Monthly Archives: May 2014


Before proceeding to the IV estimation, a decision had to be made regarding which instrumental variables to use, choosing among the age, education, and job requirements variables. The maintained assumption is that at least one set of these variables can be excluded from the starting wage equation. The question that can be addressed empirically, however, is which set of instruments provides the most predictive power for the performance rating in the first-stage regression. To assess this, the first-stage regression was estimated using each set of instruments separately. For men, only the age variables were jointly significant in the first-stage regression; as reported in column (3), for the levels specification the p-value for the test of joint significance was .03 for the age variables, .58 for the education variables, and .47 for the job requirement variables, with qualitatively similar results for the log specification. Thus, for men the first set of instruments considered is the age variables. For women, only the education coefficients were jointly significant, with p-values of .00 in both the levels and log specifications. However, the p-values for the age variables are also relatively low (.17 and .21 in the two specifications); consequently, specifications using education and age as instruments are also reported for women.


Tables 2 and 3 report the results for the test of statistical versus taste discrimination, for men and women respectively. The first column in Panel A of each table reports OLS estimates of a standard log wage regression (for starting wages) without any information on performance ratings, with controls for education, age, job requirements, and race/ethnicity. In both the male and female samples, wages of blacks are significantly lower by about 14 percent, while wages of Hispanics are not significantly lower (with the point estimates indicating wage gaps of one to four percent). These results are not fully consistent with other estimates of race and ethnic wage differentials, where it is more common to find a smaller race difference among women than among men (Blau and Beller, 1992), and Hispanic-white differences are often larger than black-white differences (Reimers, 1983). However, this sample is somewhat unique in covering four specific metropolitan areas, and the wage measure studied here is the starting wage. The starting wage differentials associated with schooling appear relatively similar to those observed in other data sets for contemporaneous wages, although the considerably higher wage premium for male college graduates compared with female college graduates is unusual. The relationship between age and the starting wages also parallels the usual relationship. Among the job requirements, both specific experience and training are associated with significantly higher wages, while general experience is not.


Statistical Versus Taste Discrimination

Table 1 reports descriptive statistics on log starting wages, performance ratings, and log performance ratings. The wage differences between race and ethnic groups are similar for men and women, with whites earning about 19 percent more than blacks, and four to eight percent more than Hispanics. The difference in starting wages between men and women is about 10 to 14 percent, toward the lower figure for Hispanics. These sex-related differentials are somewhat small compared with representative samples of the U.S. work force, but the data here refer to starting wages of relatively young workers (29.5 years old, on average); existing work with other data sets documents the lower sex differences in wages for workers early in their careers (e.g., Light and Ureta, 1995).


The data set includes three types of variables that can be potentially be used as predictors of productivity that do not themselves (i.e., independently of productivity) affect wages: age, education, and job requirements. The information on job requirements comes from survey questions asking the employer whether specific experience, general experience, and vocational education or formal training are ’’absolutely required, strongly preferred, mildly preferred, or does not matter.” It seems reasonable to suppose that if these are absolutely required of a hire, then that hire must possess these qualifications, and the data are used in this manner. In fact, this supposition could be checked using another question on whether a high school diploma was required and corresponding information on the reported actual education of the worker hired; only 1.4 percent of those hires for which a high school diploma was absolutely required (27 percent of the hires in the sample) did not actually have a high school diploma. so


i (1)
Finally, the performance ratings pose some pure measurement problems, because they may vary for reasons other than the worker’s actual performance. In particular, the ratings that particular respondents provide may vary for random reasons, with some tending to give higher and some lower ratings for equally-productive workers. This case may be interpreted as one of pure measurement error in the performance rating. Unfortunately, in this case the instrumental variables procedure may be correcting for this pure measurement error, rather than that which arises in the imperfect information story because of discrepancies between current and expected productivity. Because the IV procedure may simply correct for standard measurement error bias, it could lead to spurious evidence in favor of the statistical discrimination model, with a,v larger in absolute value than aoLS, and bIV falling in absolute value relative to b0LS.


Respondents are asked about the last worker hired, whether or not that worker is still with the employer. The recorded characteristics of the last worker hired include race/ethnicity, sex, age, educational attainment, starting and current wages, and job requirements. In addition, a supervisor’s performance rating of the worker is also provided, measured on a scale of one to 100. These ratings are used to measure productivity (P).


These questions are studied using an employer data set stemming from the Multi-City Study of Urban Inequality (MCSUI). This data set contains information on starting and current wages, worker characteristics, and employers’ ratings of employees. The information available in the MCSUI conforms quite closely to the data required to implement the tests described in the previous section, with some exceptions discussed below. one hour payday loans


Thus, this model explains why groups of workers with identical expected productivity could earn different wages, and therefore provides yet another reason why labor market information may generate wage differences between similar workers. This is perhaps most pertinent for sex differences in wages, because women do not, on average, receive lower performance ratings than men. While simple statistical discrimination therefore cannot explain women’s lower wages, such extensions of the statistical discrimination model, coupled with worse information about women, can generate group discrimination against women.


An additional important component of the empirical analysis, which is described following the discussion of the results testing for taste versus statistical discrimination, concerns whether the differences between aoLS and aIV indicated by the data stem from imperfect information, or simply from measurement error in the performance rating available in the data set as a measure of true, known productivity. This is potentially important because the latter type of measurement error could generate evidence of statistical discrimination, despite employers (but not econometricians) having perfect information about workers. Electronic Payday Loans Online


What is required to correct the estimates of equation (4) for the bias from using P instead of Ps* is a variable that is correlated with productivity but uncorrelated with r|s, and that does not appear in equation (2). Because r|s is orthogonal to the information set Is, any variable that is in Is and correlated with productivity satisfies the first two criteria. However, to satisfy the third criteria, this variable has to be unrelated to starting wages conditional on expected productivity. Given that the null hypothesis is that there is taste discrimination, only variables that measure characteristics not subject to taste discrimination are valid instruments. The instruments considered include education, age, and training or experience. Age is potentially objectionable, given that there may well be age discrimination in the workplace (e.g., Johnson and Neumark, forthcoming). However, this is unlikely to be an issue in the present context, both because the sample consists of relatively young workers, and because most age discrimination claims concern discharges, layoffs, and hiring, rather than wage discrimination. website

Pages: 1 2 Next