Abstract
Despite concern regarding labor market discrimination against Hispanics, previously published estimates show that Hispanic women earn higher hourly wages than white women with similar observable characteristics. This estimated wage premium is likely biased upwards because of the omission of an important control variable: cost of living. We show that Hispanic women live in locations (e.g., cities) with higher costs of living than whites. After we account for cost of living, the estimated Hispanic-white wage differential for non-immigrant women falls by approximately two-thirds. As a result, we find no statistically significant difference in wages between Hispanic and white women in the NLSY97.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Other papers find no difference in wages between white and Hispanic women, conditional on observable worker characteristics (e.g., Fryer 2011—NLSY97 estimates). Recent estimates of the wage gap between Hispanic and white men range from no difference (Fryer 2011) to a penalty of approximately 0.10 log points (Hirsch and Winters 2014; Black et al. 2012).
A group of workers with higher wages due to their tendency to locate in places with high firm productivity and high land prices are not better off as a consequence of their high wages than they would be with lower wages and lower costs of living.
On average, Hispanics have lower AFQT scores than whites, and since there is a wage return to a higher score on the AFQT, omitting the test score results in Hispanics appearing to perform more poorly, relative to whites. Neal and Johnson (1996) find that including a respondent’s AFQT score can boost relative wages for young Hispanic workers, relative to whites, by between 0.11 and 0.14 log points.
The NLSY97 survey is fielded annually, and the NLSY79 is fielded every other year. Therefore, these are the most recent years of data available for each survey.
Recent research notes that self-reports of race and ethnicity may be inaccurate due to selective group identification and ethnic attrition (Antman and Duncan 2015; Duncan and Trejo 2011). We identify respondent race and ethnicity using as much information from our data sets as we can, including self-reports from multiple surveys (fielded years apart) and assessments made by screeners.
To be in the OLS analysis, women must have worked and have valid wage information in 2011, 2010, or 2009. The wage measure we use is the hourly wage at the current or most recent job. If the respondent does not report current wages, we impute the most recent wage from the prior two years (adjusted for inflation by the Consumer Price Index).
Our education variable (EDUC i ) is the respondent’s years of completed schooling. We start with the highest grade completed that was reported in the most recent survey. We impute a value of 12 for respondents with 11 years of school who ever received a high school diploma or equivalent. We impute a value of 16 for respondents with 15 years of school who ever received a bachelor’s degree. We impute a value of 15 for respondents with more than 15 years of school but no bachelor’s or higher degree.
Since schooling and experience influence AFQT scores, our AFQT score variable is standardized by birth year (or equivalently, age when taking the test). We calculate the mean and standard deviation of raw AFQT scores within each birth year cohort. Our AFQT variable is the difference between a respondent’s raw score and the cohort mean, divided by the cohort’s standard deviation. The method follows Neal and Johnson (1996).
We identify spouse income for the past 5 years in the NLSY79 (from the prior three biannual surveys) and the past 3 years in the NLSY97 (from three annual surveys).
We only use the years that a respondent reports nonzero spousal earnings between t-5 and t-1 to identify high earning spouses. For example, a non-working college graduate woman who was single from years t-5 through t-2 would be assigned zero spousal earnings for those years. But if she married a high-earning spouse one year before the survey, then she gets a high imputed wage.
We do not impute wages for 160 (204) non-working women in the NLSY97 (NLSY79). 79 (83) of these women had at least some college education. Our main results are robust to changes in imputation. If we impute all missing values as either very high or very low, we obtain similar results to those shown below.
The smallest identifiable area in the ACS is the public use microdata area (PUMA), a Census-defined place with population over 100,000. Some PUMA boundaries do not perfectly align with counties. When this is the case, we assign PUMA characteristics to a CZ based on the PUMA’s population share in the CZ (see McHenry 2014). The housing cost variable is similar to the one in Moretti (2013).
