shifts in product demand across industries, shifts in net international trade that affect the domestic share of output at fixed relative wages, and so on). • Shifts within industries change relative factor intensities within industries at fixed relative wages (for example, changes in the prices of nonlabor inputs, outsourcing, and so on). To measure the role of changes in relative demand, the following demand shift indicator is constructed: d ∆X k = ∆Dk = ∑α j jk ∆E j , (B4.2.2) Ek Ek where k indicates demographic group, and j indexes a combination of industry and occupation. The overall demand shift index is constructed by combining industry and occupation j = industry*occupation; the between-industry demand shift index is calculated over industry; and the within-industry index is the difference between the overall demand shift and the between-industry shift indexes and reflects changes in employment among occupations within industries. Rising Inequality in Wages among Individuals 91 (World Bank 2015b). The gender gap has closed: girls in 2011–15. By contrast, women with lower-secondary have passed boys in enrollments in secondary education. education did not experience large changes in relative labor Such exceptionally important changes occurring among the supply (-0.5 percent in 2004–15). However, the first and population are reflected in the shifts in the relative supply the last subperiods saw the relative supply drop by about of labor. Figure 4.14, panels a and b illustrate a massive 22.5 percent and 29.0 percent, respectively, a reduction growth, by 79 percent, in the relative supply of highly that was offset by the strong gain observed in 2007–10 educated (upper-secondary or higher education) women (+58 percent). The relative labor supply of men followed a between 2004 and 2015. The largest push in the increase different pattern. Although the relative supply of low- and in the relative supply occurred in the most recent period mid-educated men declined as observed among women (2011–15), when the expansion was by about 28 percent, (by 48 percent and 27 percent, respectively, in 2004–15), compared with 1.6 percent in 2004–06 and 14.2 percent in the relative supply of highly educated men did not exhibit 2007–10. In parallel, the relative supply of women with up an increase as large as the one observed among women. to completed primary education declined by 40 percent over Over the entire past decade, this relative supply rose by a the entire period. The bulk of the reduction was concentrated meager 2.5 percent. FIGURE 4.14. Changes in the Relative Supply of Workers, by Gender, Education, and Experience, 2004–15 a. Changes in relative supply by education, women b. Changes in relative supply by education, men 90 78.9 90 70 58.1 70 50 50 28.5 30 30 14.2 Percent Percent 10.4 10 2.4 1.6 10 1.6 0.7 0.2 2.5 –10 –0.5 –10 –5.4 –3.6 –12.6 –16.0 –15.0 –30 –22.5 –21.3 –30 –29.1 –29.1 –27.1 –50 –39.6 –50 –48.0 –70 –70 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 Up to complete primary Up to complete primary Lower secondary Lower secondary Upper secondary/post-secondary/tertiary Upper secondary/post-secondary/tertiary c. Changes in relative supply by experience, women d. Changes in relative supply by experience, men 120 120 103.8 100 100 80 80 63.1 60 49.6 60 44.2 Percent Percent 40 28.6 40 34.7 21.6 21.1 20 8.0 5.7 11.3 14.3 20 10.4 9.6 11.3 11.4 4.2 3.0 0 0 –20 –4.2 –20 –13.5 –3.7 –6.7 –15.2 –40 –40 –25.4 2004–2006 2007–2010 2011–2015 2004–2015 2004–2006 2007–2010 2011–2015 2004–2015 0–14 15–34 35+ 0–14 15–34 35+ Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The numbers in the panels represent percentage changes in the share of each group in total labor supply measured in efficiency units. Labor supply in efficiency units is calculated by multiplying annual hours by the relative wage of the group in 2004–15. 92 Mauritius: Addressing Inequality through More Equitable Labor Markets While the changes in the relative supply of low- and mid- Following the definition of demographic groups described educated workers could explain the trends observed in above, workers are categorized in groups defined by gender, the relevant wages, the impressive expansion in the supply educational level, and potential work experience. Changes of highly educated women seems to be at odds with the in relative supply against changes in relative wages are equally large wage gains of this group. A possible explana­ plotted in figure 4.15. Panel a shows the relation between tion might be related to what occurred on the demand side: changes in relative quantities and changes in relative wages substantial shifts in labor demand for this type of worker for the full period (2004–06 to 2012/15), while panels b need to have happened to reconcile the observed trends in and c display the same relationship for 2004/06 to 2007/11 labor supply and wages. and 2007/11 to 2012/15, respectively.13 For the whole period and particularly for 2007/11 to 2012/15, the groups The Mauritian population is aging. This is reflected in the with the largest increases in relative supply had the smallest shifts in relative labor supply among workers with differ­ increases in relative wages. This means that differences in ent numbers of years of potential experience. The relative supplygrowth have the potential to explain the observed supply of the least experienced workers (0–14 years of changes in relative wages. experience) and the most experienced workers (35+ years) rose considerably among women, whereas the relative sup­ Breaking down these patterns by gender shows that dif­ ply of workers with 15–34 years of experience expanded ferences across groups in relative supply growth played less, and it declined considerably among men (figure 4.14, a major role among women (figure 4.16, panels a and b). panels c and d). As in the case of education groups, the Over the whole period, the groups of women with the larg­ largest changes are observed in the most recent years, that est increases in relative supply, particularly women with is, in 2011–15. Overall, the relative supply of experienced lower- or upper-secondary education and over 35 years of women increased by over 100 percent in 2004–15, and the experience, exhibited the smallest increases (or a decline) growth was rapid over the last five years (44 percent). The in relative wages. Conversely, changes in the relative supply relative supply of middle-experienced women increased by of men workers do not have the potential to explain much of about 50 percent. This compares with a growth by 63 percent the changes observed in relative wages, with the exception of among the youngest women. Among men, the overall increase highly educated men with significant experience. However, in the supply of the most and least experienced workers in the case of men, the product of the relative changes in was substantially smaller (34.7 percent and 11.4 percent, supply and the relative changes in wages is barely different respectively), whereas the labor supply of workers with from zero. In the case of both men and women, a pure sup­ 15–34 years of potential experience fell by 25 percent. These ply shift scenario is unlikely to be able to account fully for changes in the age structure of the supply of labor might the observed changes in wages. It is, rather, a combination be an important explanation for the modest increase in the of relative supply and relative demand shifts that have wages of older workers. However, the strong wage gains contributed to the wage inequality patterns. posted by younger women cannot be accounted for solely by the large expansion in their relative supply. 4.5.1  THE ROLE OF FOREIGN LABOR The Role of Shifts in the 4.5  It is important to take into account the role of foreign labor in the analysis of changes in labor supply.14 Mauritius, Relative Supply of Labor as a small open economy, has been increasingly relying on immigrant labor. According to data of Statistics Mauritius, Between-group inequality can be traced back to changes the number of valid work permits increased from around in the relative wages of groups of workers defined by 24,700 in 2004 to 36,800 in 2015. Despite this substan­ demographic characteristics and treated as distinct labor tial rise in number, the share of work permits in total inputs. The relative wages of demographic groups can be employment changed only modestly, from 5.0 percent to thought of as the by-product of the interaction of the relative 6.5 percent between 2004 and 2015 (figure 4.17, panel a). supplies of and the relative demands for these groups (see box 4.2).11 Under the assumption of stable labor demand A large majority of all work permits in 2015 were issued and two labor inputs, an increase in the relative supply of a to workers employed in manufacturing (79.7 percent) and certain group must lead to a reduction in the relative wage construction (12.3 percent) (figure 4.17, panel b). The of that same group.12 remaining permits are issued mostly to immigrants working Rising Inequality in Wages among Individuals 93 FIGURE 4.15. Price Versus Quantity Changes, All Workers, by Period, 2004–15 a. 2004/06 to 2012/15 b. 2004/06 to 2007/11 3.00 1.00 Change in log relative wage Change in log relative wage 2.00 0.50 1.00 0.00 0.00 –1.00 –0.50 –2.00 –1.00 –3.00 –4.00 –1.50 –0.40 –0.30 –0.20 –0.10 0.00 0.10 0.20 0.30 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 Change in log relative supply Change in log relative supply c. 2007/11 to 2012/15 1.50 Change in log relative wage 1.00 0.50 0.00 –0.50 –1.00 –1.50 –2.00 –0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 Change in log relative supply Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 4.16. Price Versus Quantity Changes, by Gender, 2004–15 a. Women workers b. Men workers 2.00 1.00 0.50 Change in log relative wage Change in log relative wage 1.00 0.00 0.00 –0.50 –1.00 –1.00 –1.50 –2.00 –2.00 –2.50 –3.00 –3.00 –4.00 –3.50 –0.40 –0.30 –0.20 –0.10 0.00 0.10 0.20 0.30 –0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 Change in log relative supply Change in log relative supply Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 94 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.17. Foreign Workers, Overall and by Sector, 2004–15 a. Share of foreign workers in total employment 12 10 8 Percent 6 4 2 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 b. Distribution of valid work permits, by sector 100 11.0 8.3 9.2 7.7 8.1 7.6 7.8 7.4 7.6 12.8 13.9 13.1 90 8.1 13.0 16.1 14.1 14.1 12.9 12.3 9.4 6.6 6.9 10.6 17.4 80 70 60 Percent 50 40 79.2 79.7 83.4 77.9 79.3 79.7 77.6 78.1 77.5 75.9 74.1 77.7 30 20 10 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture, hunting and forestry Manufacturing Construction Services Source: Based on Statistics Mauritius data. Note: The number of foreign workers is captured by the number of valid work permits. in wholesale and retail trade and hotels and restaurants. in elementary occupations, including machine operator, This pattern has been relatively stable. Thus, in 2004, masons, carpenters, plumbers, and electricians. manufacturing contributed around 77.0 percent of total work permits, and construction accounted for 9.4 percent, The largest share of immigrant workers consists of men, while services accounted for a larger share in 2004 than in who accounted for over two-thirds of total foreign labor in 2015, 12.8 percent and 7.6 percent, respectively. 2015. With the exception of manufacturing, notably textiles, where some 50 percent of immigrant workers were women Most foreign workers take up jobs in low-skilled occupa­ in 2004, the rest of the sectors employ a small share of tions. Over 80 percent of valid work permits as of December women foreign workers (figure 4.18, panel a). In services, 2015 and December 2016 were held by workers employed women accounted for about 16 percent of foreign labor, Rising Inequality in Wages among Individuals 95 FIGURE 4.18. Foreign Workers, by Gender, Sector, and Country of Origin, 2004– or 2005–15 a. Share of women foreign workers, by sector 60 50 40 Percent 30 20 10 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture, hunting and forestry Manufacturing Construction Services b. Distribution of foreign workers, by country of origin 100 2.6 2.2 2.2 2.4 2.5 2.7 2.4 3.7 3.5 3.2 2.9 3.1 6.5 5.2 5.1 4.8 4.2 4.2 4.0 7.1 8.2 8.8 9.9 9.2 0.3 0.3 0.3 0.3 0.3 0.2 90 0.3 1.1 1.0 0.9 0.4 6.8 7.6 9.6 9.3 9.3 9.4 10.1 0.4 3.0 2.5 3.6 4.0 5.1 80 20.3 20.5 21.4 70 23.0 29.4 25.4 35.0 32.3 30.3 35.4 38.0 60 41.0 8.3 7.4 6.1 11.2 Percent 50 13.4 18.1 40 24.9 20.8 29.6 30 32.7 24.4 54.5 54.9 55.2 35.5 48.1 20 43.0 36.1 25.9 26.4 10 18.8 12.8 15.3 6.8 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Bangladesh China India Madagascar South Africa Sri Lanka Others Source: Based on Statistics Mauritius data. Note: The number of foreign workers is captured by the number of valid work permits. 