New Century, Old Disparities Gender Wage Gaps in Latin America Hugo Ñopo (based on work with Juan Pablo Atal, Alejandro Hoyos, and Natalia Winder) The One Slide Presentation What is this paper about? » » Harmonized and comparable measures of gender and ethnic wage gaps for 18 countries in the region A refined answer to the old question: » » To what extend gender and ethnic differences in individuals’ characteristics can explain the differences in earnings? Findings? New Insights? » Gender wage gaps between 8% and 28%, dropping around 4 points in 15 years (1992-2007). » Higher gaps among older people, those with lower income, with secondary education, self-employed, informal and those in small firms » Surprising role for occupational and sector segregation » An across-the-board reduction over the last decade, but especially among those with kids at home, parttime workers and those previously found with higher gaps Methodological improvements: » Matching and a decomposition that recognizes not only differences on average characteristics but also on their distribution; and most importantly, on their supports Outline » Setup of the Problem. Methodological considerations » Blinder-Oaxaca decompositions » Matching » Combining the two tools » Empirical Results. LAC (circa 2005; 1992-2007) » The Data » For Ethnic and Gender Gaps » » » » For Gender Gaps » » Averages Distributions The role of occupational and sector segregation Evolution Conclusions 1. Setup of the Problem. Methodological considerations »Blinder-Oaxaca decompositions »Matching »Combining the two tools Gender Differences in: » Wages » Individual Characteristics » Age » Education » Individual Characteristics » » » » Urban and Rural Area Presence of children in the HH Presence of other income earner in the HH Job Characteristics » Occupation » Sector » Type of employment » Part – time » Formality » Firm size Blinder-Oaxaca Decompositions » The wage gap is separated into two additive components » » One attributable to the existence of differences in the average characteristics of females and males The other attributable to the existence of differences in the rewards that females and males get for the same characteristics » » Discrimination Unobservable characteristics Blinder-Oaxaca Decomposition. Linear Setup Critiques » Recent data violates key implications of the Mincerian model » Hansen (1999) » Heckman, Lochner and Todd (2001) » B-O is informative only about the average decomposition, no clues about the distribution of the components » Jenkins (1994) » DiNardo, Fortin and Lemieux (1996) » Donald, Green and Paarsch (2000) » Bourguignon, Ferreira and Leite (2008) » The comparison should be restricted only to comparable individuals. The failure to recognize this fact may bias the estimates in the gap decomposition » Barsky, Bound, Charles and Lupton (2001) » The relationship governing characteristics and wages is not necessarily linear Matching. Impact Evaluation » » » Treatment effects Identification of counterfactual situations Extensively used in the Program Evaluation literature » » » » » Rubin (1977) Heckman, Ichimura and Todd (1998) Heckman, LaLonde and Smith (1998) Angrist (1998) Dehejia and Wahba (1998) The Main Counterfactual Question What would the distribution of earnings for males be, in the case that their individual characteristics follow the distribution of the characteristics for females? The Matching Algorithm For each possible value of the vector of characteristics x: » Select all females with these characteristics nF(x) » Select all males with these characteristics nM(x) » If nF(x)=0 and nM(x)>0 unmatched males » If nF(x)>0 and nM(x)=0 unmatched females » If nF(x)>0 and nM(x)>0 reweight: » Each female with 1 » Each male with