Banzhaf and Farooque (2012) compare alternative methods for measuring local housing costs and find that average rental prices perform well: they are closely associated with housing transaction price data (which are more costly to collect), and rental prices are closely associated with measured local amenities and average incomes.
That is, the CZ housing cost measure is computed as follows:
\( {\mathrm{HousingCost}}_{CZ}=\frac{{\mathrm{MeanRent}}_{CZ}}{N^{-1}{\displaystyle {\sum}_{CZ=1}^N{\mathrm{MeanRent}}_{C{Z}^{\cdot }}}} \) and the cost of living is computed as CostofLiving CZ =.4146 * HousingCost CZ +.5854 * 1. The 41.46 % housing expenditure share in 2011 is from the Bureau of Labor Statistics web page (http://www.bls.gov/cpi/cpiri2011.pdf; Bureau of Labor Statistics 2007).
The one exception is that the OLS coefficient estimate for the NLSY97 now achieves statistical significance.
The corresponding estimates are somewhat larger in magnitude and achieve statistical significance in the older cohort in the NLSY79: 0.45 with a standard error of 0.12. This suggests that differential patterns in educational attainment by ethnicity have gotten smaller in younger cohorts.
In results not shown, we confirm that the addition of cost of living has a similar effect when we instead add a control for years of education to equation (1) first, and a control for cost of living second. That is, adding a control for years of education only reduces the estimated wage premiums by between 0.004 and 0.016 log points, but adding a control for cost of living reduces the estimated wage premium by between 0.074 and 0.125 log points, and the estimate loses statistical significance.
In results not shown, we find that the same qualitative conclusions are upheld with a less conservative treatment of selection out of the labor force. For those results, we alternatively impute a low (high) wage of $1 ($45) for all women with missing wages who did not meet our imputation criteria. The only difference is that we now have a statistically significant coefficient estimate (at the 10 % level) for the NLSY79 median regression results when we impute a low wage of $1 for all women with missing wages (analogous to the column (6) result in Panel III of Table 4).
Trejo (1997) examines wage differentials among Mexican and white men and shows labor market returns to education and work experience differ for the newly arrived (i.e., first or second generation) vs. those in the third generation or higher. Further, wage gaps for Mexican American men are over three times larger for first generation Mexican men (relative to first generation white men) than they are for men of the third generation or higher.
In the NLSY97, we exclude 85 immigrants (70 Hispanic, 15 white), and in the NLSY79 we exclude 153 immigrants (122 Hispanic, 31 white).
Not only is there a smaller share of Hispanic immigrants in the NLSY97 with lower levels of education, but the less-educated Hispanic immigrants in the NLSY97 have higher mean hourly wages and more years of education. Mean hourly wages in this group are $15.17, versus $11.28 for their counterparts in the ACS. Less-educated Hispanic women in the NLSY97 acquired 11.35 years of education, on average, versus 10.28 among their counterparts in the ACS.
Respondents of Mexican descent in the NLSY97 are the subset of respondents we previously identified as Hispanic who also selected “Mexican” as their primary ethnicity in the 1999 survey.
References
Albouy D (2009) The unequal geographic burden of federal taxation. J Polit Econ 117(4):635–667
Antecol H, Bedard K (2002) The relative earnings of young Mexican, black, and white women. Ind Labor Relat Rev 56(1):122–135
Antecol H, Bedard K (2004) The racial wage gap: the importance of labor force attachment differences across black, Mexican, and white men. J Hum Resour 39(2):564–583
Antman F, Duncan B (2015) Incentives to identify: racial identity in the age of affirmative action. Rev Econ Stat
Banzhaf HS, Farooque O (2012) Interjurisdictional housing prices and spatial amenities: which measures of housing prices reflect local public goods?” NBER working paper #17809
Black DA, Haviland AM, Sanders SG, Taylor LJ (2008) Gender wage disparities among the highly educated. J Hum Resour 43(3):630–659
Black DA, Kolesnikova N, Sanders SG, Taylor LJ (2012) The role of location in evaluating racial wage disparity. Federal Reserve Bank of St. Louis Working Paper #2009-043C
Bureau of Labor Statistics (2007) The Consumer price index. In Handbook of Methods. Washington, D.C.: U.S. Department of Labor. URL: http://www.bls.gov/opub/hom/pdf/homch17.pdf.