96 Mauritius: Addressing Inequality through More Equitable Labor Markets whereas, in agriculture and construction, they contributed Examples of between-industry shifts in demand are shifts less than 3 percent and 1 percent, respectively. in product demand across industries and shifts in net international trade affecting the domestic share of output. There was a change in the patterns of country of origin (figure 4.18, panel b). First, over 95 percent of foreign The effect of between-industry changes in labor demand labor arrives from eight countries: Bangladesh, China, clearly depends on differences across demographic groups India, France, Madagascar, South Africa, Sri Lanka, and in the distribution of sectoral (or industrial) employment. the United Kingdom. Second, the relative contribution of Think, for example, of a case in which all working women these top sending countries changed dramatically over the with upper-secondary or higher education are employed in decade. In 2004, India was the first foreign labor contribu­ manufacturing. Under this scenario, a shift in labor demand tor, with a share of over 60 percent, followed by China, across industries, especially between manufacturing and Sri Lanka, and Bangladesh; in 2015, Bangladesh climbed other industrial sectors, would enormously affect the rela­ the ranking and contributed 55 percent, while the share of tive wages of that group of workers. India was reduced to slightly above 20 percent. There are considerable differences in the sectoral and occu­ Overall, foreign labor in Mauritius accounts for only pational distribution of each demographic group defined a modest share of total employment, is predominantly by gender and educational attainment. Figure 4.19, panel a low-skilled, and is employed in elementary occupations in illustrates the distribution of employment across nine indus­ manufacturing and construction. It may marginally moder­ trial sectors. Figure 4.19, panel b displays the distribution of ate the effect of the changes in the labor supply of workers employment across three major occupational categories with low education. Certainly, accounting in the analysis among six demographic groups defined by gender and for the contribution of foreign labor would not alter the education. Women with up to completed primary education direction of the findings. are concentrated in manufacturing (39.4 percent), trade (13.1 percent), agriculture (11.9 percent), and services, mostly household services (20.9 percent), while low-educated The Role of Relative 4.6  men are largely employed in construction (22.5 percent), Demand Shifts manufacturing (19.2 percent), and agriculture (13.9 percent). Conversely, professional activities, public administration, and Shifts in relative labor supply can partially account for the trade and other services in the case of women attract most observed changes in the wage structure and therefore in highly educated workers, that is, workers with, at minimum, between-inequality. However, substantial changes in rela­ an upper-secondary education. tive labor demand occurred over the course of the 2004 to 2015 period. There are a number of factors that may have The patterns are even more striking if one considers the occu­ contributed to changes in relative labor demand. Among pational distribution of workers in different demographic the most frequently cited in the literature are changes in groups. Between 73 percent and 80 percent of women the structure of product demand, increased international and men with low education are employed in craftwork, competition (or changes in the terms of trade because of pro­duction, and elementary occupations, whereas highly shocks in trade agreements), and skill-biased technological educated workers, notably, women, take up professional, change. All these potential factors may explain shifts in technical, and managerial occupations. About one woman in labor demand in favor of more well educated workers. two with lower-secondary education is employed in low-end occupations, and 40 percent in clerical, sales, and service Following the approach of Katz and Murphy (1992), occupations. By contrast, men with the same educational relative labor demand shifts can be divided into changes attainment are primarily engaged in elementary occupa­ that occur within industries and changes that occur between tions (74.6 percent). industries. Within-industry shifts are changes that affect the relative intensities of the use of production inputs within Because different demographic groups are not evenly dis­ industries; between-industry shifts impact the allocation tributed across sectors and occupations, any change in the of total labor demand between industries. Examples of sectoral or occupational distribution of employment will within-industry shifts are nonneutral technological change, have a different impact on each group. Figure 4.20, panel c changes in the prices of nonlabor inputs, and outsourcing. delineates a clear trend of movement out of traditional FIGURE 4.19. Sectoral and Occupational Distribution of Employment, by Demographic Group, Average, 2004–15 a. Sectoral distribution F, ≤ complete primary 11.9 39.4 0.5 13.1 3.9 9.1 0.9 0.4 20.9 F, lower secondary 3.0 36.9 0.4 22.2 12.5 3.6 1.1 0.6 19.9 F, ≥ upper secondary 1.0 16.8 1.0 20.7 8.1 6.2 14.0 7.4 24.9 M, ≤ complete primary 13.8 19.2 22.4 12.5 5.9 10.7 5.9 3.2 6.3 M, lower secondary 6.7 22.3 18.4 15.9 8.1 13.5 6.5 2.7 6.0 M, ≥ upper secondary 3.8 14.3 8.5 14.5 9.8 14.1 11.3 13.7 10.0 0 20 40 60 80 100 Percent Agriculture/mining Manufacturing Construction Trade Hotels/restaurants Transport/utilities Professional activities Public administration Other services b. Occupational distribution F, ≤ complete primary 3.8 22.6 73.6 F, lower secondary 6.2 40.5 53.3 F, ≥ upper secondary 36.2 49.0 14.8 M, ≤ complete primary 2.4 17.7 79.9 M, lower secondary 4.6 20.8 74.6 M, ≥ upper secondary 29.2 35.5 35.3 0 20 40 60 80 100 Percent Professional, technical, and managers Clerical, sales, and service workers Craft, production, and elementary occupations Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 98 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.20. Sectoral and Occupational Distribution of Total Employment, by Period, 2004–15 and Period Average a. Sectoral distribution, by period b. Occupational distribution, by period 2004–2015 6.8 20.0 10.3 15.3 8.4 10.0 9.0 7.3 12.9 2004–2015 18.5 31.2 50.3 2004–2006 7.9 24.0 10.4 14.2 7.9 9.5 7.3 7.2 11.6 2004–2006 16.1 29.2 54.7 2007–2011 6.9 20.3 10.7 15.1 8.3 9.8 9.2 6.9 12.9 2007–2011 17.4 31.4 51.2 2012–2015 5.9 16.5 9.6 16.5 8.9 10.7 10.1 7.7 13.9 2012–2015 21.7 32.3 45.9 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent Agriculture/mining Manufacturing Professional, technical, and managers Construction Trade Clerical, sales, and service workers Hotels/restaurants Transport/utilities Craft, production, and elementary occupations Professional activities Public administration Other services c. Change in sectoral distribution, 2004–15 d. Change in occupational distribution, 2004–2015 Agriculture/mining –2.0 Professional, Manufacturing –8.9 technical 6.9 and managers Construction –2.2 Trade 1.4 Clerical, sales Hotels/restaurants 1.8 and service 5.8 workers Transport/utilities 1.1 Professional activities 4.8 Craft, production Public administration 0.8 and elementary –12.7 Other services 3.2 occupations –12 –10 –8 –6 –4 –2 0 2 4 6 8 –12 –10 –8 –6 –4 –2 0 2 4 6 8 Percentage change 2004–2015 Percentage change 2004–2015 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. sectors, including manufacturing, construction, and agri­ relative labor demand. The advantage of this approach is culture, toward services, notably, professional activities, that it allows one to look at within-industry shifts in labor tourism-related activities, and trade, over the entire period demand that are captured by changes in occupations within 2004–15. In parallel, there was also a decline in the rela­ each industry, in addition to between-industry changes, tive importance of production workers in favor of sales which are measured by changes across industries. and service workers, professional activities, and managers. These patterns are indicative of a demand shift in favor of In 2004–15, the overall indicator of demand shifts increased highly educated workers and against low-educated workers, monotonically by educational level among both women particularly among men. and men workers and, within each educational level except the lowest, shifted in favor of women (figure 4.21, panels a Subdividing the economy into industry-occupation categories and b). Overall labor demand shifts generated a rise in the that are treated as different sectors facilitates an assessment demand for men and women workers with upper-secondary of the magnitude of between- and within-industry shifts in education by 13 percent and 16 percent and a decline in the Rising Inequality in Wages among Individuals 99 FIGURE 4.21. The Between, Within, and Overall Labor Demand Shift Index, by Demographic Group and Period, 2004–15 a. Women, 2004–15 b. Men, 2004–15 15 15 10 10 Percent Percent 5 5 0 0 −5 −5 ≤ Complete Lower ≥ Upper ≤ Complete Lower ≥ Upper primary secondary secondary primary secondary secondary Overall Between Within Overall Between Within c. Women, 2004–06 d. Men, 2004–06 3 3 2 2 Percent Percent 1 1 0 0 −1 −1 ≤ Complete Lower ≥ Upper ≤ Complete Lower ≥ Upper primary secondary secondary primary secondary secondary Overall Between Within Overall Between Within e. Women, 2012–15 f. Men, 2012–15 8 8 6 6 4 4 Percent Percent 2 2 0 0 −2 −2 ≤ Complete Lower ≥ Upper ≤ Complete Lower ≥ Upper primary secondary secondary primary secondary secondary Overall Between Within Overall Between Within Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 100 Mauritius: Addressing Inequality through More Equitable Labor Markets demand for low-educated workers, namely, workers with The observed increase in wage inequality can be explained some primary education or with a certificate of primary by changes in the inequality between groups of workers education, by almost 5 percent for both men and women. defined by demographic characteristics, including gender, Between-industry shifts raised the demand for workers with education, and age. First, the expanding premium for highly upper-secondary or higher education significantly more than educated workers is attributable to the larger increase in the demand for workers with lower-secondary education, the hourly wages of these workers relative to low-educated notably so in the case of women. These differences are workers, particularly among men between 2007 and 2011. ascribable to the higher concentration of highly educated Second, the decline in the experience premium was driven women in expanding industries such as trade, professional by the larger rise in the hourly wages of young workers activities, and other services. By contrast, within-industry relative to their older counterparts mostly between 2007 and shifts induced a modest decline in the demand for men and 2010. In addition to the rise in between-group inequality, women workers in general, but particularly for the least well particularly between high- and low-skilled workers, there educated, and generated an increase in the demand for highly was an increase in inequality within groups, that is, an educated men workers. This is attributable to the reduc­ increased in the inequality in hourly wages within groups tion in the relative importance of production occupations defined by gender, education, or experience. where large shares of low- and mid-educated workers are employed. The size of the demand shifts for highly educated Changes observed in the relative hourly wages of high- workers is, however, smaller than the growth of relative and low-skilled workers and older and younger workers labor supply in the case of women. This means significant are attributable to structural changes in the economy that within-industry and within-occupation demand shifts in generated considerable shifts in relative labor supply and favor of highly educated women workers are the driving demand. Thus, the rapid expansion in the relative demand factor behind the large increase in the relative demand for high-skilled labor outpaced the expansion in the relative for these women workers. supply of this labor. Likewise, there was massive growth in the relative supply of highly educated women, particularly The patterns in overall demand shifts present differences. between 2007 and 2010, but the relative supply of low- and The size of the demand shifts favoring workers with upper- mid-educated workers, both men and women, declined secondary education grew. For example, overall demand appreciably. shifts in favor of highly educated women were larger during 2007–11 and 2012–15. A similar time pattern is observed While relative shifts in labor supply can account for the among highly educated men. This reflects the acceleration rise in hourly wages among low-educated workers if their in the expansion of trade, hotels and restaurants, profes­ relative supply declined, the supply shift scenario is not sional activities, and high-skill occupations during 2007–15. able to explain the large growth in hourly wages among Between-industry demand growth in favor of highly edu­ highly educated women because this was accompanied cated workers accelerated between 2004 and 2011 and by a parallel and similarly large expansion in the relative then lost steam between 2012 and 2015. By contrast, a supply of these women workers. within-industry shift acted against highly educated workers between 2004 and 2011, but then turned in their favor Changes in relative labor demand can square the circle. between 2012 and 2015. Between 2004 and 2015, there was an increase in the relative demand for highly educated workers, particularly women, These findings mirror explanations of the steady decline and a decline in the relative demand for low-educated in income inequality in Latin America since the 2000s workers. Such demand shifts are largely attributable to (Rodríguez-Castelán et al. 2016). There, the decline in changes occurring between industries, notably in 2012–15. labor income inequality was associated with more rapid growth rates in the earnings of less well paid jobs relative Policies targeted at closing the skills shortage have the to labor incomes among more well paid earners. The drop potential to reduce wage inequality and are also beneficial in the higher educational attainment premium relative to in terms of productivity and economic growth. Key are primary educational attainment and the acceleration in investments in skills that are in high demand. This calls for the decline of the secondary-school premium relative to accurate assessments of the country’s current and future primary educational attainment are part of the scenario skill needs, followed by adjustments in education and in the reduction of labor income inequality. training systems to ensure they are responsive to changing Rising Inequality in Wages among Individuals 101 skill needs. A comprehensive strategy to reduce the skills complex system of ROs in place in the private sector is shortage requires, first, securing good-quality public educa­ described in box 4.3. tion. This calls for an approach to providing education that acknowledges the labor market relevance of both medium Appendix E, tables E.1 and E.2 clearly illustrate the sec­ skills (acquired through technical and vocational educa­ toral and occupational variations in legislated wages that tion) and high skills (acquired through tertiary education). underlie this complex wage system. Appendix E, table E.2 Guaranteeing the relevance of education and training for illustrates the complexity of the system by showing the the labor market means there must be effective channels of number of job title categories and the total number of communication between education and workplace actors, wage rates specified within each RO in 2016. Overall, over as well as public-private partnerships. 