AN ECONOMIC ANALYSIS OF ANTI-HISPANIC DISCRIMINATION IN THE AMERICAN LABOR MARKET: 1970s-1990s - Statistical Data Included
...a].sub.ji] the corresponding coefficients. The [[Beta].sub.ji] coefficients are measures of the change in the dependent variable for changes in each independent variable. For example if an individual gains a year of education, ceteris paribus, the human capital model says that individual's earnings will rise. The [Beta] coefficient on the education variable measures how much earnings will increase for a unit increase in education. The symbol e represents the stochastic error term for each individual i. In the empirical analysis below, a version of this equation is estimated for white males, white females, Hispanic males and Hispanic females in order to obtain the data required for measuring the total earnings difference between white males and the reference groups. This total earnings difference is then separated into the portion attributable to differences in market characteristics (the endowment effect) and a residual portion, which may be due to labor market discrimination. The following equation shows the earnings differential broken down into the endowment and residual proportions: Equation 2: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], where [y.sup.wm] is the estimated earnings level for white males and [y.sup.r] is the estimated earnings level for each reference group (white females, Hispanic males, and Hispanic females). The Use of B* incorporates the methodological extension developed separately by Cotton (1988)(9) and Neumark (1988).(10) The B* coefficient estimates are those which would occur in a discrimination-free world, while the B coefficient estimates are from the equations estimated for white males (wm) and reference groups (r). If there were no discrimination, then white male coefficients would fall while reference group coefficients would rise due to changes in labor market supply and demand. B* represents a weighted average of the coefficients for the reference group and white males. It is obtained from a regression that includes all groups of the population (white males, white females, Hispanic males, and Hispanic females). The equation thus compares the current estimated earnings of all groups with what they would be in a discrimination-free world. The various means ([x.sub.j]) are the means of different characteristics for white males (wm) and reference group males and females (r). Using such a model, Durden and Gaynor found that the 1990 discrimination-based earnings difference (in logarithmic values) between white and Hispanic males was about 1.1% and for Hispanic females about 22.2%.(11) Reimers found a 15% differential between white and Hispanic males,(12) and Pagan and Cardenas found a similar difference in logarithmic values of .1919 for 1992 Hispanic males and .2916 for 1992 Hispanic females.(13) In dollar values, Torres found a residual of $3,808 for 1980 Puerto Rican-born Hispanic males and -$310 for females; for 1980 U.S.-born Hispanics the discrimination residuals were $1,590 for males and -$641 for females.(14) Verdugo found a larger difference in 1987 for Mexican-American workers of $2,098.(15) In this paper, we will also be examining earnings in dollar, rather than logarithmic, values. The Variables and Equation Using 1970, 1976, and 1995 Census data from the Inter-University Consortium for Political and Social Research, the following human-capital based regression models (based on equation l) were estimated for each year: Equation 3: [Y.sub.i] = [b.sub.0] + [b.sub.1] ([X.sub.1]) + [b.sub.2]([X.sub.2]) + [b.sub.3]([X.sub.3]) + [b.sub.4]([X.sub.4]) + [b.sub.5]([X.sub.5]) + [e.sub.i], where, for white and Hispanic males and females, [Y.sub.i] is estimated earnings, [X.sub.1] - [X.sub.5] are vectors of independent variables proxies used to estimate the effects of differences in worker productivity and other respondent characteristics, and [b.sub.i] are the corresponding coefficient estimates for human-capital variables, spatial-related variables, industry and occupation variables, family-related variables, and other influences. The variables included in each of the five categories are summarized in Table 1. Table 1: Variables included in the regression equations HGHGRADE The highest grade of schooling completed EXPERIENCE Number of years of on-the-job experience SMSA Value = 1 if the individual resides in the SMSA, otherwise 0 EXPSQ Number of years of on-the-job experience squared WEST Value = 1 if the individual resides in the West, otherwise 0 SOUTH Value = 1 if the individual resides in the South, otherwise 0 MIDWEST Value = 1 if the individual resides in the Midwest, otherwise 0 GOVT Value = 1 if the individual is a government employee, otherwise 0 WHTRADE Value = 1 if the individual works in wholesale trade, otherwise 0 RETRADE Value = 1 if the individual works in retail trade, otherwise 0 CONST Value = 1 if the individual works in construction, otherwise 0 FIRE Value = 1 if the individual works in finance, insurance, and real estate, otherwise 0 DUMANUF Value = l if the individual works in durable goods manufacturing, otherwise 0 NDMANUF Value = 1 if the individual works in non-durable goods manufacturing, otherwise 0 TRANSCOM Value = 1 if the individual works in transportation and communications, otherwise 0 PROFSER Value = 1 if the individual works in the professional services, otherwise 0 PROF_TEC Value = 1 if the individual works in professional or technical fields, otherwise 0 MANAGER Value = 1 if the individual works as a manager, otherwise 0 CLERICAL Value = 1 if the individual works in the clerical fields, otherwise 0 OTH_SER Value = 1 if the individual works in other services, otherwise 0 LAB_OPER Value = 1 