nF(x)/nM(x) Maids CEOs The Matching Algorithm Result: A sample of matched females and males with the same distribution of observable individual characteristics (but not necessarily the same distribution of earnings). A sample of unmatched females and another of unmatched males This Matching Approach is… A non-parametric alternative to B-O decompositions that has advantages in terms of: » » » » Simplicity Avoiding the estimation of earnings equations Flexibility It “contains” all possible propensity scores Identification/Correct specification Recognizing that the supports of empirical distributions of characteristics do not completely overlap (the failure to recognize this leads to an overestimation of the unexplained component of the wage gap) Information Allowing us to compute directly the distribution of the unexplained effects, not just the average The New Decomposition and Matching 2. Empirical Results. LAC (circa 2005) The Data Gender Gaps Averages Distributions The role of occupational and sector segregation The Data 2006 Number of Observations* 41,287 31 urban regions Encuesta Continua de Hogares (ECH) 2006 4,959 National Brasil Pesquisa Nacional por Amostra de Domicilio (PNAD) 2007 133,764 National Chile Encuesta de Caracterizacion Socioeconomica Nacional (CASEN) 2006 85,968 National Colombia Encuesta Continua de Hogares (ENH) 2005 52,388 National Costa Rica Encuesta de Hogares de Propositos Multiples (EHPM) 2006 13,810 National Dominican Republic Encuesta Nacional de Fuerza de Trabajo (ENFT) 2003 9,718 National Ecuador Encuesta de Empleo, Desempleo y Subempleo (ENEMDU) 2007 15,611 National Guatemala Encuesta Nacional de Condiciones de Vida (ENCOVI) 2006 18,865 National Honduras Encuesta Permanente de Hogares de Propositos Multiples (EPHPM) 2007 23,278 National Mexico Encuesta Nacional Empleo (ENE), Segundo Trimestre 2004 131,348 National Nicaragua Encuesta Nacional de Hogares sobre medicion de Niveles de Vida (EMNV) 2005 9,838 National Panama Encuesta de Hogares (EH) 2003 17,368 National Paraguay Encuesta Permanente de Hogares (EPH) 2006 5,592 National Peru Encuesta Nacional de Hogares (ENAHO) 2006 27,665 National El Salvador Encuesta de Hogares de Propositos Multiples (EHPM) 2005 16,856 National Uruguay Encuesta Continua de Hogares (ECH) 2005 20,351 Urban Venezuela Encuesta de Hogares Por Muestreo (EHM), Segundo Semestre 2004 47,880 National Country Name Of The Survey Year Argentina Encuesta Permanente de Hogares (EPH), Segundo Semestre Bolivia * Workers between 18 and 65, after eliminating observations with incomplete data or outliers in wage Coverage The pooled data set » » » » Covering all Latin American countries (except rural Argentina and Uruguay) Use of expansion factors, so the size of the economies are properly represented (all but Mexico) Income measures are normalized to 2002 PPP USD, deflated by nominal GDP After that, average females (minorities) income is normalized to one Relative Wages by Characteristics All Age 18 to 24 25 to 34 35 to 44 45 to 54 55 to 65 Education None or Primary Incomplete Primary Complete or Secondary Incomplete Secondary Complete or Tertiary Incomplete Tertiary Complete Presence of children (<12) in the household No Yes Presence of other household member with labor income No Yes (Base: Average female wage = 100) Male Female 110.00 100.00 (Base: Average minority wage = 100) Non Minority Minority 137.78 100.00 79.62 106.57 122.45 127.15 113.02 74.94 100.90 108.72 111.30 97.84 98.44 133.62 149.45 159.80 151.24 77.86 98.18 109.45 113.49 100.08 73.06 95.27 141.67 201.99 71.08 75.98 118.10 178.94 108.72 113.36 155.67 223.68 74.67 90.79 127.12 160.16 117.03 102.20 105.04 95.92 144.65 130.73 104.40 96.32 108.78 110.81 101.95 99.40 140.48 136.67 96.32 101.90 Source: Authors’ calculations using Household Surveys circa 2005. Relative Wages by Characteristics All Urban No Yes Type of Employment Employer Self - Employed Employee Part time No Yes Formality No Yes Small fim No Yes (Base: Average female wage = 100) Male Female 110.00 100.00 (Base: Average minority wage = 100) Non Minority Minority 137.78 100.00 91.34 116.76 92.45 101.60 92.47 145.73 67.96 108.13 195.34 95.94 109.59 180.11 88.81 101.53 264.33 134.96 130.84 215.35 95.12 97.81 105.04 158.32 92.22 123.55 133.00 169.18 94.31 132.72 95.81 128.38 86.82 116.70 113.45 160.03 83.90 120.98 115.90 85.28 113.72 78.13 152.10 122.92 113.79 87.60 Source: Authors’ calculations using Household Surveys circa 2005. Relative Wages by Characteristics All Occupation Professionals and technicians Directors and upper management Administrative personnel and intermediary level Merchants and sellers Service workers Agricultural workers and similar Non-agricultural blue-collars, drivers and similar Armed forces Occupations not classified above Economic Sector Agriculture, Hunting, Forestry and Fishing Mining and Quarrying Manufacturing Electricity, Gas and Water supply Construction Wholesale and Ratail Trade and Hotels and Restorants Transport, Storage Financing Insurance, Real Estate and Business Services Community, Social and Personal Services (Base: Average female wage = 100) Male Female 110.00 100.00 (Base: Average minority wage = 100) Non Minority Minority 137.78 100.00 208.68 212.50 134.02 106.60 93.43 63.41 95.59 105.58 110.52 182.18 176.66 107.66 93.28 70.87 80.37 70.41 116.23 89.93 236.98 271.72 136.48 117.47 94.99 85.29 126.07 409.12 170.26 180.33 210.97 114.02 102.22 79.85 57.69 102.14 260.14 161.39 59.13 144.27 115.51 153.89 97.33 106.62 115.73 150.50 153.91 54.03 175.90 85.42 165.60 109.31 88.84 125.02 149.12 110.13 87.64 195.63 136.94 178.43 124.16 132.34 158.21 196.78 153.21 58.31 144.83 103.91 151.34 94.51 102.71 129.27 143.38 112.32 Source: Authors’ calculations using Household Surveys circa 2005. Descriptive statistics Age Education (%) None or Primary Incomplete Primary Complete or Secondary Incomplete Secondary Complete or Tertiary Incomplete Tertiary Complete Presence of children in the household (%) Presence of other member with labor income (%) Urban (%) Type of Employment (%) Employer Self - Employed Employee Part time (%) Formality (%) Small fim (%) Source: Authors’ calculations using Household Surveys circa 2005. Non Minority Minority 37.04 36.37 Men Women 37.05 36.59 20.90 44.51 29.08 5.51 47.41 60.18 73.40 15.89 37.60 37.96 8.56 55.26 76.41 82.53 14.93 38.65 38.39 8.04 49.34 70.81 85.08 24.82 42.99 27.61 4.59 54.49 65.96 79.76 4.93 27.96 67.11 9.30 43.56 52.39 2.30 26.22 71.49 24.84 44.11 54.22 4.46 24.06 71.48 13.21 52.23 49.23 2.51 28.19 69.30 14.81 43.42 70.08 Descriptive statistics Age Occupation (%) Professionals and technicians Directors and upper management Administrative personnel and intermediary level Merchants and sellers Service workers Agricultural workers and similar Non-agricultural blue-collars, drivers and similar Armed forces Occupations not classified above Economic Sector (%) Agriculture, Hunting, Forestry and Fishing Mining and Quarrying Manufacturing Electricity, Gas and Water supply Construction Wholesale and Ratail Trade and Hotels and Restorants Transport, Storage Financing Insurance, Real Estate and Business Services Community, Social and Personal Services Non Minority Minority 37.04 36.37 Men Women 37.05 36.59 9.62 3.32 5.02 9.16 11.84 15.55 32.04 0.77 12.67 15.10 2.76 10.52 17.19 32.52 7.05 9.41 0.08 5.39 13.62 4.75 9.61 12.39 18.96 11.96 27.59 0.01 1.11 8.48 2.27 6.48 11.40 24.30 16.69 29.00 0.00 1.38 18.07 0.95 16.70 0.85 12.08 20.96 8.97 3.10 18.32 3.78 0.14 15.31 0.22 0.79 27.86 1.94 