2000 individual wage rates were specified in that year. Mauritius has been historically characterized by signifi­ In real terms, legislated minimum wages have fallen over cant emigration. According to OECD data, about 96,000 time in most of the RO sectors. Appendix E, table E.3 shows Mauritians ages 15 or above resided abroad in 2000 the real hourly legislated minimum wage for each of the (IOM 2014). Large Mauritian diasporas have been estab­ 30 ROs over 2004–14. The log of the real minimum wage lished in Australia, Canada, France, Italy, South Africa, and for each RO, defined as the lowest stipulated wage rate for the United Kingdom. Every year, an increasing number of each RO, is shown for the 2004–14 period relative to the Mauritian students go abroad for educational purposes: value in 2004 in figure 4.22. about 11,000 in 2015 according to Statistics Mauritius. While more evidence is needed on the size, pattern, and For some of the ROs, the real minimum wage was lower characteristics of the Mauritian diaspora, providing incen­ in 2014 than in 2004. Among these are ROs for attorney tives to Mauritian, who emigrated abroad for educational and notary workers, workers in the baking industry, in reasons, to return to the island and simultaneously incentiv­ retail trades, in newspapers and periodicals, in printing, izing firms that operate in Mauritius to hire these returning among security guards, and in the sugar (nonagricultural) migrants might contribute to reducing the skills shortage. industry. For the ROs that had higher real minimum wages in 2014, only three of the minimum wages are more than 20 percent higher than in 2004, namely, the ROs for domestic The Role of 4.7  workers, livestock workers, and public transport workers Remuneration Orders (+65 percent). However, of the three key RO categories in terms of numbers of covered workers employed (distributive In Mauritius, wages and conditions of employment for a trades, construction, and manufacturing), the construction large portion of the workforce are still determined centrally and the manufacturing ROs (comprising the factory and rather than by the employer firms. This applies to work­ export enterprise ROs) have seen legislated minimum wage ers in both the public and private sectors, where separate increases over the period. wage-setting mechanisms specify wage grids and the duties of every type of worker in the firm in great detail. There­ Trends in the actual mean wages earned by workers in fore, wages do not necessarily reflect the productivity of these RO groups are varied. The mean wage earned by individual firms. workers employed in these RO sectors relative to the mean wage in 2004 is shown in figure 4.23. Changes in aver­ There are three key wage determination institutions in the age earned wages are more erratic and do not seem to be country. The Pay Research Bureau is a permanent body consistent with the trends in the legislated minimum wage. responsible for reviewing the pay and grading structures Average worker wages in more than half the RO-covered and conditions of service in the public sector. The National sectors increased over the period (18 of the 30 RO categories). Remuneration Board (NRB) is a quasi-judicial body respon­ Average wages have declined in real terms in distributive sible for determining the minimum wage and conditions of trade and construction, two of the three key RO categories employment in various sectors and industries in the private in terms of number of covered workers. The third category, sector. The National Tripartite Committee is a high-level manufacturing, which comprises both the factory and export committee responsible for making annual pay adjustments enterprise ROs, is the only major RO sector to have seen based on cost-of-living indicators. The adjustments are both average earned income and legislated minimum wage applicable in both the private and the public sectors. The increases in real terms over the period. BOX 4.3. Wage Setting in the Private Sector The framework for setting the minimum wage and conditions of employment in the private sector is determined by the remuneration regulations of the NRB, collective agreement, or an award of the Employment Relations Tribunal (figure B4.3.1).a The NRB acts as a specialized wage committee or advisory body on wages. The minimum wage is set based on the consumer price index. The consumer price index of the last reviewed base year is chosen and then compared to the wages in the specific sector to determine the level of loss in terms of real wages. The percentage for compensation is calculated and is applied to the last minimum wage of the worker. The NRB is also guided by the principles set out in Section 97 of the Employment Rights Act. The other important factors taken into consider- ation by the NRB are the need to promote decent work and living standards; the need to promote gender equality and to fix wages on the basis of job content; the need to ensure the continued ability of the government to finance development programs and recurrent expenditures in the public sector; the capacity of enterprises to pay; the need to develop payment schemes based on results, and, as far as possible, the need to relate increased remuneration to increased labor productivity. FIGURE B4.3.1. Framework for Establishing a Minimum Wage in the Private Sector IF AGREEMENT IS REACHED, the wages and working conditions are governed by the collective agreement above the remuneration order. Collective Agreements NO AGREEMENT IS REACHED National Remuneration Board Commission for (NRB) – Remuneration Conciliation and Orders (ROs) Mediation (CCM) No agreement is reached and parties do not want to refer the matter to the Employment Relations Tribunal (ERT) Employment Relations Tribunal (ERT) – Come with a judgment of the ERT which is rarely challenged STRIKE Minimum wage rates and working conditions among private sector employees are set by the NRB through ROs. There is no systematic, established time interval for reviewing ROs. For example, of the 30 ROs, only seven have been updated in the last five years. Others, such as the ROs for export-oriented industries and private secondary-school teachers have not been updated in over 30 years. The minimum wage rates specified in the ROs are, however, automatically adjusted every year in line with the salary compensation paid to employees following tripartite negotiations and enacted in the Additional Remuneration Act.b Thirty ROs are currently applied both at the sectoral and occupational level, stipulating different wages to workers covered by each order.c This complexity is further compounded by the fact that some ROs are set for an industry or sector with variations in occupations, while others are set by occupations that vary in the associated occupations. Thus, an office messenger in the sugar industry could potentially be covered by the sectoral RO or the general RO for office assistants. Under the current minimum wage architecture, multiple coverage among workers is therefore common. A further layer of complexity is added by the fact that, in many occupations or occupational categories, wages are stipulated by number of years of work experience. A vertical division has been established between organized sectors in which wages are regulated by collective bargaining and sectors in which the government considers that workers do not have the real bargaining power necessary to fix minimum wages. The fixing of minimum wages by collective bargaining can be provided for by law or result from national practice. After a collective bargaining process, a collective agreement is reached.d However, the collective agreement cannot contain a provision reducing the wage below that provided in the ROs. Workers covered by the ROs can also be covered by collective agreements. If a collective agreement is not reached, the matter is referred to the Commission for Conciliation and Mediation. The sectors covered by a collective agreement in Mauritius are the sugar industry, bus transport, construction, port services, hotels, and catering services. a. The NRB is a quasi-judicial body formerly recognized under Section 45 of the Repealed Industrial Relations Act, but now deemed to have been established under Section 90 of the Employment Relations Act 2008. b. The salary compensation system in Mauritius is a cost-of-living adjustment mechanism. The mechanism for salary compensation takes into account the rate of inflation. Every year, the government issues a decree fixing minimum wage increases that apply to all workers, even those not covered by ROs. The increase in wages is higher among those workers in the lower-wage brackets and lower among those workers at the upper end of the distribution. The salary compensation system focuses on supporting low wages and caters for the vulnerable segments of society by raising purchasing power. Furthermore, the quantum for salary compensation is fixed after various tripartite consultations. c. Wages among public sector workers (namely, those working in parastatals and local authorities) and workers in the finance and banking sector are excluded from these ROs. Furthermore, managerial positions in all sectors are excluded. d. The collective agreement can be drawn up wherever a recognized trade union, a group of recognized trade unions, or a joint negotiating panel and an employer reach an agreement on the terms and conditions of employment. FIGURE 4.22. Changes in the Real Hourly Minimum Wage, by RO, 2004–14 60 40 Percent 20 0 −20 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Baking Blockmaking Catering Cleaning Distributive trade Export Factory Field-crop Light metal Livestock Nursing Printing Pre-primary Private secondary Public transport Sugar-ag Sugar-nonag Tailoring Tea Domestic Security Electrical Newspapers Road haulage Of ce attendants Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: Wages are expressed in 2012 prices. Four ROs with a small number of identified workers are omitted from this figure. These ROs are banks fishermen, cinema employees, salt workers, and travel agent employees. FIGURE 4.23. Changes in Average Real Earned Hourly Wages, by RO, 2004–14 40 20 Percent 0 −20 −40 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Baking Blockmaking Catering Cleaning Distributive trade Export Factory Field-crop Light metal Livestock Nursing Printing Pre-primary Private secondary Public transport Sugar-ag Sugar-nonag Tailoring Tea Domestic Security Electrical Newspapers Road haulage Of ce attendants Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: Wages are expressed in 2012 prices. 104 Mauritius: Addressing Inequality through More Equitable Labor Markets Legislated RO wages have lagged behind the actual wages high-skilled non–RO-covered wage workers more often earned by RO workers (see appendix E, figure E.1).For than less-skilled RO-covered wage workers. example, in 2014, the average legislated RO wage in Mauri­ tius across all the published ROs was MUR 7,200. This was The estimated effect throughout the distribution of hourly 71 percent and 85 percent below, respectively, the actual wages in the RO-covered sectors indicates that RO wages average and median wages earned by wage workers in had a significant positive effect on inequality, particularly RO covered sectors. This sluggish growth in legislated wages in the lower tail up to the 30th percentile (figure 4.24, arises from a combination of factors, namely, these wages panel a). There is also a positive relationship between are adjusted intermittently, and, between wage revisions, RO wages and upper-tail inequality. However, this is more workers only receive inflationary adjustments. However, likely attributable to bias in the estimates.15 The inequality- this disparity reflects an economy in which labor demand increasing effect is larger among men, particularly at the has become increasingly skill-biased in that it rewards bottom of the distribution (below the 20th percentile). FIGURE 4.24. Estimates of the Effect of Remuneration Orders on Hourly Wage Inequality, 2004–15 a. All workers b. Women 1.5 1.5 1 1 Marginal effect Marginal effect .5 .5 0 0 −.5 −.5 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Percentile of log hourly wage Percentile of log hourly wage Con dence intervals in dotted lines Con dence intervals in dotted lines c. Men 1.5 1 Marginal effect .5 0 −.5 0 10 20 30 40 50 60 70 80 90 100 Percentile of log hourly wage Con dence intervals in dotted lines Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: Inequality is measured as the distance between each percentile and the 70th percentile and is a function of the gap between the legislated RO wage and the 70th percentile. The 70th percentile has been chosen as a level of earnings unaffected by legislated RO wages. A positive marginal effect indicates that an increase in legislated RO wages is associated with an increase in wage inequality in the covered sectors. Rising Inequality in Wages among Individuals 105 THE ROLE OF REMUNERATION 4.7.1  decline in employment in the covered sector by 0.57 per­ ORDERS ON EMPLOYMENT cent (table 4.1). Among men, the effect associated with a AND WORKING HOURS 10 percent increase is a decline of employment of 0.77 per­ cent, while, among women, the effect associated with a While the debate and the analysis of the impact of the 10 percent increase is a decline of 1.06 percent in women’s minimum wage regime on the labor market in the United employment in the covered sector. States have been extensive, there is little research in develop­ ing countries. However, a review of 98 studies covered in In terms of hours of work, the results indicate that a Neumark and Wascher (2007), along with 17 more recent 10 percent increase in RO wages is associated with a studies focused on low- and middle-income countries, has 2.3 percent increase in average work hours among men in found a number of employment effects, typically elastici­ the covered sector, but a 1.8 percent decline in average work ties (DPRU and CSDA 2016). The results include aggregate hours among women in the covered sector (table 4.2). This impacts among all workers, coupled with the employment suggests that industries in which women are more likely to impacts among specific demographic groups, regions, and be represented, such as domestic work and services, may sectors.16 Overall, employment elasticities in the studies be responding to increased minimum wages by decreasing reviewed range from 2.17 to -4.60. The mean and median the number of hours worked by women employees rather of all of the cumulative elasticities are -0.22 and -0.11, than firing workers, despite the overall positive employ­ respectively, suggesting that, on average, the impacts of vari­ ment effect. In the uncovered sector, a significant overall ous minimum wage hikes in the countries under review were effect is found, with a 10 percent increase in the minimum marginally negative. Minimum wage-employment elasticity wage associated with an increase in the average number estimates found in 11 low- and middle-income countries range of hours worked of 4.2 percent. The effect is driven by an from a negative value of -1.3 to a positive value of 1.0. The increase in average work hours among men workers in the median elasticity was -0.08 and the mean elasticity was -0.11. uncovered sector. Based on the sample of 59 developed and 32 developing The elasticity of -0.057 is within the range shown for a country estimates reviewed in the study, 81 percent of the number of low- and middle-income countries in the review elasticities were negative, while 19 percent were positive. (DPRU and CSDA 2016). This indicates that increases in Furthermore, the absolute value of these coefficients was minimum wages have a small negative employment effect. small, on average. This suggests that, in general, increases While the overall estimated employment effects seem to be in wages will have either benign or only slightly negative employment effects. With respect to the impact on hours worked, Gindling and TABLE 4.1. Estimates of the Effects of Minimum Terrell (2007) note that the expected sign of this impact Wages on Employment is ambiguous both in theory and in