|
| Stasz, C., & Brewer, D. J. (1999). Academic Skills at Work: Two Perspectives (MDS-1193). Berkeley: National Center for Research in Vocational Education, University of California. |
In this section, we offer some preliminary evidence on whether labor market outcomes are related not just to academic skills, but also to non-academic skills. We look only at the group of students who graduate high school with their class and go directly into the sub-baccalaureate labor market directly after high school. This is done for several reasons. First, the NELS:88 data only contain information on students two years after high school, so that labor market information is not yet available for students who go on to college. (On the other hand, the HSB data only contain earnings--rather than wage--data from 1992, ten years after high school graduation, and no hours worked information, so it is less than ideal; see below.)[30] Second, this sample restriction has the advantage of reducing heterogeneity in the sample--we are comparing a group with identical education and almost identical labor market experience levels. This reduces the likelihood that omitted factors bias our estimated results. It also means, of course, that one should not generalize from the results presented here to other populations such as high school dropouts or college-bound youth.
We confine our attention to the wage rates earned by individuals two years after high school. Examining the wages of students just two years after high school is problematic in part because previous research has suggested that this may not necessarily be an indicator of long-term success (Murnane et al., 1995). Some students in our sample will no doubt ultimately return to college and achieve greater earning potential than is indicated here, and the variation among the sample used in terms of wages will grow as their careers develop. In addition, of course, wages are only one indicator of labor market performance, and subsequent studies might look at wage growth, job status, job satisfaction, or other indicators.
Tables 4.5 and 4.6 show selected estimated regression coefficients from wage regressions for the variables of interest. As is standard in this type of study, we report the results separately for men and women. Major differences in the labor force behavior of men and women mean that pooling the two groups can yield misleading results. Table 4.5A shows the results for HSB men, Table 4.5B for NELS:88 men, Table 4.6A for HSB women, and Table 4.6B for NELS:88 women. All the HSB and NELS:88 models include a standard set of background control variables as noted in the notes to the tables.[31] The tables report a selection of all the models estimated, limiting attention to a manageable number of measures--mathematics test score in the 12th grade, mathematics units, 12th-grade total number of extracurricular activities, 12th-grade hours spent on extracurricular activities, and 12th-grade hours of work. (Other results not shown that are of interest--for example, using 10th-grade extracurricular activities and work, or the reading test score--will be discussed only to the extent that they yield contrasting findings to those reported in the tables.)
| (1) | (2) | (3) | (4) | (5) | |
| 12th grade math test score | -.001 (.182) | -.000 (.109) | -.001 (.190) | .001 (.183) | .000 (.158) |
| Number of mathematics course credits | -- | -.003 (.132) | -- | -- | -- |
| 12th grade total extracurricular activities | -- | -- | .001 (.163) | -- | .003 (.285) |
| 12th grade hours of work | -- | -- | -- | .007 (4.158) | .007 (4.160) |
| Sample size 1,269, male high school graduates who entered the labor market directly after high school only. |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | |
| 12th grade mathematics test score | -0.002 (1.532) | -0.002 (1.364) | -0.002 (1.533) | -0.002 (1.565) | -0.002 (1.646) | -0.002 (1.647) | -0.002 (1.700) |
| Number of mathematics course credits | -- | 0.001 (0.114) | -- | -- | -- | -- | -- |
| 12th grade total extracurricular activities | -- | -- | -0.0004 (0.056) | -- | -- | -0.001 (0.121) | -- |
| 12th grade hours extracurricular activities | -- | -- | -- | 0.003 (1.360) | -- | -- | 0.003 (0.580) |
| 12th grade hours of work | -- | -- | -- | 0.002 (1.692) | 0.002 (1.695) | 0.002 (1.900) |
| Sample size 1,831, male high school graduates who entered the labor market directly after high school only. |
Table 4.5A shows the estimated coefficients from OLS regression models with the natural logarithm of the hourly wage rate in 1984 (two years after high school) as the dependent variable. All models also include race/ethnicity dummy variables, family income, number of siblings, a single parent household dummy variable, and region dummies. For more details, see Appendix II.
Absolute value t statistics are shown in parentheses. For these sample sizes, a t statistic greater than 1.645 indicates that the coefficient estimate is statistically significant at the 10% level, and a t statistic of 1.960 indicates that the coefficient estimate is statistically significant at the 5% level.