3.10 46.86 12.21 0.78 16.83 0.64 7.30 23.98 6.56 3.72 27.99 16.92 0.70 14.49 0.50 9.63 21.89 5.35 1.70 28.83 Source: Authors’ calculations using Household Surveys circa 2005. Gender Wage Gap Decompositions ∆ ∆0 ∆M ∆F ∆X % Men in CS % Women in CS Age + Education + Presence of children in the HH 10.00% 8.88% 0.00% 0.00% 1.11% 100.00% 100.00% 10.00% 17.16% 0.07% -0.03% -7.20% 99.80% 99.94% 10.00% 17.40% 0.20% -0.11% -7.49% 99.28% 99.78% Source: Authors’ calculations using Household Surveys circa 2005. + Presence of other income earner in the HH + Urban 10.00% 17.93% 0.24% -0.38% -7.80% 97.66% 99.12% 10.00% 18.80% -0.28% -0.58% -7.94% 94.67% 97.90% Gender Wage Gap Decompositions by Job Related Characteristics ∆ ∆0 ∆M ∆F ∆X % Men in CS % Women in CS Demographic Set & type of empl. & Part-time & Formality & Sector & Occupation & Small firm Full Set 10.00% 18.80% -0.28% -0.58% -7.94% 94.67% 97.90% 10.00% 17.23% 1.10% -1.19% -7.14% 87.25% 95.12% 10.00% 27.30% -0.29% -2.03% -14.98% 91.26% 93.46% 10.00% 17.99% -0.14% -1.03% -6.82% 90.82% 96.36% 10.00% 23.59% -5.02% -0.33% -8.25% 64.26% 87.96% 10.00% 16.84% -0.82% -1.12% -4.89% 72.96% 86.79% 10.00% 18.83% -0.19% -0.88% -7.75% 90.75% 96.28% 10.00% 19.47% -2.02% -2.92% -4.53% 27.26% 44.71% Source: Authors’ calculations using Household Surveys circa 2005. Unexplained Gender Wage Gaps by country [%] Argentina Bolivia Brasil Chile Colombia Costa Rica Dominican Republic Ecuador Guatemala Honduras Mexico Nicaragua Panama Peru Paraguay El Salvador Uruguay Venezuela Latin America 0.5 -5.5 20.5 10.9 -0.9 -5.8 -3.1 -3.2 -3.3 5.6 2.6 1.5 -8.6 18.3 6.2 3.3 5.7 0.4 10.0 + Presence of + Part-time, Formality, children in the HH, Occupation, Economic Age and Presence of other Sector, Type of education [%] income earner in the Employment and Small HH and Urban [%] Firm [%] 14.2 -1.8 29.7 19.3 7.1 13.7 16.6 16.4 0.3 16.3 7.8 20.3 13.6 19.4 16.0 11.9 26.3 13.9 17.2 * * * * * * * * * * * * * * * * 12.6 3.0 31.4 18.6 6.3 13.6 17.3 13.6 -0.7 16.3 10.5 19.3 16.2 25.9 13.8 16.0 27.5 13.8 18.8 * * * * * * * * * * * * * * * * 10.8 17.8 26.4 13.1 7.3 17.9 23.9 5.6 17.7 24.2 15.3 28.4 10.4 23.5 6.9 11.3 23.4 12.3 19.5 Source: Authors’ calculations using Household Surveys circa 2005. *Statistically different than zero at the 99% level †Statistically different than zero at the 95% level * * * * * * * * * * † * * * * Gender Wage gap Decompositions by Country NIC (∆=1.5%) BRA (∆=20.5%) HON (∆=5.6%) DOM (∆=-3.1%) PER (∆=18.3%) URU (∆=5.7%) CRI (∆=-5.8%) BOL (∆=-5.5%) GUA (∆=-3.3%) MEX (∆=2.6%) CHL (∆=10.9%) VEN (∆=0.4%) SLV (∆=3.3%) ARG (∆=0.5%) PAN (∆=-8.6%) COL (∆=-0.9%) PRY (∆=6.2%) ECU (∆=-3.2%) -60% ∆0 ∆M ∆F -40% -20% 0% 20% 40% 60% ∆X Venezuela Uruguay El Salvador Paraguay Peru Panama Nicaragua Mexico Honduras Guatemala Ecuador Dom. Republic Costa Rica Colombia Chile Brasil Bolivia Argentina Percentage of Femlae Wage Confidence Intervals for the Unexplained Gender Gap 35% 30% 25% 20% 15% 10% 5% 0% -5% -10% Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% Females have more schooling, but they do not earn more 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment + Part Time Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment + Part Time + Formality Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment + Part Time + Formality + Occupation Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment + Part Time + Formality + Occupation + Sector Unexplained Gender Wage Gaps by Percentiles of the Wage Distribution 70% 60% 50% 40% 30% 20% 10% 0% 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 Wage Percentile Age + Education + Presence of Children in HH + Presence of other wage earner in HH + Urban + Type of Employment + Part Time + Formality + Occupation + Sector + Small Firm