the empirical literature. Higher costs of employment could result in cost-minimizing Employment Employment Employment All Men Women employers reducing the number of workers employed at a fixed monthly or weekly rate and increasing the hours of Log RO -0.0572** -0.0766* -0.106*** the fewer workers that they do employ. However, employ­ (0.0251) (0.0400) (0.0353) ers may also respond to increases in wage rates by either Constant 0.889*** 1.039*** 0.699*** reducing employment at the extensive (cutting the number (0.130) (0.192) (0.226) of employees) or intensive margin (reducing total hours Observations 92,871 58,070 34,801 worked). The regulatory-induced costs of firing workers R-squared 0.163 0.163 0.182 would be one reason, for example, why employers may Note: Explanatory variables in the regressions also include years of choose to keep the wage bill and headcount of employees education, age, age squared, age cubed, gender, dummies for constant, but reduce the number of hours worked. RO categories, dummies for years, and value added by broad indus- try. The covered sector is wage-earning workers in the private sector. The uncovered sector are employers and own-account workers in In the case of Mauritius, the estimates point to a negative the RO sectors covering wage-earning employees, as well as the unemployed who have worked in these sectors at one time. Reported employment effect arising from RO wages across the board. significance levels are based on robust standard errors. A 10 percent increase in the RO wage is associated with a *** p < 0.01 ** p < 0.05 * p < 0.1 106 Mauritius: Addressing Inequality through More Equitable Labor Markets TABLE 4.2. Estimates of the Effects of Minimum Wages on Hours Worked Covered, Covered, Uncovered, Uncovered, Uncovered, Covered, men women all men women Variables all Log hours Log hours Log hours Log hours Log hours Log hours Log RO -0.0429 0.232*** -0.181*** 0.419*** 0.414*** -0.0631 -0.0376 -0.0389 -0.0575 -0.0953 -0.0989 -0.343 Constant 3.647*** 2.232*** 5.138*** 0.454 1.049** 1.648 -0.165 -0.167 -0.334 -0.453 -0.466 -1.972 Observations 67,548 41,345 26,203 17,719 13,436 4,283 R-squared 0.333 0.157 0.358 0.16 0.077 0.191 Note: Explanatory variables in the regressions also include years of education, age, age squared, age cubed, gender, dummies for RO categories, dummies for years, and value added by broad industry. The covered sector is wage-earning workers in the private sector. The uncovered sector are employers and own-account workers in the RO sectors covering wage-earning employ- ees, as well as the unemployed who have worked in these sectors at one time. Reported significance levels are based on robust standard errors. *** p < 0.01 ** p < 0.05 * p < 0.1 negative, this may be the result of context-specific factors wage coverage and gaps in minimum wage compliance in that interact with the minimum wage to ensure that the 11 low- and middle-income countries (Rani et al. 2013).17 effects are small, on average (Bhorat, Kanbur, and Stanwix The study shows that simple national minimum wage 2015). Ultimately, the impact of any enforced change in systems are typically associated with higher compliance wage levels on any particular sector depends on a range rates. There are also countries that combine a national of factors. These include the level of the minimum wage minim wage with sectoral minimum wages, and their relative to average wages, the size of the wage increase, compliance has grown thanks to increasing awareness of the sector under consideration, the timing of wage changes, the rates that apply in each case among both employers the change post-law in worker productivity levels, and finally, and workers. enforcement and compliance. There are four sets of variables that are important in under­ The main argument typically offered in favor of a mini­ standing the factors influencing a gap in compliance in mum wage is that it helps poor and low-income families. the developing world (Bhorat, Kanbur, and Mayet 2013). However, minimum wages often bring about some nega­ There are, first, institutional factors such as the penalty tive employment effects and therefore create winners and structure for noncompliance, the complexity of the wage losers. If the gains are large and concentrated among low- schedule, and the resources allocated to enforcement ser­ income families, some losses can be acceptable to some vices. Second, the individual characteristics of inspectors, policy makers. Empirical evidence on the United States including their level of education, can influence the extent to has shown that minimum wages are not a good instrument which they are effective at achieving compliance. Third, firm to help the poor. This is either because the policy target is characteristics such as size, distance from the enforcement wrong, that is, low-wage workers instead of low-income agency, the number of previous violations, and the level of families when the two groups do not overlap, or because foreign ownership will affect the levels of enforcement and many low-income families have no workers. The latter is violation. Finally, local labor market characteristics such certainly the case of Mauritius, where poor families are as the unemployment rate, the average wage rate relative less likely to have working household members: in 2012, to the minimum wage, and the levels of unionization also about 73 percent of the poor were unemployed or inactive play a role. (World Bank 2015a). In addition, the economic environment and the implemen­ Among the key decisions around a minimum wage system tation of collaborative social policies that coincide with is not only the level of a minimum wage, but also the minimum wage policy can greatly influence compliance complexity of the wage regime and the intensity of enforce­ and enforcement as well as the overall economic impact ment. A recent study has explored issues of minimum of minimum wages. Rising Inequality in Wages among Individuals 107 Skill Mismatch among 4.8  8 percent to 13 percent and a decline in the share of the undereducated (from 38 percent to 34 percent). the Employed and Rising Unemployment While the share of undereducated workers in total employ­ ment declined across all age-groups and notably among Besides skills shortage, an additional increasingly relevant workers ages 15–29, the share of overeducated workers source of skill mismatch is attributable to the difference increased substantially among youth from about 10 percent between the educational level that the employed have and in 2006 to over 20 percent in 2015 (figure 4.26, panels a the educational level required in the jobs or tasks they and b). Women are spearheading such trends. Working perform, the education mismatch (box 4.4).18 An educa­ women are increasingly more well educated than their job tion mismatch can take the form of under- or overeduca­ requires. In 2006, about 6 percent of young working women tion. The first occurs if workers are employed in jobs that were overeducated; this figure was almost four times as require an educational level higher relative to the level they high in 2015 (around 23 percent) (figure 4.26, panels c). hold; by contrast, the latter is realized if workers hold an It appears that, despite their high educational attainment, educational level higher relative to the level necessary to young cohorts of workers encounter increasing difficulties perform the jobs they do. This type of mismatch is on in obtaining jobs that match their educational level. the rise in Mauritius and could, in the long run, prevent the country from realizing the full potential of its labor In parallel, unemployment is on the rise, markedly among force and ultimately constrain productivity and economic youth (figure 4.27, panel a). The unemployment rate among growth. youth ages 15–24 has regularly been three times as high as the overall unemployment rate and significantly higher Overall, the share of mismatched workers, either over- or compared with the unemployment among individuals undereducated, was roughly constant over the last decade ages 25–29. The unemployment rate among youth increased at about 47 percent (figure 4.25). However, there was from about 19 percent in 2008 to 25 percent in 2015. an increase in the share of overeducated workers from The latter compares with about 11 percent among the BOX 4.4. International Standard Classification of Occupations: Definitions of Skill Levels • Skill Level 1. Occupations at skill level 1 typically require the performance of simple and routine physical or manual tasks. They may require the use of hand held tools, such as shovels, or of simple electrical equipment, such as vacuum cleaners. They involve tasks such as cleaning; digging; lifting and carrying materials by hand; sorting, storing or assembling goods by hand (sometimes in the context of mechanized operations): operating nonmotorized vehicles; and picking fruits and vegetables. Many occupations at skill level 1 may require physical strength and endurance. For some jobs, basic skills in literacy and numeracy may be required. If required, these skills would not be a major part of the job. • Skill Level 2. Occupations at skill level 2 typically involve the performance of tasks such as operating machinery and electronic equipment; driving vehicles; maintenance and repair of electrical and mechanical equipment; and manipulation, ordering, and storage of information. For almost all occupations at skill level 2, the ability to read information such as safety instructions, to make written records of work completed, and to perform simple arithmetical calculations accurately is essential. Many occupations at this skill level require relatively advanced literacy and numeracy skills and good interpersonal communication skills. In some occupations, these skills are required for a major part of the work. Many occupations at this skill level require a high level of manual dexterity. • Skill Level 3. Occupations at skill level 3 typically involve the performance of complex technical and practical tasks that require an extensive body of factual, technical, and procedural knowledge in a specialized field. Occupations at this skill level generally require a high level of literacy and numeracy and well-developed interpersonal communication skills. These skills may include the ability to understand complex written material, prepare factual reports, and communicate with people who are distressed. • Skill Level 4. Occupations at skill level 4 typically involve the performance of tasks that require complex problem solving and deci- sion making based on an extensive body of theoretical and factual knowledge in a specialized field. The tasks performed typically include analysis and research to extend the body of human knowledge in a particular field, diagnosis and treatment of disease, imparting knowledge to others, design of structures or machinery and of processes for construction and production. Occupations at this skill level generally require extended levels of literacy and numeracy, sometimes at a high level, and excellent interpersonal communication skills. These skills generally include the ability to understand complex written material and communicate complex ideas in media such as books, reports, and oral presentations. Source: ILO 2012. 108 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE 4.25. Trends in Education Mismatch, 2006–15 100 80 52 52 53 52 52 53 53 51 51 51 60 Percent 40 39 37 36 41 41 40 40 40 37 36 20 10 11 11 12 13 7 8 7 8 8 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Overeducated Undereducated Matching skill pro le Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. Note: The time period is restricted to 2006–15 because of the lack of detailed information on education qualifications according to the International Standard Classification of Education before 2006. FIGURE 4.26. Trends in the Education Mismatch, by Gender and Age-Group, 2006–15 a. Overeducated workers, by age-group b. Undereducated workers, by age-group 22 50 20 45 18 16 40 Percent Percent 14 35 12 30 10 25 8 6 20 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 30−44 16−24 30−44 25−29 45−64 25−29 45−64 (continued) FIGURE 4.26. Trends in the Education Mismatch, by Gender and Age-Group, 2006–15 (continued) c. Overeducated workers, by gender and age-group d. Undereducated workers, by gender and age-group Female Male Female Male 25 25 50 50 22.5 22.5 45 45 20 20 40 40 17.5 17.5 35 35 Percent Percent 15 15 30 30 12.5 12.5 25 25 10 10 20 20 7.5 7.5 15 15 5 5 10 10 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 30−44 16−24 30−44 25−29 45−64 25−29 45−64 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE 4.27. Trends in Unemployment Rates, by Age-Group and Gender, 2006–15 a. Unemployment rate, by age-group b. Unemployment rate, by gender, ages 16–24 30 35 25 30 25 20 Percent Percent 20 15 15 10 10 5 5 0 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 16−24 30−44 Female Male 25−29 45−64 c. Unemployment rate, by gender, ages 25–29 d. Unemployment rate, by gender, ages 30–44 35 35 30 30 25 25 Percent Percent 20 20 15 15 10 10 5 5 0 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Female Male Female Male Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 110 Mauritius: Addressing Inequality through More Equitable Labor Markets 25–29 age-group, 5 percent among the 30–44 age-group, An additional hypothesis can be advanced: the unemployed and less than 2.5 percent among the oldest age-group do not have the skills needed to get a job. In other words, (45–64). The unemployment rate is consistently higher although the Mauritian population has made consider­ among women across all age-groups. Yet, among young able improvement in educational attainment, the formal girls, the gender gap has increased over the last five years education system might not be providing youth with the and reached 10 percentage points in 2015 (figure 4.27, high-quality learning that is required in the labor market. panels b, c, and d). According to the results of the OECD Program for Inter­ national Student Assessment conducted in 2010, Mauritius Unemployed youth are increasingly more well educated. is behind the OECD average and also some middle-income In 2006, more than 25 percent of unemployed youth countries in terms of learning achievements. Mauritius (ages 16–29) held, at most, a primary-school certificate; attained a mean score of 407 on the reading literacy around 10 percent had lower-secondary education; and scale, below the OECD average (493) and below that most of the rest had upper-secondary education (figure 4.28, of Costa Rica, Mexico, Malaysia, Colombia, and Brazil. panel a). In 2015, the share of unemployed youth with up Only 53 percent of students were estimated to have a pro­ to completed primary education had dropped to 5 percent. ficiency in reading literacy that is at or above the baseline This was accompanied by a significant contraction in the level needed to participate effectively and productively in share of unemployed youth with upper secondary (down life. As