Chandra A (2003) Is the convergence of the racial wage gap illusory?. The National Bureau of Economic Research, Cambridge
Duncan B, Trejo SJ (2011) Tracking intergenerational progress for immigrant groups: the problem of ethnic attrition. Am Econ Rev 101(3):603–608
Duncan B, Hotz VJ, Trejo SJ (2006) Hispanics in the U.S. Labor Market. In: Marta T, Faith M (eds) Hispanics and the future of America. National Academies Press, Washington, D.C, pp 228–290
Fisher JD, Houseworth CA (2012) The reverse wage gap among educated white and black women. J Econ Inequal 10(4):449–470
Fitzpatrick K, Thompson JP (2010) The interaction of metropolitan cost-of-living and the federal earned income tax credit: one size fits all? Nat Tax J 63(3):419–446
Fryer R (2011) Racial inequality in the 21st century: the declining significance of discrimination. In: Card D, Ashenfelter O (eds) Handbook of labor economics, vol 4b. Elsevier, Amsterdam, pp 855–971
Hirsch BT, Winters JV (2014) An anatomy of racial and ethnic trends in male earnings in the U.S. Review of Income and Wealth 60(4):930-947
Hoynes H, Miller DL, Schaller J (2012) Who suffers during recessions? J Econ Perspect (Summer) 26(3):27–47
Johnson W, Kitamura Y, Neal D (2000) Evaluating a simple method for estimating black-white gaps in median wages. Am Econ Rev 90(2):339–343
Lang K, Manove M (2011) Education and labor market discrimination. Am Econ Rev 101(4):1467–96
McHenry P (2014) The geographic distribution of human capital: measurement of contributing mechanisms. J Reg Sci 54(2):215–248
McHenry P, McInerney M (2014) The importance of cost of living and education in estimates of the conditional wage gap between black and white women. J Hum Resour 49(3)
Moretti E (2011) Local labor markets. In: Card D, Ashenfelter O (eds) Handbook of labor economics, vol 4b. Elsevier, Amsterdam, pp 1237–1313
Moretti E (2013) Real wage inequality. Am Econ J Appl Econ 5(1):65–103
Neal D (2004) The measured black-white gap among women is too small. J Polit Econ 112(1):S1–S28
Neal DA, Johnson WR (1996) The role of premarket factors in black-white wage differences. J Polit Econ 104(5):869–895
Roback J (1982) Wages, rents, and the quality of life. J Polit Econ 90(6):1257–1278
Ruggles S, Trent Alexander J, Genadek K, Goeken R, Schroeder MB, Sobek M (2010) Integrated public use microdata series: version 5.0 [Machine-readable database]. University of Minnesota, Minneapolis
Tolbert CM, Sizer M (1996) U.S. commuting zones and labor market areas: a 1990 update. Rural Economy Division, Economic Research Service, U.S. Department of Agriculture. Staff Paper AGES-9614
Trejo S (1997) Why do Mexican americans earn low wages? J Polit Econ 105(6):1235–1268
United States Department of Labor (2012) The Latino labor force at a glance
Acknowledgments
The authors gratefully acknowledge support from a W.E. Upjohn Institute Early Career Research Grant. This research was conducted with restricted access to Bureau of Labor Statistics (BLS) data. The views expressed here do not necessarily reflect the views of the BLS. We thank Alison Courtney and Sarah Gault for excellent research assistance.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Rights and permissions
About this article
Cite this article
McHenry, P., McInerney, M. Estimating Hispanic-White Wage Gaps Among Women: The Importance of Controlling for Cost of Living. J Labor Res 36, 249–273 (2015). https://doi.org/10.1007/s12122-015-9210-3
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12122-015-9210-3