Table 4.5B shows the estimated coefficients from OLS regression models with the natural logarithm of the hourly wage rate in 1994 (two years after high school) as the dependent variable. All models also include race/ethnicity dummy variables, family income, number of siblings, a single parent household dummy variable, and region dummies. For more details, see Appendix II.
Absolute value t statistics are shown in parentheses. For these sample sizes, a t statistic greater than 1.645 indicates that the coefficient estimate is statistically significant at the 10% level, and a t statistic of 1.960 indicates that the coefficient estimate is statistically significant at the 5% level.
| (1) | (2) | (3) | (4) | (5) | |
| 12th grade composite test score | .009 (2.305) | .009 (1.993) | .010 (2.349) | .009 (2.334) | .010 (2.358) |
| Number of mathematics course credits | -- | .010 (0.429) | -- | -- | -- |
| 12th grade total extracurricular activities | -- | -- | -.003 (0.295) | -- | -.004 (0.385) |
| 12th grade hours of work | -- | -- | -- | .009 (4.149) | .009 (4.147) |
| Sample size 1,298, female high school graduates who entered the labor market directly after high school only. |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | |
| 12th grade mathematics test score | .001 (1.0154) | .002 (1.402) | .001 (1.064) | .001 (1.1057) | .001 (1.075) | .001 (1.076) | .001 (0.999) |
| Number of mathematics course credits | -- | .009 (1.127) | -- | -- | -- | -- | -- |
| 12th grade total extracurricular activities | -- | -- | -.002 (0.183) | -- | -- | -.0004 (0.050) | -- |
| 12th grade hours extracurricular activities | -- | -- | -- | -.001 (0.281) | -- | -- | .0001 (0.033) |
| 12th grade hours of work | -- | -- | -- | .004 (2.438) | .004 (2.433) | .004 (2.443) |
| Sample size 1,745, female high school graduates who entered the labor market directly after high school only. |
Table 4.6A shows the estimated coefficients from OLS regression models with the natural logarithm of the hourly wage rate in 1984 (two years after high school) as the dependent variable. All models also include race/ethnicity dummy variables, family income, number of siblings, a single parent household dummy variable, and region dummies. For more details, see Appendix II.
Absolute value t statistics are shown in parentheses. For these sample sizes, a t statistic greater than 1.645 indicates that the coefficient estimate is statistically significant at the 10% level, and a t statistic of 1.960 indicates that the coefficient estimate is statistically significant at the 5% level.
Table 4.6B shows the estimated coefficients from OLS regression models with the natural logarithm of the hourly wage rate in 1994 (two years after high school) as the dependent variable. All models also include race/ethnicity dummy variables, family income, number of siblings, a single parent household dummy variable, and region dummies. For more details, see Appendix II.
Absolute value t statistics are shown in parentheses. For these sample sizes, a t statistic greater than 1.645 indicates that the coefficient estimate is statistically significant at the 10% level, and a t statistic of 1.960 indicates that the coefficient estimate is statistically significant at the 5% level.
Turning first to the results for the 1982 cohort, Table 4.5A suggests that academic ability as proxied by a mathematics test score has no impact on early career wages for men. For the 1982 women, Table 4.6A suggests that higher 12th-grade mathematics test scores are associated with higher wages (although it should be noted that this is not the case if the composite HSB test score is used instead of the math score). These results should not be interpreted as meaning that academics do not play a role, rather that they are not critical in early career labor market performance. The results are very similar to those of Murnane et al. (1995) who, using the 1980 HSB senior cohort, found math scores have a statistically significant effect on wages two years after high school for women but not for men.
The effect of adding extracurricular and part-time work activity to this model can be seen in columns (3)-(5). Three things are striking: (1) as these indicators are added to the model, the estimated effect of test score on wages barely changes. This provides some limited evidence that previous results may not have been biased in their restricted focus on academic skills; (2) for this group, participation in extracurricular activities does not yield any positive return in the labor market; and (3) the most important factor among the skill proxies shown in the tables is the extent to which the individual worked prior to entering the labor market. Every additional hour of part-time work in the 12th grade of high school generates a small (.7% in the case of men and .9% in the case of women) but statistically significant wage premium. This result holds for models including the test score only, extracurricular measures only, or measures of both. It is also insensitive to replacing 12th-grade work with 10th-grade work.