to the mathematical literacy scale, Mauritius attained from 55.1 percent in 2006 to 42.5 percent in 2015) and a score of 420, which is the same as Chile and Mexico, the a highly marked increase in the share of youth with post­ two lowest performing OECD countries, and below that secondary (12.6 percent) or tertiary education (26.9 percent). of Azerbaijan. In terms of scientific literacy, the attained score was 417 compared with the OECD average of 501 A number of explanations can be advanced to explain the and the score of Mexico, but below the scores of Chile, patterns observed so far, namely, an increasing skills short­ Costa Rica, and Malaysia. About 53 percent of students age, growing overeducation among workers, notably youth, were proficient in science to the baseline level that is con­ and rising unemployment, particularly among the highest sidered necessary to demonstrate competencies that enable educated youth. One way to address this conundrum is people to participate actively in life situations related to to introduce a distinction between voluntarily and invol­ science and technology. Girls had a 12-point higher score untarily unemployment. Individuals may be unemployed in scientific literacy compared with boys. Although these if they have a high reservation wage and are not able to figures concern students who were 15 years old in 2010, find job offers satisfying their requests or the jobs they are they potentially reflect the level of formal learning achieve­ offered do not match their expectations in terms of work­ ments of large segments of the population that has developed ing conditions, working hours, benefits, and so on. This is over the course of the years their educational background sometimes observed among youth with a high socioeconomic under the same system and that appears to lack adequate background, that is, among youth from affluent families literacy and mathematical and scientific skills needed in who can afford to wait for the job that best fits their the labor market. ambitions. However, this situation might also be common among less well off households that receive considerable The lack of a workforce with adequate skills is also reflected public transfers, in the form of noncontributory pensions, in the responses of enterprises surveyed for the World Eco­ for example. Figure 4.28, panel b displays the distribution nomic Forum’s Global Competitiveness Report 2016–2017 of unemployed youth by quintile of household income. (Schwab 2016). Respondents were asked to select the five About 60 percent of youth looking for a job belong to the most problematic factors for doing business in their countries two poorest quintiles, whereas youth from the most well and to rank these factors between 1 (most problematic) and 5. off households contribute less than 10 percent to total About 14 percent of the enterprises surveyed identified an youth unemployment. Overall, the hypothesis of voluntarily inadequately educated workforce as the third most prob­ unemployment does not seem compelling in the case of lematic factor (after insufficient government bureaucracy Mauritius, at least among the richest youth. It is possible and insufficient capacity to innovate). that households benefiting from public transfers might allow their youngest members to wait for their preferred A series of surveys on the labor shortage and the skills gap job. However, it is likely that the answer lies elsewhere or were conducted in 2011 among selected industries, including is a combination of multiple factors. agriculture, financial intermediation, ICT, manufacturing, Rising Inequality in Wages among Individuals 111 FIGURE 4.28. Distribution of Unemployed Youth (Ages 16–29), by Education and Quintile of Household Income, 2006–15 a. By educational level 100 3.1 3.7 4.7 5.0 6.1 3.7 4.2 3.9 4.3 15.8 7.1 19.9 21.4 24.5 26.9 80 8.7 14.5 14.0 13.2 55.1 12.6 53.7 59.0 53.9 60 53.4 Percent 46.2 42.5 40.3 40 42.2 42.5 11.7 14.7 17.6 15.7 15.9 20 17.2 13.5 15.2 26.4 13.1 22.3 19.8 12.9 17.4 17.5 12.1 9.6 9.2 7.0 5.0 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Up to complete primary Lower secondary Upper secondary Post-secondary Tertiary b. By quintile of household income 100 7.6 8.2 8.5 6.0 8.8 9.1 9.2 8.4 10.4 10.0 12.7 15.1 14.2 14.9 14.2 12.6 16.3 17.0 14.6 15.1 80 19.4 19.0 19.4 18.3 18.1 20.5 19.5 21.1 20.7 20.1 60 Percent 26.1 24.4 25.5 25.3 23.2 28.5 23.4 26.1 25.4 28.0 40 20 35.8 33.9 32.6 31.6 32.9 33.0 31.5 28.6 26.2 29.1 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Q1 Q2 Q3 Q4 Q5 Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. 112 Mauritius: Addressing Inequality through More Equitable Labor Markets and export-oriented enterprises (HRDC 2012a–2012d).19 employers in skills development and utilization are key and Between 17 percent and 52 percent of surveyed employers, make the response more likely to be effective. Moreover, depending on the sector, reported a labor shortage in their with a view to enhancing efficiency, the government could organizations: the highest percentage was in export-oriented review the existing range of incentives, including collective enterprises, and the lowest in agriculture. At least 60 percent training funds, tax incentives, and payback clauses, as well of employers in all sectors described labor shortage as a lack as the international evidence on what works. of workers with qualifications and past working experience. By contrast, labor shortage means lack of workers with qualifications only to a minority of employers, typically NOTES less than 10 percent. Employers value not only technical 1. The analysis is restricted to 2004–15, the period of the most rapid skills, but also soft skills, including the ability to work on increase in household labor income inequality. 2. For an extensive discussion of the types and the measurement of the a team, communication skills, the ability to understand skills mismatch, see Bartlett (2012); Cedefop (2010); Johansen and the needs of customers, and ability to innovate and create. Gatelli (2012). 3. The labor module of the CMPHS requires individuals who report to have worked at least one hour during the reference week to report their When asked about the main reasons for the labor shortage, total income, including overtime pay, derived from their job during most employers report difficulty in finding people with the previous month. The module also asks about individuals with jobs who are temporarily absent from work. proficiency in technical skills, the fact that training and 4. The variance in earnings is equal to the variance in wages, plus the education systems do not meet market demands, insufficient variance in hours, plus twice the covariance between wages and hours. 5. Hourly earnings are calculated by dividing monthly earnings by the proficiency in languages, lack of adequate attitudes toward number of hours worked during the week preceding the interview, work, unfavorable conditions with respect to other sectors, multiplied by 4.33. This does not give a precise measure of the amount of time worked each month by wage workers. However, information unavailability to work in shifts or overtime or use flexible from the Survey of Employment and Earnings indicates that the share time arrangements, insufficient job security, or low wages. of wage workers paid by the month increased and was as high as 87 percent in 2014; the rest were paid by the day (8 percent) or by the In agriculture and in export-oriented enterprises, a consider­ hour (3 percent). For inequality, what matters is the distribution of able share of employers report that people have a negative these workers along the distribution of earnings. Wage workers paid by the hour or by the day and not employed full time every week of each opinion about these sectors and therefore are not willing month are likely concentrated in the bottom of the earnings distribu­ to take up jobs in the sectors. Thus, there are problems in tion. This means the pattern of hourly wage inequality shown here is probably an upper-bound estimate of the true distribution of hourly the education and training system that do not appear to wages that would prevail if working time could be fully observed in provide individuals with the adequate learning required by hours, days, weeks, and months worked. 6. The hourly wage premiums illustrated in figure 4.5 are calculated using employers; there are also factors more strictly connected composition-adjusted hourly wages. The adjustment holds constant with the willingness of individuals to take up jobs that the relative employment shares of demographic groups defined by gender, educational attainment, and potential experience across all require flexible working arrangements, that are considered years between 2004 and 2015, thus making sure the premiums are economically unappealing, and that carry a social stigma, not mechanically affected by shifts in the gender, experience, and particularly in agriculture and export-oriented enterprises. education composition of the workforce. The data are sorted into sex, education, and experience groups of two sexes, three education categories (up to completed primary, lower secondary, and upper The education mismatch on the job needs to be addressed secondary and above), and three potential experience categories (0–14, 15–34, and 35+ years). Log hourly wages of workers ages 16–64 with targeted training and retraining programs for under­ not in education are regressed in each year separately by sex on dummy educated workers, largely middle-age and older workers. variables for the three education categories, a quartic in experience, and interactions of the experience quartic with the education catego­ The Mauritian population is aging. This means it is even ries. The composition-adjusted mean log hourly wage for each of the more important to adopt a life-cycle approach to learning. 18 groups in a given year is the predicted log hourly wage from these regressions evaluated at the relevant experience level (7, 25, and 40 years, By contrast, young cohorts of workers who are overeducated depending on the experience group) and educational level. Mean log for their jobs or unemployed are calling for improvements hourly wages for broader groups in each year are calculated as weighted averages of the relevant composition-adjusted cell means using a fixed in the effectiveness of targeted youth employment policies set of weights, equal to the mean share of total hours worked by each and the functioning of employment services. Medium skills group over 2004–15. acquired through technical and vocational education and 7. Residual wage inequality is calculated on the distribution of residuals from a regression of log hourly wages on dummy variables for three training need to be made attractive and relevant to the education categories, a quartic in experience, and interactions of the changing needs of the labor market because they are often experience quartic with education categories separately by gender. The distribution of residuals measures the dispersion of wages within viewed as considerably less prestigious than academic the demographic groups. education certificates. This requires a promotion and com­ 8. To assess the impact of prices and composition effects on overall and residual wage inequality, a reweighting approach is employed munication effort, accompanied by enhanced and continuous (see box 4.1). The composition of the workforce in each year between career guidance. The active involvement and ownership of 2004 and 2015 is applied to a price function from the years 2004, Rising Inequality in Wages among Individuals 113 2010, and 2015. This allows a hypothetical set of cases to be simulated RO wages and p. The correlation is likely to fade as one moves further whereby the composition of the employed population changes as it from the percentile p. See appendix A for details. actually did over time, while labor market prices are held as they were 16. Where a given study produced elasticity estimates for more than one at the beginning of the period (2004), the middle of the period (2010), cohort of workers, the authors included each estimate separately. and the end of period (2015). 17. Coverage gaps represent the proportion of wage earners that are not 9. The validity of the exercise resides on the assumption of partial equi­ covered by minimum wage legislation. Compliance gaps represent librium: prices and quantities can be viewed as independent. Thus, the proportion of wage earners who are covered by minimum wage changes in labor market quantities do not affect labor market prices. legislation, but still make subminimum wages. The assumption is not appealing given the changes in labor supply; 18. This analysis relies on a measure of the match between skills and yet, it may be viewed as an informative exercise. job tasks and duties, the International Standard Classification of 10. For a detailed discussion of the skills mismatch, see Bartlett (2012); Occupations (box 4.4). This normative measure is based on a division Cedefop (2010); Johansen and Gatelli (2012). of major occupations into broad groups. It assigns a level of educa­ 11. The framework is clearly partial equilibrium because the determinants tion to each occupational group in accordance with the International of relative factor supplies are not specified. The only requirement is that Standard Classification of Education. Workers in a group who have the observed prices and quantities be on the demand schedule. assigned level of education are considered well matched. Those who 12. Changes in labor quantities, net of demand shifts, and changes in have a higher (lower) level of education are considered overeducated wages must negatively covary if observed wages and quantities lie on (undereducated). An advantage of the measure resides in the fact that the labor demand schedule. the definition of a mismatch does not change over time; the results are 13. To reduce the number of computations (216 = 18 groups*12 years) and therefore strictly comparable. A disadvantage of the measure is that, by minimize the effect of measurement error, the 12 single-year observa­ construction, it does not allow for overeducation in major groups 1–3 or tions over 2004–15 are aggregated into three three-year intervals, and undereducation in major group 9. Moreover, formal education is only average labor supply and wages are calculated for these time intervals. one component of the measurement of skill level and can be subject Inner products of changes in these measures of supply and wages are to measurement error. calculated between each pair of these three time intervals. 