Noncognitive Skills and the Racial Wage Gap Disclaimer: The views expressed in this research, including those related to statistical, methodological, technical, or operational issues, are solely those of the author and do not necessarily reflect the official positions or policies of the Census Bureau, or the views of other staff members within these agencies. The author accepts responsibility for all errors. This paper is released to inform interested parties of ongoing research and to encourage discussion of work in progress. U S C E N S U S B U R E A U Helping You Make Informed Decisions Charles Hokayem, Housing and Household Economic Statistics Division, United States Census Bureau | 301-763-5330 | [email protected] Background Noncognitive skills, or “soft skills,” describe a person’s self-perception, work ethic, ethical orientation, and overall outlook on life. These skills have been linked to a variety of economic outcomes such as educational attainment, earnings, and work habits in the general population (Heckman et al 2006). They are important to the design of early childhood policies and adult work training programs. Less well understood is the impact of these skills on subgroups of the general population, specifically racial groups. This paper adds two measures of noncognitive skills, locus of control and self-esteem, to a simple wage specification to determine the effect of noncognitive skills by gender on the racial wage gap (white, black, and Hispanic) and the return to noncognitive skills across the wage distribution. Data and Skills Measures Data The analysis data come from the National Longitudinal Survey of Youth 1979 (NLSY79). Collected by the Bureau of Labor Statistics (BLS) the NLSY79 is a panel survey that contains 12,686 individuals between the ages of 14 and 21 at the time of first interview in 1979. The NLSY79 collects information on labor market outcomes, cognitive skills, and noncognitive skills. Additional wage data come from the Current Population Survey (CPS). This analysis uses NLSY79 and CPS observations for 1991-2006. Cognitive Skills Measure Armed Forces Qualifying Test (AFQT) measured in 1979 Noncognitive Skills Measures • Rotter Internal-External Locus of Control measures the degree to which a person has control over their life in 1979. • Rosenberg Self-Esteem measures an individual’s self-esteem in 1980. Empirical Methods Empirical Specification • Key Idea: AFQT i,1980 and Noncog i,1979/1980 are measured before labor market entry (Neal and Johnson 1996) i i i i i t i i i t i Noncog Noncog AFQT AFQT Age Hispanic Black wage ε β β β β β β β β + + + + + + + + = 2 1980 / 1979 , 7 1980 / 1979 , 6 2 1980 , 5 1980 , 4 , 3 2 1 0 , ln Results and Implications Quantile Regression Results After controlling for locus of control, the • wage gap for black and Hispanic men shrinks by 1-2 percentage points • wage gap for black and Hispanic women shrinks by 1-5 percentage points After controlling for self-esteem, the • wage gap for black men and women mostly widens by 1-5 percentage points • wage gap for Hispanic men and women mostly shrinks by 1-4 percentage points After controlling for cognitive and noncognitive skills, • the male black-white wage gap persists across the wage distribution • the female black-white wage gap exists at higher quantiles of wage distribution • Hispanic men earn less than white men at lower quantiles but earn more at higher quantiles • Hispanic women earn more than white women across the entire wage distribution • the return to cognitive skills still remains greater than the return to noncognitive skills across the wage distribution Implications Noncognitive skills have generally been found to determine wage levels in the general population and across both genders (Heckman et al 2006). In this context, the finding in this paper that noncognitive skills cannot affect or close some racial wage gaps presents a puzzle to the noncognitive literature. On one hand, these skills are important for wage levels; on the other hand, they do not seem to be important for wage gaps. This result has implications for education policy designed to close racial gaps. Bibliography Heckman, James, Jora Stixrud, and Sergio Urzua. 2006. The Effects of Cognitive and Noncognitive Abilities on Labor Market Outcomes and Social Behavior. Journal of Labor Economics 24, no. 3, 411-482. Neal, Derek, and William Johnson. 1996. The Role of Premarket Factors in Black-White Wage Differences. Journal of Political Economy 104, no. 5, 869-895. Cognitive and Noncognitive Measures 0 .2 .4 .6 Density -4 -2 0 2 4 Standardized AFQT Score Black Hispanic White 0 .2 .4 .6 Density -4 -2 0 2 4 Standardized Rotter Score Black Hispanic White 0 .2 .4 .6 Density -4 -2 0 2 4 Standardized Rosenberg Score Black Hispanic White Distribution of AFQT, Rotter, and Rosenberg Scores: Entire Sample Each graph displays the median hourly wage for each gender race combination for the NLSY and the CPS covering 1991-2006. Further infor- mation about the source and accuracy of the CPS and NLSY can be found at <http://www.bls.census.gov/cps/bsrcacc.htm> and <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>, respectively. Median Hourly Wages by Race, 1991-2006 Change in Racial Wage Gap Due to Noncognitive Skills Each graph shows the change in the quantile coefficient estimates for black and Hispanic by gender. Each bar represents the change in the wage gap after adding noncognitive measures to the empirical specification and not controlling for the cognitive measure. Quantile regressions are estimated on the pooled sample and include annual time dummy variables. Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>. Each graph shows the quantile coefficient estimates from the empirical specification controlling for cognitive and noncognitive skills. Quantile regressions are estimated on the pooled sample and include annual time dummy variables. Standard errors for confidence intervals are based on the nonparametric bootstrap with 100 replications. Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>. Quantile Regression Controlling for Cognitive and Noncognitive Skills Presented at the Annual Meeting of the Population Association of America Washington, DC March 30–April 2, 2011 Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>. -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Black -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Hispanic -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff AFQT -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Rotter -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff AFQT2 -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Rotter2 Men Quantile Coefficients For Rotter Locus of Control -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Black -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Hispanic -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff AFQT -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Rotter -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff AFQT2 -.4 -.2 0 .2 .4 Coefficient 0 .2 .4 .6 .8 1 Wage Quantile 95% Conf Band Quantile Coeff Rotter2 Women Quantile Coefficients For Rotter Locus of Control -10.0 -8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 10 20 30 40 50 60 70 80 90 Change in Wage Gap (percentage points) Wage Quanle Change in Wage Gap Due To Locus of Control Black Men Black Women Hispanic Men Hispanic Women -10.0 -8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 10 20 30 40 50 60 70 80 90 Change in Wage Gap (percentage points) Wage Quanle Change in Wage Gap Due To Self-Esteem Black Men Black Women Hispanic Men Hispanic Women 0 5 10 15 20 25 1991 1992 1993 1994 1996 1998 2000 2002 2004 2006 Median Wage ($/hour) Year Men (NLSY) Black White Hispanic 0 5 10 15 20 25 1991 1992 1993 1994 1996 1998 2000 2002 2004 2006 Median Wage ($/hour) Year Women (NLSY) Black White Hispanic 0 5 10 15 20 25 1991 1992 1993 1994 1996 1998 2000 2002 2004 2006 Median Wage ($/hour) Year Men (CPS) Black White Hispanic 0 5 10 15 20 25 1991 1992 1993 1994 1996 1998 2000 2002 2004 2006 Median Wage ($/hour) Year Women (CPS) Black White Hispanic
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Noncognitive Skills and the Racial Wage Gap
Disclaimer: The views expressed in this research, including those related to statistical, methodological, technical, or operational issues, are solely those of the author and do not necessarily reflect the official positions or policies of the Census Bureau, or the views of other staff members within these agencies. The author accepts responsibility for all errors. This paper is released to inform interested parties of ongoing research and to encourage discussion of work in progress.
U S C E N S U S B U R E A UHelping You Make Informed Decisions
Charles Hokayem, Housing and Household Economic Statistics Division, United States Census Bureau | 301-763-5330 | [email protected]
BackgroundNoncognitive skills, or “soft skills,” describe a person’s self-perception, work ethic, ethical orientation, and overall outlook on life. These skills have been linked to a variety of economic outcomes such as educational attainment, earnings, and work habits in the general population (Heckman et al 2006). They are important to the design of early childhood policies and adult work training programs. Less well understood is the impact of these skills on subgroups of the general population, specifically racial groups.
This paper adds two measures of noncognitive skills, locus of control and self-esteem, to a simple wage specification to determine the effect of noncognitive skills by gender on the racial wage gap (white, black, and Hispanic) and the return to noncognitive skills across the wage distribution.
Data and Skills MeasuresDataThe analysis data come from the National Longitudinal Survey of Youth 1979 (NLSY79). Collected by the Bureau of Labor Statistics (BLS) the NLSY79 is a panel survey that contains 12,686 individuals between the ages of 14 and 21 at the time of first interview in 1979. The NLSY79 collects information on labor market outcomes, cognitive skills, and noncognitive skills. Additional wage data come from the Current Population Survey (CPS). This analysis uses NLSY79 and CPS observations for 1991-2006.
Cognitive Skills MeasureArmed Forces Qualifying Test (AFQT) measured in 1979
Noncognitive Skills Measures• Rotter Internal-External Locus of Control measures the degree to which a person has control over their life in 1979.• Rosenberg Self-Esteem measures an individual’s self-esteem in 1980.