Unexplained Gender Gaps Unexplained Gender Gaps Unexplained Gender Gaps Unexplained Gender Gaps Unexplained Gender Gaps Unexplained Gender Gaps The Role of Job Tenure The Role of Job Tenure II A look at the evolution of Gender Wage Gaps ∆ ∆0 ∆M ∆F ∆X % CS Males % CS Females ∆ ∆0 ∆M ∆F ∆X % CS Males % CS Females Age + Education 16.32% 13.44% 0.00% 0.00% 2.88% 100.00% 100.00% 16.32% 25.17% 0.39% -0.01% -9.23% 99.46% 99.88% Age + Education 8.80% 9.73% 0.00% 0.00% -0.92% 100.00% 100.00% 8.80% 22.21% 0.03% 0.01% -13.44% 99.86% 99.97% Period 1 (CIRCA 1992) + Presence of + Presence of Other Wage Children in the Earner in the Household Household 16.32% 16.32% 25.42% 23.96% 0.50% 0.80% 0.05% -0.02% -9.65% -8.41% 98.20% 93.47% 99.52% 98.88% Period 2 (CIRCA 2007) + Presence of + Presence of Other Wage Children in the Earner in the Household Household 8.80% 8.80% 22.21% 21.88% 0.04% -0.25% 0.02% 0.07% -13.47% -12.90% 99.26% 97.42% 99.78% 99.41% + Urban + Type of Employment + Time Worked 16.32% 25.00% 0.02% 0.13% -8.83% 89.34% 97.40% 16.32% 23.99% 2.23% 0.26% -10.16% 79.62% 92.79% 16.32% 33.68% 1.29% -1.43% -17.22% 65.55% 80.66% + Urban + Type of Employment + Time Worked 8.80% 22.56% -0.89% 0.16% -13.03% 95.28% 98.74% 8.80% 20.75% -0.33% 0.37% -11.98% 89.61% 96.36% 8.80% 29.56% -2.07% 0.43% -19.12% 79.42% 89.04% Drop in Gender Earnings Gaps (15-year span) The Gap has Dropped in Most Countries And it Has Dropped Especially at Both Extremes of the Earnings Distribution It Has Dropped more among Low-Educated People It has dropped more in rural areas It Has Dropped more among the Self-Employed It Has Dropped more among those with Children at Home It Has Dropped more among Part-Time Workers It has dropped for older cohorts (matching after matching) Cohort Decomposition Counterfactual Jump if no Change in X's 15 to 29 in 1992 Part of the Jump due to changes in X's Total Jump Counterfactual Jump if no Change in X's 30 to 44 in 1992 Part of the Jump due to changes in X's Total Jump Counterfactual Jump if no Change in X's 45 to 59 in 1992 Part of the Jump due to changes in X's Total Jump Education Presence of Children in the Household 5.21 2.47 7.67 -4.82 1.04 -3.79 -16.55 4.13 -12.42 6.98 0.70 7.67 -5.13 1.35 -3.79 -12.54 0.12 -12.42 Presence of other wage earner in the Household 7.30 0.38 7.67 -3.91 0.13 -3.79 -12.40 -0.02 -12.42 Urban Type of Employment Time Worked Full Set 6.52 1.16 7.67 -5.67 1.89 -3.79 -14.29 1.86 -12.42 7.82 -0.14 7.67 -3.05 -0.72 -3.79 -11.56 -0.86 -12.42 6.54 1.14 7.67 -4.20 0.42 -3.79 -12.91 0.49 -12.42 6.98 0.70 7.67 -6.72 2.94 -3.79 -28.75 16.33 -12.42 3. Conclusions Methodological Advantages/Disadvantages Messages Advantages/Disadvantages It is not necessary to estimate earnings equations (no functional form assumption) Better assessment. The traditional approach seems to deliver biased results when the differences in supports are not taken into account Once the matching has been done, it is straightforward to: » Explore the distribution of the unexplained wage gap » Explore not only wage gaps but also gaps for other labor market outcomes (participation, unemployment, unemployment spells, segregation) Curse of Dimensionality. The method does not allow us to use too many explanatory variables. It does not take into account selection into the labor markets Summary Gender wage gaps » Between 8% and 28%. » older people, those with lower income, with secondary education, self-employed, informal and those in small firms » Some “CEO effects” (in some countries) » Somewhat surprising segregation effects » An overall reduction over the last 15 years, especially for those segments of the labor markets with the highest disparities. » And the reductions are not explained by changes in the labor markets’ composition