19. The primary data source is a survey questionnaire administered to a 14. The CMPHS data cover only the Mauritian resident population and number of companies and stakeholders backed up by qualitative data may therefore exclude immigrant workers who are in the country on through face-to-face semistructured interviews of a few major players in work permits for short periods of time. local industry. The final number of interviews, conducted between July 15. These estimates are likely to be upward biased because the percentile p and October 2001, was 100 in agriculture and 46 in the agroprocessing in both the left- and right-hand side of the equation and transitory subsector, 185 in manufacturing, 97 in the export-oriented enterprise fluctuations in percentile p are correlated with the gap between the sector, 90 in the financial intermediation sector, and 95 in ICT.  115 REFERENCES Atkinson Anthony B. 2007. “Measuring Top Incomes: Primer.” IZA Discussion Paper 9204 (July), Institute Methodological Issues.” In Top Incomes over the for the Study of Labor, Bonn, Germany. Twentieth Century: A Contrast between Continental Blinder, Alan S. 1973. Wage Discrimination: Reduced Form European and English-Speaking Countries, edited by and Structural Estimates. Journal of Human Resources Anthony B. Atkinson and Thomas Piketty. New York: 8 (4): 436–55. Oxford University Press. Bosch, Mariano, and Marco Manacorda. 2010. “Minimum ———. 2015. Inequality: What Can Be Done? Cambridge, Wages and Earnings Inequality in Urban Mexico.” MA: Harvard University Press. American Economic Journal: Applied Economics 2 Autor, David H., Lawrence F. Katz, and Melissa S. Kearney. (4): 128–49. 2008. “Trends in U.S. Wage Inequality: Revising the Burtless, Gary. 1999. “Effects of Growing Wage Disparities Revisionists.” Review of Economics and Statistics 90 and Changing Family Composition on the U.S. Income (2): 300–23. Distribution.” European Economic Review 43 (4–6): Autor, David H., Alan Manning, and Christopher L. 853–65. Smith. 2016. “The Contribution of the Minimum Cedefop (European Centre for the Development of Vocational Wage to US Wage Inequality over Three Decades: Training). 2010. “Skill Mismatch in Europe.” Briefing A Reassessment.” American Economic Journal: Applied Note (June), Cedefop, Thessaloniki. Greece. http://www. Economics 8 (1): 58–99. cedefop.eurpa.eu/EN/Files/9023_en.pdf. Azevedo, João Pedro, Gabriela Inchauste, and Viviane Daly, Mary C., and Robert G. Valletta. 2006. “Inequality Sanfelice. 2013. “Decomposing the Recent Inequality and Poverty in United States: The Effects of Rising Decline in Latin America.” Policy Research Working Dispersion of Men’s Earnings and Changing Family Paper 6715, World Bank, Washington, DC. Behavior.” Economica 73 (289): 75–98. Azevedo, João Pedro, Minh Cong Nguyen, and Viviane Devereux, Paul J. 2004. “Changes in Relative Wages and Sanfelice. 2012. “ADECOMP: Stata Module to Estimate Family Labour Supply.” Journal of Human Resources Shapley Decomposition by Components of a Welfare 39 (3): 696–722. Measure.” Statistical Software Components 457562, DiNardo, John Enrico, Nicole M. Fortin, and Thomas Department of Economics, Boston College, Chestnut Lemieux. 1996. “Labor Market Institutions and the Hill, MA. Distribution of Wages, 1973–1992: A Semi­ parametric Azevedo, João Pedro, Viviane Sanfelice, and Minh Cong Approach.” Econometrica 64 (5): 1001–44. Nguyen. 2012. “Shapley Decomposition by Components DPRU (Development Policy Research Unit) and CSDA of a Welfare Measure.” Unpublished working paper, (Centre for Social Development in Africa). 2016. World Bank, Washington, DC. Investigating the Feasibility of a National Minimum Bartlett, Will. 2012. “Skills Anticipation and Matching Wage for South Africa. Cape Town: DPRU, University Systems in Transition and Developing Countries: of Cape Town; Johannesburg: CSDA, University of Conditions and Challenges.” ETF Working Paper Johannesburg. (February), European Training Foundation, Turin, Esquivel, Valeria. 2014. “What Is a Transformative Italy. http://www.etf.europa.eu/webatt.nsf/0/84E964 Approach to Care, and Why Do We Need It?” Gender F6CBD16532C1257AAD0038EC27/$file/Skills%20 and Development 22 (3): 423–39. matching%20systems.pdf. Fortin, Nicole M., and Tammy Schirle. 2006. “Gender Bhorat, Haroon, Ravi Kanbur, and Natasha Mayet. 2013. Dimensions of Changes in Earnings Inequality in “The Impact of Sectoral Minimum Wage Laws on Canada.” In Dimensions of Inequality in Canada, Employment, Wages, and Hours of Work in South Africa.” edited by David A. Green and Jonathan R. Kesselman, IZA Journal of Labor and Development 2 (1): 1–27. 307–46. Vancouver: UBC Press. Bhorat Haroon, Ravi Kanbur, and Benjamin Stanwix. Freeman, Richard B. 1975. “Overinvestment in College 2015. “Minimum Wages in Sub-Saharan Africa: A Training?” Journal of Human Resources 10 (3): 287–311. 116 Mauritius: Addressing Inequality through More Equitable Labor Markets Gindling, Thomas, and Katherine Terrell. 2007. “The Effects Murphy, Kevin M., and Finis Welch. 1992. “The Structure of Multiple Minimum Wages throughout the Labor of Wages.” Quarterly Journal of Economics 107 (1): Market: The Case of Costa Rica.” Labour Economics 285–326. 14 (3): 485–511. Neumark, David, and William L. Wascher. 2007. “Minimum HRDC (Human Resource Development Counci). 2012a. A Wages and Employment.” Foundations and Trends in Study on Labour Shortage in the Agricultural Sector in Microeconomics 3 (1–2): 1–186. Mauritius. Ebène Cybercity, Mauritius: HRDC. Oaxaca, Ronald L. 1973. “Male-Female Wage Differentials ———. 2012b. A Study on Labour Shortage in the Financial in Urban Labor Markets.” International Economic Intermediation sector in Mauritius. Ebène Cybercity, Review 14 (3): 693–709. Mauritius: HRDC. OECD (Organisation for Economic Co-operation ———. 2012c. A Study on Labour Shortage in the ICT/ and Development). 2011. Divided We Stand: Why BPO sector in Mauritius. Ebène Cybercity, Mauritius: Inequality Keeps Rising. Paris: OECD Publishing. HRDC. ———. 2015. In It Together: Why Less Inequality Benefits ———. 2012d. A Study on Labour Shortage in the All. Paris: OECD Publishing. Manufacturing sector in Mauritius. Ebène Cybercity, Pencavel, John. 2006. “A Life Cycle Perspective on Changes Mauritius: HRDC. in Earnings Inequality among Married Men and ILO (International Labour Organization). 2012. Structure, Women.” Review of Economics and Statistics 88 (2): Group Definitions, and Correspondence Tables. Vol. 1 232–42. of International Standard Classification of Occupations: Rani, Uma, Patrick Belser, Martin Oelz, and Setareh Ranjbar. ISCO-08. Geneva: ILO. 2013. “Minimum Wage Coverage and Compliance in ———. 2014. “Minimum Wages Fixing Challenge in Developing Countries.” International Labour Review Mauritius.” ILO, Antananarivo, Mauritius; Inclusive 152 (3–4): 381–410. Labour Markets, Labour Relations, and Working Rodríguez-Castelán, Carlos, Luis F. López-Calva, Nora Conditions Branch, ILO, Geneva. Lustig, and Daniel Valderrama. 2016. “Understanding IOM (International Organization for Migration). 2014. the Dynamics of Labor Income Inequality in Latin Migration in Mauritius: A Country Profile 2013. Geneva: America.” Policy Research Working Paper 7795, World IOM. Bank, Washington, DC. Jenkins, Stephen P. 2016. “Pareto Models, Top Incomes, Schwab, Klaus, ed. 2016. Insight Report: The Global and Recent Trends in UK Income Inequality.” IZA Competitiveness Report 2016–2017 . Geneva: Discussion Paper 10124 (August), Institute for the World Economic Forum. http://www3.weforum. Study of Labor, Bonn, Germany. org/docs/GCR2016-2017/05FullReport/ Johansen Jens, and Debora Gatelli. 2012. “Measuring TheGlobalCompetitivenessReport2016-2017_FINAL. Mismatch in ETF Partner Countries: A Methodological pdf. Note.” February, European Training Foundation, Turin, Svirydzenka, Katsiaryna, and Martin Petri. 2014. Italy. “Mauritius: The Drivers of Growth, Can the Past Juhn, Chinhui, and Kevin M. Murphy. 1997. “Wage be Extended?” IMF Working Papers 14/134 (July), Inequality and Family Labor Supply.” Journal of Labor International Monetary Fund, Washington, DC. Economics 15 (1): 72–97. UNECE (United Nations Economic Commission for Katz, Lawrence F., and Kevin M. Murphy. 1992. “Changes Europe). 2011. Canberra Group Handbook on in Relative Wages, 1963–1987: Supply and Demand Household Income Statistics . 2nd ed. Geneva: Factors.” Quarterly Journal of Economics 107 (1): UNECE. https://www.unece.org/fileadmin/DAM/ 35–78. stats/groups/cgh/Canbera_Handbook_2011_ Marrero, Gustavo A., and Juan G. Rodríguez. 2013. WEB.pdf. “Inequality of Opportunity and Growth.” Journal of van der Weide, Roy, and Branko Milanovic ´. 2014. “Inequality Development Economics 104: 107–22. Is Bad for Growth of the Poor (but Not for That of the Marrero, Gustavo A., Juan G. Rodríguez, and Roy van der Rich).” Policy Research Working Paper 6963, World Weide. 2016. “Unequal Opportunity, Unequal Growth.” Bank, Washington, DC. Policy Research Working Paper 7853, World Bank, Walker, Maurice. 2011. PISA 2009 Plus Results: Washington, DC. Performance of 15-Year-Olds in Reading, Mathematics, References 117 and Science for 10 Additional Participants. Melbourne: ———. 2015b. Mauritius: Systematic Country Diagnostic. ACER Press. Report 92703-MU (June 25). Washington, DC: World World Bank. 2011. World Development Report 2012: Bank. Gender Equality and Development. Washington, DC: ———. 2016. Poverty and Shared Prosperity 2016: Taking World Bank. On Inequality. Washington, DC: World Bank. ———. 2015a. Mauritius, Inclusiveness of Growth and ———. 2017. “Country Partnership Framework for Shared Prosperity. September. Washington, DC: GPVDR Mauritius for the Period FY17–FY21.” Report 112232– Africa, World Bank. MU (April 20), World Bank, Washington, DC.  119 APPENDIX A Identification Strategy: is important in capturing the idea that a change in ROs is likely to have more impact on the wage distribution Remuneration Orders where it is more binding. The term “effective” is used owing and Wage Inequality to the fact that it expresses the RO wage relative to some level of local earnings that are unaffected by the RO wage itself and thus effectively proxies for the real RO wage. To identify the effect of ROs on wage inequality, the analy- The specification also includes a time-invariant percentile- sis follows the approach implemented in Autor, Manning, specific RO fixed effect (α q ), a time fixed effect associated and Smith (2016) and Bosch and Manacorda (2010). The r to percentile q (α q ), and a RO-specific linear time trend analysis focuses on earnings differentials across ROs that are r (δr * timet), while εrqt is the idiosyncratic component of the subject to different ROs. The model presented here defines wage percentile differential. a function for the latent wage distribution, that is, the wage that would be observed in the absence of ROs. Let us define To operationalize the estimation, preliminary evidence indicates ωq as the logarithm of wages at percentile q in RO r at rt that earnings at or above the 70th percentile are not affected time t, and let ω* q be the latent percentile. The specification rt by ROs. For this reason, in the analysis, one imposes that below consists of a censoring model in that it assumes that p = 70 as log RO wages reach well beyond the median of log individuals with latent earnings below the minimum wage actual earnings distribution in some sectors covered by ROs. are paid exactly the minimum wage, whereas those with earnings above the minimum wage are unaffected. Let p One difficulty with the ordinary least squares estimation be a percentile sufficiently high so that for percentiles of equation (3) is that any measurement error in the q-th s ≥ p wages are unaffected by ROs, that is, ω* rt s = ω* rt s . Then, percentile of the wage distribution will lead to a spurious the model can be written as follows: positive correlation between different measures of inequal- ity and the effective RO, possibly leading to upward biased (ω q rt − ω rt p ) = (ω* rt − ω* q rt ) if p ω* rt ≥ ROrt q (A.1) estimates of the effect of the minimum wage. Lacking any credible instrument, the estimates will need to be taken (ω q rt − ω rt p ) = ( ROrt − ω rt p ) if ω* rt < ROrt , q (A.2) with caution and as indicative of an upper bound of the true effect of ROs on wage inequality. where ROrt is the log wage for RO r. The set of equations above state that the q to p differential of the actual log earnings distribution for RO r will be equal to the latent differential if the latent q-th percentile is above the RO Identification Strategy: wage; otherwise, it will be equal to the difference between Remuneration Orders, the RO wage and the p-th percentile. The actual p-th percentile is substituted in place of the latent counterpart Employment, and based on the assumption that, at percentiles s ≥ p, wages Working Hours are unaffected by the ROs. The methodology combines a series of individual-level The q to p percentile gap is estimated as a function of the cross-sectional data for the years between 2004 and 2014 effective RO wage: from the Continuous Multi-Purpose Household Survey (CMPHS). The sample is restricted to include only employed (ω q rt − ω rt p ) = β1q ( ROrt − ω rt p 2 ( ROrt − ω rt ) ) + βq p 2 individuals and those with nonzero reported earnings. Self- r + α t + δ r ∗ time t + ε rqt , + αq q (A.3) employed, employers, and unemployed who have worked before are included in the analysis as the uncovered sector which consists of a quadratic function of the difference in the estimation of minimum wage effects. For this estima- between the log RO wage (ROrt) and the p-th percentile of tion, only workers who are in the covered RO sectors and the actual log earnings distribution. The quadratic term the self-employed, employers, and unemployed in or last 120 Mauritius: Addressing Inequality through More Equitable Labor Markets employed in the RO covered sectors are considered.1 To fixed. Equation (1) is similar to a difference-in-difference retrieve the effect of minimum wages on the employment model that compares employment in RO sectors wherein of workers and number of hours worked, holding constant there was a change in the level of the minimum wage with other factors that might affect wages, the following equa- employment in RO sectors where the legislated minimum tion is estimated:2 wage did not change over time. Eit = α + βLnMWit + Xit γ + δVAzt + Ds + Dt + ε it (A.4) Equation (1) is first estimated to test for an employment effect of legal minimum wages in the covered sector. Then, where the same equation is used to estimate, via ordinary least Eit =  1 if individual i is employed in the covered sec- squares, the effect of minimum wages on the number of tor at time t hours worked per week in the covered and uncovered by Eit =  0 if individual I at time t is an own-account substituting the log of hours worked for the EMP variable. worker, an employer, or an unemployed who has worked in the past One challenge to the identification strategy is that the Xit =  individual characteristics include gender, experi- time of changes in the legislated minimum wage in specific ence, educational level, and interactions RO sectors is correlated with employment levels in those LnMWit =  logarithm of real hourly RO wage that applies same sectors. However, the procedure to review an RO to individual i at time t3 of particular sector is complex and requires a significant Ds = a set of dummies for industry/occupation