if the individual works as a laboratory operator, otherwise 0 SALES Value = 1 if the individual works in sales, otherwise 0 CRAFTS Value = 1 if the individual works in crafts, otherwise 0 MARRIED Value = 1 if the individual is married, otherwise 0 HHEAD Value = 1 if the individual is the head of the household, otherwise 0 FULLTIME Value = 1 if the individual is employed fulltime, otherwise 0 PURICAN Value = 1 if the individual is an Hispanic of Puerto Rican descent, otherwise 0 CUBAN Value = 1 if the individual is an Hispanic Of Cuban descent, otherwise 0 BLACK Value = 1 if the individual is an Hispanic of Black descent, otherwise 0 Empirical Evidence Using 1970, 1976, and 1995 U.S. Census data, we regressed the previously discussed variables on the earnings of individual Hispanic males (1976 and 1995), Hispanic females (1976 and 1995), white males (1970 and 1995), and white females (1970 and 1995). All earnings values are in 1982 dollars so that use of the specially collected 1976 Hispanic data should provide estimates, which can reasonably be compared with estimations for white males and females using 1970-CPS data. Sufficient data on Hispanics are not available in the 1970-CPS files. The results of the regressions are shown in Tables 2 and 3. All means and significant coefficients to the 0.01 level together with corresponding [R.sup.2] values are presented. The equations are robust with respect to explanatory power, and most of the important coefficients are significant and signed as expected. [TABULAR DATA 2-3 NOT REPRODUCIBLE IN ASCII] In following with our model, we then calculated the B* "discrimination-free" coefficient estimates using a weighted average of the corresponding estimates of each subgroup of the population. Equation 2 (above) was then used to compute the components shown in Table 4, which shows earnings comparisons. Column 1 shows total estimated earnings, and column 2 displays the total earnings differential between white males and each reference group. In column 3, we look more specifically at the difference attributable to skill represented in equation 2 by the expression [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. Recall that this is a measure of the earnings effects of differences in such characteristics as education, experience, industry and job category, etc. Computed this way, the skill differential is that which exists in a discrimination-free world. Column 4 shows the dollar value of the advantage associated solely with being a white male. In our equation, this factor is in equation 2, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] which shows the difference, for example, between what white males earn for an additional year of education as compared to what they would earn in a discrimination free world. If [B.sup.wm] exceeds B*, then white males enjoy an earnings premium, given current labor market conditions. Column 5 represents a similar variable for the disadvantage associated solely with being a member of the target reference group. This factor is shown in our equation by the sum of two components, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. Column 6 simply subtracts the differential attributable to skill (column 3) from the total difference between estimated earnings for white males and the reference group (column 2). This residual difference (it is also the sum of white male advantage and reference group disadvantage) is the estimated earnings difference due to discrimination. As discussed earlier, this estimate is almost surely an overstatement due to the certain omission of some quantified variables and the use of proxies for non-quantified ones. Table 4: (1) (2) (3) (4) (5) (6) Estimated Total Skill White Reference Residual Earnings Diffe- Diffe- Male Group Diffe- rence rential Advan- Disad- rence tage vantage White Males 1970: $18962 1995: $20597 White Females 1970: $8125 $10837 $4911 $3901 $2026 $5927 1995: $12823 $7775 $2701 $2851 $2222 $5073 $ Change: $4698 $-3062 $-2210 $-1050 $196 $-854 % Change: 57.8 -28.3 -45.0 -26.9 9.7 -14.4 Hispanic Males 1976: $8782 $10180 $528 $3901 $5751 $9652 1995: $13946 $6651 $4953 $2851 $-1153 $1698 $ Change: $5164 $-3529 $4425 $-1050 $-6904 $-7954 % Change: 58.8 -34.7 838.1 -26.9 -120 -82.4 Hispanic Females 1976: $4510 $14452 $2791 $3901 $7760 $11661 1995: $10416 $10181 $5589 $2851 $1740 $4591 $ Change: $5906 $-4271 $2798 $-1050 $-6020 $-7070 % Change: 131.0 -29.6 100.3 -26.9 -77.6 -60.6 Over the time period of our analysis, we find a general decrease in the total differential between white males and all reference groups of around 30% (column 2 of Table 4). The change in skill differential shown in column 3 of the table is considerably less consistent. The dollar denominated skill differential decreased from $4911 to $2701 (-45%) for white females, but actually increased for Hispanic males ($528 to $4953, +838%) and females ($2791 to $5589, 100%). We would have expected the skill differential to narrow for Hispanic males and females, as compared to white males, but this is not the case. The primary reason is that white females and males have significantly increased their education levels between 1970 and 1995, but for Hispanics, education level has been relatively constant (see Table 2). This probably means that more recent Hispanic immigrants are less well-educated than the those already in the United States, a condition which should change as newcomers become assimilated and gain more education and training. The white male relative advantage has fallen from $3901 to $2851, -27%, for the time period (column 4 of Table 4). The reference group disadvantage effects shown in column 5 show an interesting result. White females are at about the same dollar disadvantage now ($2222) as they were...