Empirical MethodsEmpirical Specification
• Key Idea: AFQTi,1980 and Noncogi,1979/1980 are measured before labor market entry (Neal and Johnson 1996)
iiii
itiiiti
NoncogNoncogAFQT
AFQTAgeHispanicBlackwage
εβββ
βββββ
+++
+++++=2
1980/1979,71980/1979,621980,5
1980,4,3210,ln
Results and ImplicationsQuantile Regression ResultsAfter controlling for locus of control, the
• wage gap for black and Hispanic men shrinks by 1-2 percentage points• wage gap for black and Hispanic women shrinks by 1-5 percentage points
After controlling for self-esteem, the• wage gap for black men and women mostly widens by 1-5 percentage points• wage gap for Hispanic men and women mostly shrinks by 1-4 percentage points
After controlling for cognitive and noncognitive skills,• the male black-white wage gap persists across the wage distribution• the female black-white wage gap exists at higher quantiles of wage distribution• Hispanic men earn less than white men at lower quantiles but earn more at higher quantiles• Hispanic women earn more than white women across the entire wage distribution• the return to cognitive skills still remains greater than the return to noncognitive skills across the wage distribution
ImplicationsNoncognitive skills have generally been found to determine wage levels in the general population and across both genders (Heckman et al 2006). In this context, the finding in this paper that noncognitive skills cannot affect or close some racial wage gaps presents a puzzle to the noncognitive literature. On one hand, these skills are important for wage levels; on the other hand, they do not seem to be important for wage gaps. This result has implications for education policy designed to close racial gaps.
BibliographyHeckman, James, Jora Stixrud, and Sergio Urzua. 2006. The Effects of
Cognitive and Noncognitive Abilities on Labor Market Outcomes and Social Behavior. Journal of Labor Economics 24, no. 3, 411-482.
Neal, Derek, and William Johnson. 1996. The Role of Premarket Factors in Black-White Wage Differences. Journal of Political Economy 104, no. 5, 869-895.
Cognitive and Noncognitive Measures
0.2
.4.6
Dens
ity
-4 -2 0 2 4Standardized AFQT Score
Black Hispanic White
0.2
.4.6
Dens
ity
-4 -2 0 2 4Standardized Rotter Score
Black Hispanic White
0.2
.4.6
Dens
ity
-4 -2 0 2 4Standardized Rosenberg Score
Black Hispanic White
Distribution of AFQT, Rotter, and Rosenberg Scores: Entire Sample
Each graph displays the median hourly wage for each gender race combination for the NLSY and the CPS covering 1991-2006. Further infor-mation about the source and accuracy of the CPS and NLSY can be found at <http://www.bls.census.gov/cps/bsrcacc.htm> and <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>, respectively.
Median Hourly Wages by Race, 1991-2006
Change in Racial Wage Gap Due to Noncognitive Skills
Each graph shows the change in the quantile coefficient estimates for black and Hispanic by gender. Each bar represents the change in the wage gap after adding noncognitive measures to the empirical specification and not controlling for the cognitive measure. Quantile regressions are estimated on the pooled sample and include annual time dummy variables. Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>.
Each graph shows the quantile coefficient estimates from the empirical specification controlling for cognitive and noncognitive skills. Quantile regressions are estimated on the pooled sample and include annual time dummy variables. Standard errors for confidence intervals are based on the nonparametric bootstrap with 100 replications. Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>.
Quantile Regression Controlling for Cognitive and Noncognitive Skills
Presented at the Annual Meeting of the Population Association of AmericaWashington, DCMarch 30–April 2, 2011
Source: NLSY. Further information about the source and accuracy of the NLSY can be found at <http://www.nlsinfo.org/nlsy79/docs/79html/NLSY79%20Tech%20Samp%20Rpt.pdf>.
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Black
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Hispanic
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
AFQT
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Rotter
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
AFQT2
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Rotter2
Men Quantile Coefficients For Rotter Locus of Control
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Black
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Hispanic
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
AFQT
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Rotter
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
AFQT2
−.4
−.2
0.2
.4C
oeffi
cien
t
0 .2 .4 .6 .8 1Wage Quantile
95% Conf Band Quantile Coeff
Rotter2
Women Quantile Coefficients For Rotter Locus of Control
-10.0
-8.0
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
10.0
10 20 30 40 50 60 70 80 90
Chan
ge in
Wag
e G
ap (
perc
enta
ge p
oint
s)
Wage Quantile
Change in Wage Gap Due To Locus of Control
Black Men
Black Women
Hispanic Men
Hispanic Women
-10.0
-8.0
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
10.0
10 20 30 40 50 60 70 80 90
Chan
ge in
Wag
e G
ap (
perc
enta
ge p
oint
s)
Wage Quantile
Change in Wage Gap Due To Self-Esteem
Black Men
Black Women
Hispanic Men
Hispanic Women
0
5
10
15
20
25
1991 1992 1993 1994 1996 1998 2000 2002 2004 2006
Med
ian
Wag
e ($
/hou
r)
Year
Men (NLSY)
Black
White
Hispanic
0
5
10
15
20
25
1991 1992 1993 1994 1996 1998 2000 2002 2004 2006
Med
ian
Wag
e ($
/hou
r)
Year
Women (NLSY)
Black
White
Hispanic
0
5
10
15
20
25
1991 1992 1993 1994 1996 1998 2000 2002 2004 2006
Med
ian
Wag
e ($
/hou
r)
Year
Men (CPS)
Black
White
Hispanic
0
5
10
15
20
25
1991 1992 1993 1994 1996 1998 2000 2002 2004 2006
Med
ian
Wag
e ($
/hou
r)
Year
Women (CPS)
Black
White
Hispanic
Related Documents