amount of time. The International Labour Organization Dt = a set of year dummies (ILO 2014) summarizes the process as follows: (1) trade VAzt = value added of sector z at time t unions, workers, or other stakeholders make an official request to the Minister of Labor whenever they consider The coefficient β is an estimate of the effect on employ- necessary a revision of wage levels or working condi- ment in the covered sector of changes in the RO wage.4 tions of a certain sector; (2) the minister of labor invites The RO wage is assigned to each worker based on their representatives of employers, employees, and others as RO category. It was not possible to assign specific wages necessary to a discussion; (3) based on the meetings, the based on job title and years of experience. The matching NRB designs specific questionnaires to conduct a survey exercise has taken the lowest stipulated wage in each RO with the aim of collecting data regarding wages and work- for that category of RO workers as the binding RO wage ing conditions; (4) after the fieldwork, a technical report for each RO category regardless of job title and number is prepared with the information from the survey and the of years of experience. proposal of employers’ and employees’ representatives; (5) the report is submitted to the NRB which decides A set of dummy variables for each RO category are included whether and how to proceed; (6) the report is then made to control for RO category specific fixed effects and for public and counter proposals are invited; (7) the NRB the endogenous correlation of employment and RO wages makes final recommendations to the Minister of Labor across RO categories. Value added for each broad industry taking into account all the counter proposals received; category is included to control for changes in demand over (8) the Minister accepts, rejects, amends, or refers back the time. To control for endogenous changes in yearly average recommendations of the NRB; (9) the recommendations minimum wages as well as other year-specific factors, a and the decision of the Minister goes to the Cabinet for dummy variable is included for each year. approval; and (10) after receipt of the Cabinet’s approval the matter is passed to the State Law Office in charge of Equation (A.1) can be consistently estimated through a drafting the legislation. In other words, there is a time lag in linear probability model provided there is no correlation the process that makes it unlikely that changes in legislated between the error term and the regressors. The introduction minimum wages respond promptly to contemporaneous of year and RO categories dummies allows to exploit the economic conditions. within-year within-RO variation in the level of the minimum wage to identify its effect on employment, while controlling One drawback in our estimation strategy is that, due to for a host of individual-level characteristics. Controlling data limitations, it is not possible to assign the exact RO for the level of value added at the level of broad industrial wage schedule within the broader RO category to each sector allows to capture changes in employment with output worker in the sample. The preferred choice was to take the Appendix A 121 lowest stipulated minimum wage in each RO as the relevant NOTES minimum wage for that category of RO workers. In other 1. Four RO categories are excluded, namely, banks fishermen, cinema words, it was not possible to assign specific minimum wages employees, salt workers, and travel agent employees, for which only for specific job titles and years of experience.5 a small number of covered workers (fewer than 50 covered workers identified in any year) is identified in the dataset. 2. This approach is similar to a number of approaches in the literature, With respect to the pool of unemployed who worked in the including, most notably, that of Gindling and Terrell (2007), who esti- past and who are included as part of the uncovered sector mate employment effects for a similarly complex system of minimum wages in Costa Rica. in the analysis, the limitation is that there is no complete 3. Minimum wages are set in hourly, daily, or monthly terms. Where they occupation data for all years in the period of analysis. The are not in hourly terms, they have been standardized to a real hourly minimum wage using 8 hours as the reference number of hours for a CMPHS collects the required industry data for all individu- workday and 168 hours as the reference number of hours for a work als used in the analysis. However, for the unemployed who month. All real wages are in 2012 prices. have worked before, we have occupational data about 4. Note that this specification, including self-employed and unem- ployed workers who have worked before in the covered sectors, their last job for only eight of the eleven years considered. assumes that workers who lose their jobs in the covered sectors These individuals are not in large numbers in the CMPHS either become unemployed or self-employed in those covered sectors they left. If some workers who lose their jobs in the covered sample and are more likely than not to be workers in the sector find employment in a different RO sector, these estimates occupational classes covered by ROs based on data from are affected. the years with full available information. The implication 5. See appendix E, table E.1, which shows the number of wage rates there were in 2016. Ideally, each worker would be assigned the exact of this minor data limitation is thus that, in the years 2004, minimum wage rate to fully exploit the variation in minimum wages 2006, and 2007, for which there is no occupational data in the complex system of ROs. 6. This was true for only two RO categories: tailors and office attendants. for the unemployed, the number of unemployed who had These are however not categories with a significant number of identified worked before is possibly slightly overstated.6 RO-covered workers.  123 APPENDIX B Data Sources (Box B.1) •  Continuous Multi-Purpose Household Survey (CMPHS): •  Remuneration orders: Government of Mauritius, Depart- data from 2001 to 2015 ment of Labor, Government Notices 2004–2014 •  Administrative data: work permits BOX B.1. Data Overview and Definitions of Labor Market Variables Data The main dataset employed in this analysis is derived from the Continuous Multi-Purpose Household Survey (CMPHS) for the years 2001–2015. In each wave of the survey, question- naires were administered to 11,280 households through face-to-face interviews covering about 40,000 individuals. The CMPHS is a continuous survey, strongly comparable over time in terms of labor market variables, that therefore does not pose seasonality issues. Furthermore, its design allows for time-consistent definitions of labor market indicators. Employment The definition of employment is based on three filter questions asked to all respondents aged 12 and above. Individuals are classified as employed if, during the reference period consisting of the 7 days preceding the interview, they either: 1. worked for pay, profit or family gain, even if it was only for one hour; 2. performed other activities for sale or pay; 3. were temporarily absent from a job or business because of holidays, sickness or any other reason. Unemployment Individuals are defined as unemployed if they have not worked during the reference period and they have been looking for work or trying to set up a business during the 4 weeks prior to the interview. Labor force participation All individuals who fall into either the employment or unemployment category are defined as actively participating in the labor force. Otherwise, they are classified as inactive, that is, not in the labor force. Working hours The measure of working hours employed in the analysis refers to the total number of hours (including overtime) worked at the main job during the reference period. Employment category The list of self-reported employment categories includes wage worker, employer, self- employed, contributing family worker, apprentice/intern, and other. Earnings Individual earnings, received during the last month preceding the interview, are made up of three components: 1. income from paid employment (including bonus, overtime, and so on) 2. income from self-employment (trade, business, plantation, and so on) 3. income from backyard-produced goods (vegetables, fruits, eggs, fish, and so on) Earnings are expressed in 2015 prices. Hourly wages Among wage workers, hourly wages are constructed by dividing monthly wages from the main wage job by the product of working hours (as defined above) and the maximum possible number of working weeks in a month (that is, 4.33, due to the lack of information on the number of weeks worked over the last month). Hourly wages are expressed in 2015 prices. 124 Mauritius: Addressing Inequality through More Equitable Labor Markets FIGURE B.1. Growth Incidence Curve, Total Household Income: Comparing HBS and CMPHS Data, 2007 and 2012 Growth incidence curves; 2007–2012 Household total income (per adult equivalent) 6 Annual growth rate (percent) 4 2 0 −2 −4 0 20 40 60 80 100 Percentiles CMPHS labor + property + transfer income HBS disposable income Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS) and Household Budget Survey (HBS), Statistics Mauritius.  125 APPENDIX C FIGURE C.1. Demographic and Labor Market Factors and Changes in Household Labor Income Inequality, 2001–15 1.0 6.8 P90/P10 –28.3 113.9 6.0 –7.0 –1.7 0.3 0.2 P90/P50 –87.1 165.4 36.4 –2.6 –3.3 3.4 1.2 P50/P10 –9.8 94.4 14.3 –5.3 1.9 1.7 1.3 Gini –99.9 140.0 39.5 17.6 –0.3 0.7 Variance –53.2 108.8 25.5 12.0 5.9 0.3 –150 –100 –50 0 50 100 150 200 250 Men's labor income inequality Women's participation Women's labor income inequality Assortative mating Family mix Family characteristics Residual Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius.  127 APPENDIX D FIGURE D.1. Sectorial Distribution of Wage Workers, by Gender and Main Sector, 2004–15 a. Women wage workers, public sector 20040.4 0.2 2.3 5.7 0.2 0.1 3.1 5.10.0 50.9 32.1 20050.4 0.2 0.6 0.2 2.9 0.0 3.6 6.8 0.0 52.0 33.2 20060.4 0.2 1.6 4.90.4 0.1 2.7 4.80.0 52.8 32.1 20070.3 0.1 3.1 5.40.04.7 0.0 7.2 0.0 52.4 26.9 2008 0.7 3.4 0.0 3.4 6.1 0.0 6.8 0.0 47.9 31.7 2009 1.0 1.5 5.6 0.0 0.0 3.4 7.8 0.0 51.3 29.2 2010 0.8 1.4 4.60.0 0.0 3.6 7.9 0.0 54.4 27.4 20110.4 0.1 1.1 5.5 0.0 0.0 4.0 6.1 0.0 52.5 30.4 2012 0.8 0.6 1.9 0.0 8.7 1.04.2 8.3 0.0 44.6 29.9 2013 1.0 0.3 2.5 5.10.0 6.0 2.6 0.4 0.0 49.9 32.3 2014 0.6 0.2 0.22.6 6.6 0.1 2.9 7.1 0.0 50.8 28.8 2015 0.7 0.0 1.3 0.1 7.6 0.0 2.6 7.4 0.0 51.9 28.4 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Textile manufacturing Other manufacturing Trade and transports Hotels and restaurants Information and communication Household activities Other services Public administration Other secondary Professional activities (continued) FIGURE D.1. Sectorial Distribution of Wage Workers, by Gender and Main Sector, 2004–15 (continued) b. Women wage workers, private sector 2004 5.5 37.2 10.2 0.8 13.1 7.0 1.7 6.5 10.0 8.1 0.0 2005 6.0 34.3 9.8 0.9 14.7 7.9 2.0 7.3 8.6 8.6 0.0 2006 4.8 31.3 10.3 0.7 15.3 9.1 2.6 7.8 9.3 8.8 0.0 2007 4.5 29.1 9.8 0.8 15.9 9.5 3.0 8.3 9.9 9.2 0.0 2008 4.1 28.3 9.5 1.3 15.0 8.6 2.0 11.3 9.4 10.4 0.0 2009 4.8 25.1 9.4 1.3 14.7 10.2 1.6 13.5 9.3 10.2 0.0 2010 4.1 22.8 8.4 1.3 16.2 10.8 2.2 13.8 9.9 10.5 0.0 2011 3.8 20.1 10.0 1.4 19.3 9.9 1.1 13.9 9.7 10.6 0.2 2012 3.2 18.5 10.7 1.1 21.3 10.4 1.5 13.8 9.8 9.6 0.1 2013 2.7 17.0 9.3 1.1 20.4 12.3 3.2 12.7 10.1 11.2 0.0 2014 2.6 16.4 9.2 1.7 19.9 10.8 1.9 13.6 11.9 12.1 0.0 2015 2.3 14.2 10.2 1.3 19.0 11.1 2.2 16.2 10.5 12.8 0.1 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Textile manufacturing Other manufacturing Trade and transports Hotels and restaurants Information and communication Household activities Other services Public administration Other secondary Professional activities c. Men wage workers, public sector 2004 4.0 0.0 13.6 11.7 0.3 2.4 7.6 0.0 18.0 42.3 2005 3.2 0.3 0 .0 15.0 11.3 2.5 0.2 7.8 0.0 17.3 42.3 2006 3.4 0.4 0.0 14.3 10.6 0.1 2.6 6.3 0.0 19.3 43.0 2007 2.3 0.4 0.0 12.4 10.7 0.1 2.6 6.7 0.0 22.7 42.0 2008 3.1 0.3 0 .0 12.6 12.8 2.4 0.0 7.4 0.0 21.6 39.9 2009 2.4 0.4 0.0 13.5 12.5 0.0 2.5 6.3 0.0 20.2 42.2 2010 2.8 0.7 0 .0 12.5 11.7 2.7 0.0 6.7 0.0 19.2 43.7 2011 2.71.1 6.6 0.0 13.8 3.2 4.50.0 0.0 19.6 48.4 2012 3.0 0.9 0 .1 9.6 15.0 0.83.3 4.50.0 19.2 43.6 2013 2.0 0.5 0.0 10.5 13.1 0.1 3.5 3.1 0.0 20.0 47.2 2014 2.4 0.4 0.1 9.6 13.5 0.2 3.5 4.0 0.0 18.3 48.1 2015 3.3 0.5 0.1 9.4 13.4 0.1 3.4 5.40.0 18.2 46.1 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Textile manufacturing Other manufacturing Trade and transports Hotels and restaurants Information and communication Household activities Other services Public administration Other secondary Professional activities (continued) FIGURE D.1. Sectorial Distribution of Wage Workers, by Gender and Main Sector, 2004–15 (continued) d. Men wage workers, private sector 2004 9.5 10.9 17.0 16.8 19.2 10.6 1.2 9.6 1.14.2 0.0 2005 8.8 9.1 17.3 18.1 18.1 11.6 1.3 10.5 1.14.0 0.0 2006 8.6 9.7 17.8 16.7 18.2 12.0 1.6 9.4 1.2 4.8 0.0 2007 9.3 9.2 17.6 18.7 18.0 11.6 1.9 8.9 0.83.8 0.0 2008 7.9 8.2 17.0 19.0 18.2 11.7 1.8 10.9 0.74.6 0.0 2009 7.4 7.2 15.8 18.9 19.6 11.8 1.6 11.9 1.3 4.5 0.0 2010 7.3 6.3 15.8 19.2 19.7 11.6 1.6 12.5 1.1 5.0 0.0 2011 7.5 5.9 16.2 18.1 19.8 11.6 1.6 12.9 1.1 5.3 0.0 2012 6.4 5.2 16.2 18.8 21.3 11.7 1.7 12.7 1.0 4.8 0.1 2013 5.9 5.1 14.5 18.7 23.6 12.8 2.1 11.4 1.3 4.6 0.0 2014 6.1 5.2 13.5 17.9 22.7 12.7 2.3 13.1 1.1 5.1 0.1 2015 6.7 4.4 15.1 15.6 21.4 12.7 2.8 15.2 1.2 4.8 0.0 0 10 20 30 40 50 60 70 80 90 100 Percent Agriculture Textile manufacturing Other manufacturing Trade and transports Hotels and restaurants Information and communication Household activities Other services Public administration Other secondary Professional activities Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius. FIGURE D.2. Educational Distribution of all Wage Workers, 2004–15 2004 39.6 11.7 41.1 3.6 4.0 2005 38.7 11.2 41.5 4.2 4.4 2006 37.7 11.3 42.0 4.3 4.7 2007 36.0 11.3 43.1 4.5 5.1 2008 35.6 11.3 42.9 4.3 5.8 2009 34.6 11.5 42.9 4.2 6.7 2010 32.8 11.4 43.5 4.8 7.5 2011 30.7 11.7 38.7 6.3 12.6 2012 29.9 10.9 36.6 9.4 13.1 2013 29.1 10.5 35.5 10.7 14.1 2014 27.0 10.2 36.0 10.5 16.3 2015 24.9 10.7 37.1 9.9 17.4 0 10 20 30 40 50 60 70 80 90 100 Percent Up to Complete Primary Upper Secondary Lower Secondary Post−Secondary Tertiary Source: Based on data of the Continuous Multi-Purpose Household Survey (CMPHS), Statistics Mauritius.  131 APPENDIX E FIGURE E.1. Earned Mean Wages and Legislated Mean RO Wages in Covered Sectors, 2004–14. 10,000 8,000 6,000 4,000 2,000 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Earned minimum wages Legislated minimum wages Source: CMPHS data and Government of Mauritius, Department of Labour, Government Notices 2004–2014. Note: Wages are measured in 2012 prices. 