A similar pattern of results is suggested by the 1992 cohort results shown in Tables 4.5B and 4.6B. Table 4.5B again shows no evidence that academic skills--as proxied by mathematics test score or curriculum units--have any influence on wages; if anything, there is a negative relationship for men in the most complete model (column [7]). (This result is repeated for reading test scores. It also appears to be relatively insensitive to the inclusion of earlier grade level test scores in place of the 12th-grade test score.) Similarly, the effect of adding extracurricular and part-time work activity to this model (columns [3]-[7]) suggests that an individual's work in high school pays off subsequently, although the estimated premia are even smaller than in the 1982 case. Extracurricular activities do not pay off--at least in this sample--even when a more refined measure indicating hours spent on extracurricular activities (rather than simply the number of activities) is used; using 10th-grade extracurricular activities (not shown) also fails to reveal any positive statistically significant effect. Adding extracurricular activity and part-time work measures to the models does not alter the estimated effects of test scores on wages.
There are several reasons to be cautious of these results. First, and most importantly, the measures we have included in the models which we suggest might reasonably be interpreted as proxies for other kinds of skill, and, thus, could reflect many other factors; without more direct measures of generic skills, there is no way to know whether the suggested interpretation is accurate. In particular, as noted earlier, since students choose their skill bundles to some extent, it may be that unobservable factors correlated with extracurricular activities or part-time work bias the estimated coefficients in these statistical models.
Second, it is also important to bear in mind that these results are limited to high school graduates directly entering the labor market, and inferences cannot be made about the payoff to academics, extracurricular activities, and part-time work for other groups. For example, it may be that a high school dropout with a high level of extracurricular participation would be more attractive on the margin to an employer than a dropout with no extracurricular activities. Likewise, there may be a payoff to extracurricular activities for those going on to college given that some institutions may consider these as part of the student's application. As has been suggested elsewhere (e.g., Stern et al., 1997), the short-term payoff in the sub-baccalaureate labor market for this group of high school graduates must be set against the far greater returns they may have achieved had they gone on to postsecondary schooling.
Third, we have only examined wages as a labor market outcome. Although wages are a valid indicator of performance in the market, they are only one. Future studies might explore alternatives in more depth. As noted earlier, we have confined our attention to wages immediately after entering the labor market, which may not be a reliable indication of long-term labor market prospects. We, therefore, re-estimated our statistical models for the HSB sample for which we have earnings information ten years later in 1992 (information is not yet available for the 1992 cohort beyond 1994). Unfortunately, this earnings information is far from ideal since individuals were not asked about hours of work.[32] Log earnings models estimated using the same specifications as those shown for 1984 log wages and reveal the same pattern as those reported. In other words, hours of work in high school are positively associated with wages for both men and women, while extracurricular activities have no payoff; academics (as measured by the mathematics test score), however, are now statistically significant for men but not for women. (Murnane et al. [1995], using data six years after high school graduation, found statistically significant positive effects of the mathematics test score for both men and women.) However, when additional controls are included for total months employed and highest grade of schooling completed by 1992, this latter result on academics disappears. Thus, for this group of high school graduates entering the labor market, we do not find that academics pay off, even if later career earnings are considered. In contrast, hours worked in high school continue to earn a premium even when a person is ten years into the labor market.
[30]As an alternative, one could examine the senior cohort of HSB--that is, students in the high school class of 1980. For this cohort, labor market outcome data, including hourly wage rates, exist for the sample, six years after high school graduation. This is the sample used by Murnane et al. (1995). We have not analyzed this sample here because we wanted to examine the range of academic, extracurricular, and work experiences of students before the 12th grade, which is only possible with the HSB sophomore cohort.
[31]The adjusted r2 for these models indicate that the explanatory variables can generally explain less than 5% of the variance in log wages. This is typical of these kinds of cross section log wage models.
[32]In other words, there is no way to obtain an accurate measure of each individual's wage rate, and labor supply decisions cannot be separated from the wage rate. Consequently, earnings are almost certainly measured with error, which biases upwards the standard errors on the coefficient estimates in the model.
|
| Stasz, C., & Brewer, D. J. (1999). Academic Skills at Work: Two Perspectives (MDS-1193). Berkeley: National Center for Research in Vocational Education, University of California. |