132 Mauritius: Addressing Inequality through More Equitable Labor Markets TABLE E.1. Number of Wage Rates Specified within Remuneration Orders, 2016 Number Total number Corresponding broad industry of job title Years of experience categories per of wage rates RO category sector categories job title specified Attorney and notary Finance and professional services 2 15 years (clerk), 20 years (secretary) 35 employees (administration and support) Baking industry Manufacturing 24 3 job titles specified by years of 39 experience; up to 8 years Banks fisherman and Agriculture, forestry and fishing 2 NA 2 frigo-workers Block-making and Construction 26 Most job titles specified by year; 100 construction (Possibly, mining and quarrying) up to 8 years Catering and tourism Transport, food, accommodation and 52 Most job titles specified by year: 196 ICT (food and accommodation); up to 7 years; 4 years for most CSP (Arts, entertainment and recreation) Cinema employees CSP (Arts, entertainment and 12 NA; rates for some by number of 19 recreation) shows per month Cleaning enterprises Across all sectors 13 8 years for all 104 Retail trades Wholesale and retail 34 8 years for all 272 Domestic workers Private households 8 NA 8 Electrical, engineering, Manufacturing 11 Most job titles specified by year; 63 mechanical up to 7 years Export enterprises Manufacturing 11 Most job titles specified by year; 47 up to 9 years Factory employees Manufacturing 10 Most job titles specified by year; 70 up to 8 years Field crop and orchard Agriculture, forestry and fishing 6 NA 6 workers Light metal and wooden Manufacturing 11 All job titles specified by year; up to 64 furniture 8 years Livestock workers Agriculture, forestry and fishing 4 Supervisor specified by year (5 years) 8 Newspapers and Manufacturing 10 All job titles specified by year; up to 90 periodicals 15 years Nursing homes CSP (Health) 16 Most job titles specified by year; 74 up to 10 years Office attendants Finance and professional services 2 Both job titles specified by year 20 (admin and support); CSP (admin) (10 years) Pre-primary school CSP (Education) 7 Most job titles specified by year; 26 employees up to 10 years Printing industry Manufacturing 16 Most job titles specified by year; 85 up to 10 years Private secondary school CSP (Education) 13 Most job titles specified by year; 111 up to 15 years Public transport (buses) Transport, food, accommodation and 35 Most job titles specified by year; 322 ICT (transport and storage) up to 10 years Road haulage industry Transport, food, accommodation and 6 All job titles specified by year; 8 years 48 ICT (transport and storage) Salt manufacturing Manufacturing 7 NA 7 industry Security guards Across all sectors 1 8 years 8 (continued) TABLE E.1. Number of Wage Rates Specified within Remuneration Orders, 2016 (continued) Number Total number Corresponding broad industry of job title Years of experience categories per of wage rates RO category sector categories job title specified Sugar industry Agriculture, forestry and fishing 18 NA 18 (agricultural) Sugar industry Manufacturing 35 Many job titles specified by year; 58 (nonagricultural) up to 6 year. Some job titles further categorised acc to grade Tailoring trade Manufacturing 5 Learner specified by year; 5 years 9 Tea industry Agriculture, forestry and fishing, 32 NA 32 manufacturing Travel agents and tour Finance and professional services 17 All except Trainee and Watchman 62 operators (administration and support) specified by year; 4 years Total 446 2003 Source: Government of Mauritius, Department of Labor, Government Notices 2004–2014. Note: NA = not available. TABLE E.2. Estimation Approach to Determining RO Worker Coverage Using CMPHS Data Relevant occupations and/or Remuneration order broad occupation groups* Relevant activity, industry or sector** Attorney and notary employees (RO) Specific occupations: various legal Attorney, notary regulations (last revised 2010) clerk and legal secretary related occupations Baking industry (RO) regulations Craft workers, operators and assemblers Bread (with or without pastry), pastries and (last revised 2003) cakes, biscuits, other bakery products Banks fisherman and frigo-workers (RO) Craft workers, operators and assemblers, Banks fishers regulation (last revised 1997) elementary Block-making, construction, stone crushing Technicians, clerks, craft workers, Construction (industry), stone, stone-crushing, and related industries (RO) regulations operators and assemblers manufacture of articles of cement, stone (last revised 2008) cutting, shaping, and finishing Catering and tourism industries (RO) Clerks, services/sales, craft, operators/ Various hotel, accommodation, restaurant and regulations (last revised 2004) assemblers, elementary; specific food related activities occupations: chefs, skippers, masseurs, gardeners, entertainers Cinema employees (RO) regulations Technicians, service/sales, operators/ Motion picture projection (last revised 2005) assemblers, elementary Cleaning enterprises (RO) regulations Clerks, operators/assemblers, Refuse disposal, building-cleaning activities, (last revised 2013) elementary cleaning services, care and maintenance activities Distributive trades (RO) regulations Clerks, service/sales, operators/ Wholesale and retail (industry) (last revised 2007) assemblers, elementary Domestic workers (RO) regulations Service/sales, operator/assembler, Private households (last revised 2010) elementary; specific occupations: gardener Electrical, engineering and mechanical Clerks, craft, operators/assemblers, Various maintenance and repair-related workshops (RO) regulations elementary activities (last revised 2013) Export enterprises (RO) regulations Clerks, operators/assemblers, specific A number of activities that are export oriented. (last revised 2003) occupation: cashier, watchman These include activities which fall under the general scope of the following: yarn and thread spinning, weaving and dyeing, knitting, fabrics, textiles and garments Factory employees (RO) regulations Clerks, operators/assemblers, A number of activities which have a substantial (last revised 2001) elementary; specific occupation: amount of factory workers (apart from watchman those covered by the export enterprises RO above). These include activities which fall under the general scope of the following: clothing, jewelry, fish processing, and chemical manufacturing (continued) 134 Mauritius: Addressing Inequality through More Equitable Labor Markets TABLE E.2. Estimation Approach to Determining RO Worker Coverage Using CMPHS Data (continued) Relevant occupations and/or Remuneration order broad occupation groups* Relevant activity, industry or sector** Field-crop and orchard workers (RO) Elementary Various crop, flower and fruit related activities regulations (last revised 2008) Light Metal and Wooden Furniture Clerks, craft, operators/assemblers; Manufacture of bodies for motor vehicles, Workshops (RO) Regulations (Last specific occupation: watchman furniture (wooden), furniture (metal), revised 2002) furniture (other, not plastic) Livestock workers (RO) regulations (last Elementary; specific occupations: Various livestock related activities revised 2008) livestock farmer Newspapers and periodicals employees Professionals, clerks, operators/ Various publishing related activities (RO) regulations (last revised 2001) assemblers; specific occupation: cashier Nursing homes (RO) regulations (last Technicians, service/sales, operators/ Hospital activities – private hospitals, revised 1990) assemblers, elementary; specific residential nursing care activities occupations: receptionist, gardener Office attendants (RO) regulations (last Specific occupation: office attendant revised 2013) Preprimary school employees (RO) Specific occupations: teacher, cook, Preprimary education regulations (last revised 2000) gardener, handyman, caretaker Printing industry (RO) regulations (last Clerks, craft, operators/assemblers, Printing, service activities related to printing, revised 2003) elementary; specific occupation: printing of labels, printing on metals watchman Private secondary-school employees (RO) Specific occupations: education officer General secondary education regulations (last revised 1984) (teacher), typist, secretary, librarian, gardener, cleaner, caretaker Public transport (buses) workers (RO) Clerks, service/sales, craft, operators/ Bus transport regulations (last revised 2008) assemblers, elementary Road haulage industry (RO) regulations Operators/assemblers Freight transport by road: lorry, van, other (for (last revised 2009) example, handcarts) Salt-manufacturing industry (RO) Craft, operators/assemblers, elementary Salt extraction regulations (last revised 1994) Security guards (RO) regulations (last Specific occupations: various security Investigation and security activities, private revised 1997) guard and watchman-related security activities occupations Sugar industry (agricultural workers) (RO) Elementary; specific occupations: Sugarcane regulations (last revised 2010) watchman, gardener Sugar industry (nonagricultural workers) Clerks, craft, operators/assemblers Manufacture of sugar (RO) regulations (last revised 2010) Tailoring trade (RO) regulations (last revised Specific occupations: various tailoring- 2002) specific occupations Tea industry workers (RO) regulations (last Clerks, operators/assemblers, Tea revised 1992) elementary; specific occupations: watchman Travel agents and tour operator workers Clerks, service/sales, operators/ Activities of travel agencies, tour operators, (RO) regulations (last revised 2009)µ assemblers, elementary tourist assistance activities, tour operator activities Note: * Note the following NASCO occupation codes are associated with the broad occupation categories listed in the table above: Codes starting with 1 denote managers, codes starting with 2 denote professionals, codes starting with 3 denote technicians, codes starting with 4 denote clerks, codes starting with 5 denote sales/service workers, codes starting with 6 denote skilled agricultural workers, codes starting with 7 denote craft workers, codes starting with 8 denote operators/assemblers and l codes starting with 9 denote elementary workers. ** For some ROs the occupation matched code was sufficient to identify the group of workers (for example, “office attendants”). In these cases, we only matched on that code and not an activity or sector code as well. In the cases where we could not just match on the occupation code because the code was still too broad, we matched on an activity, industry or sector/establishment type as well. So if, for example, the job title to be matched was “Accounting Clerk” in the export oriented enterprise (EOE) RO, we would first match on the occupation code for Accounting Clerk and then on the activity codes for EOE to capture accounting clerks working within EOEs. We also used an activity, industry or sector to isolate workers where we more broadly estimated the workers covered by ROs by broad occupation types. µ The Travel Agents and Tour Operators RO regulation first came into effect in 2009. We thus only include the relevant occupations as being covered by a RO in the years 2009 to 2014. All other ROs have been in effect for the full 2004 to 2014 period for which we have access to CMPHS data. Appendix E 135 TABLE E.3. The Real Hourly Minimum Wage, by Remuneration Order, 2004–14 MUR 2012 Job title of minimum wage stipulated in Remuneration order RO 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Attorney and notary Clerk / secretary 46.14 46.40 43.73 42.47 40.79 41.14 41.37 43.92 45.03 45.36 45.62 employees Baking industry Handpacker 29.13 29.48 27.97 27.55 26.79 27.23 27.39 26.58 27.29 27.75 28.23 Banks fisherman and Frigo-worker 38.41 38.65 36.39 35.29 33.84 33.01 32.07 30.11 28.99 28.00 27.13 frigo-workers Block-making, Learner 40.59 41.08 38.84 37.98 43.64 43.92 44.16 42.54 43.67 43.99 44.29 construction, stone crushing, and related industries Catering and tourism Trainee 30.43 35.51 33.73 33.28 32.41 32.97 33.16 32.21 33.07 33.66 34.28 industries Cinema employees Cleaner (04) / café 26.01 21.10 20.34 20.33 20.03 20.53 20.65 20.01 21.23 22.23 23.21 assistant (05–14) Cleaning enterprises Vehicle attendant 34.19 34.60 32.89 32.51 31.71 32.29 32.47 31.46 32.29 31.19 39.46 Distribution trades Attendant / cleaner 41.35 41.84 39.54 38.62 37.28 37.72 37.93 36.69 37.66 38.10 38.58 Domestic workers Household helper 19.65 20.26 19.54 19.52 19.23 19.72 19.83 25.68 26.69 27.50 28.32 (04–10) / household worker (11–14) Electrical, engineering, Apprentice 28.98 29.33 28.05 28.07 27.66 28.34 28.50 27.61 28.55 29.30 34.69 and mechanical workshops Export enterprises Unskilled worker 17.18 17.79 17.15 17.13 16.88 17.31 17.41 16.87 18.06 19.03 19.98 Factory employees Unskilled worker 22.95 23.29 22.42 22.40 22.06 22.62 22.75 22.04 23.04 23.84 24.64 Field crop and orchard Women worker 32.02 32.42 30.88 30.67 33.91 34.43 34.62 33.59 34.48 35.02 35.60 workers Light metal and Apprentice 22.40 22.68 21.73 21.81 21.56 22.13 22.25 21.56 22.35 22.97 23.61 wooden furniture workshops Livestock workers Young worker (04–08) 28.48 28.83 27.59 27.64 27.27 34.43 34.62 33.59 34.48 35.02 35.60 / grade II (09–14) Office attendants Office attendant 43.72 44.10 41.61 40.52 39.02 39.41 39.63 38.29 39.30 39.68 48.19 Newspaper and Receptionist 43.43 43.83 41.36 40.30 38.81 39.21 39.43 38.10 39.11 39.49 39.93 periodical employees Nursing homes Attendant / kitchen 26.33 26.66 25.60 25.81 25.43 26.07 26.22 25.41 26.43 27.25 28.07 help Printing industry Unskilled worker 40.54 41.04 38.80 37.94 36.66 37.12 37.32 36.12 37.08 37.53 36.36 Preprimary school Handyman 24.63 25.01 24.08 24.06 23.70 24.30 24.44 23.68 24.76 25.64 24.84 employees Private secondary- Caretaker 30.14 30.51 29.14 29.06 28.57 29.22 29.39 28.47 29.38 30.10 30.83 school employees Public transport (buses) Apprentice 28.75 29.10 27.84 27.87 36.44 36.91 37.11 35.92 36.88 37.34 47.34 workers Road haulage industry Lorry attendant 36.45 36.90 35.00 34.45 33.48 32.66 38.48 37.20 38.19 38.61 39.08 (04–08) / tanker assistant (10-) Salt-manufacturing Woman worker 23.37 23.66 22.63 22.64 22.32 22.86 22.99 22.28 23.04 23.64 24.26 industry (continued) 136 Mauritius: Addressing Inequality through More Equitable Labor Markets TABLE E.3. The Real Hourly Minimum Wage, by Remuneration Order, 2004–14 (continued) MUR 2012 Job title of minimum wage stipulated in Remuneration order RO 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Security guards Security guard 48.87 49.10 46.21 44.75 42.86 43.17 43.41 41.83 42.94 43.26 43.58 Sugar industry Fieldworker (special) 25.80 26.13 25.11 25.36 24.99 25.62 25.77 24.97 26.01 26.84 27.68 (agricultural) Sugar industry Messenger / 44.25 44.60 42.07 40.95 39.40 39.79 40.01 38.65 39.67 40.04 40.46 (nonagricultural) weighbridge Tailoring trade Learner 20.06 20.65 19.91 19.89 19.60 20.10 20.21 19.58 20.82 21.83 22.82 Tea industry workers Young person 28.48 28.83 27.59 27.64 27.27 27.96 28.12 27.24 28.20 28.96 29.72 Travel agents and tour Cleaner / vehicle 39.35 39.57 38.23 39.24 39.63 40.06 operator workers attendant Source: National Remuneration Board, Remuneration Orders 2004–2014.