Archived Information

Interpreting the Evidence

Research on academic course work and labor market outcomes offers evidence of positive effects, but the effects are small and not entirely consistent across studies. There is some tendency for studies that addressed more of the theoretical and statistical complications that confront the research question to observe more often positive effects. Studies that utilized test score controls obtained after course work occurred almost universally found zero or negative effects of course work, whereas studies with controls prior to course exposure (and otherwise reasonably specified) exhibited significant positive effects in two of three studies of wages (for females only: Grasso and Shea, 1979; Daymont and Rumberger, 1982), as well as in the single well-specified study of employment (for both males and females: Daymont and Rumberger, 1982), and in two of three studies of occupational attainment (for males: Griffin and Alexander, 1978; for males and females combined: Hotchkiss and Dorsten, 1987). One other analysis with no controls for test scores, but a statistical model that addressed selection bias, obtained positive effects on wages for foreign language enrollment, and positive effects of math, science, and foreign language combined, but the overall impact was very small (for males and females combined: Altonji, 1992).

A plausible interpretation of this pattern of results is that academic experiences help prepare students to find work, in occupations that are of higher status (relative to jobs held by non-college youth), but have minimal effects on wages, especially for men, in the early years after high school. This conclusion, while very tentatively drawn from studies of course work and job success, is substantiated by research on academic skills and labor market outcomes. Although academic skills have virtually no impact immediately after high school (Bishop, 1985, 1993), in the long run, academic skills are associated with higher earnings, even among persons with the same amount of schooling (Crouse, 1979; Blackburn and Neumark, 1993; Cameron and Heckman, 1993). In the short run, those with higher test scores are more likely to find jobs (Kang, 1984; Kang and Bishop, 1986; Rivera-Batiz, 1992; Rosenbaum and Binder, 1994). Effects on wages in the early career appear larger for women than men, and the impact of skills on wages six years after high school may be increasing (Murnane, Willett, and Levy, 1993). To the extent that academic course work bolsters academic skills, as indicated by numerous studies, it is reasonable to conclude that research on skills and job success supports the conclusion that course work contributes to labor market outcomes.

Combining both sources of evidence yields partial support for predictions stated earlier: (1) Effects of academic skills and courses on wages are weak at labor market entry (non-linear effects of courses were not examined with controls for prior ability); (2) Effects on wages increase with age, but effects on employment do not; (3) No studies considered variation in effects across occupational fields, but effects on wages differ for men and women, which may reflect occupational differences; (4) One study of math skills suggests that positive effects on wages may have grown between the 1970s and the 1980s (Murnane, Willett, and Levy, 1992), and another study suggested that academic program effects on wages may have increased during the 1960s and early 1970s (Grasso and Shea, 1979).

The most troubling contradiction to this assessment and interpretation of the evidence is that no study found significant positive effects of academic course work on annual earnings. This includes not only studies that considered earnings in the early career (Grasso and Shea, 1979; Gustman and Steinmeier, 1982; Gardner, 1984), but also one study with a follow-up 13 years after high school (Griffin and Alexander, 1978). Results obtained by Gustman and Steinmeier (1982) and Gardner (1984) might be explained away by sample and measurement problems, but no such objections were raised for Grasso and Shea (1979), who reported non-significant positive coefficients for men and women, or Alexander and Griffin (1978), who observed a significant negative effect for men. Since other work indicated positive effects on employment in the short term and on wages in the long term, the finding of zero or negative effects on earnings was surprising. Moreover, researchers have reported positive effects of academic skills on earnings, so one would expect the pattern to hold for academic course work. The lack of comparable effects for course work is unresolved, and remains an important qualification to the tentative conclusions of positive effects on wages, employment, and occupational status.

Another concern is that results from studies that bear directly on the question—that is, models of academic curriculum effects on labor market outcomes—have been less consistent than results from studies of curriculum and achievement, and achievement and labor market outcomes. Why is this the case? On the one hand, it is not surprising that studies of more proximate causes and effects exhibit stronger relationships. Each step in the causal chain is weakened by outside complications, not all of which are addressed in the statistical models. Some slippage may be expected.

On the other hand, model specification problems may have led to overestimates in one literature or underestimates in the other. Effects of course taking on achievement may overstate the benefits of academic courses for students who do not pursue further schooling. This is because most results in this literature pertain to the average student, but students who do not pursue postsecondary schooling are probably below average in how much they learn in academic courses (Alexander and Pallas, 1984; Gamoran, 1987a). In that case, generalizing from studies of course work on achievement may yield overestimates of the impact of course work on labor market outcomes for non-college youth.

Another way of looking at the situation, however, suggests that effects of academic courses on job success may be underestimated in many cases. Among students who enroll in academic courses, those who benefit most are presumably more likely to enroll in postsecondary education. Those who take a full complement of academic work but who do not advance to further education may well be those who, for reasons unobserved in the data, benefit least of all from each additional academic course. Analyses of courses and work outcomes may underestimate the impact of academic courses, particularly when the sample is restricted to non-college students, because of an over-representation in the analysis of students who failed to benefit from academic courses. Without information on whether students profited from courses—that is, whether course enrollment led to cognitive gains for the different groups of students, including those who tend to take relatively few academic courses—this ambiguity cannot be resolved.

Empirical evidence gives some credence to the suggestion that negative selection biases have depressed the observed impact of academic courses. First, observed effects with no controls for ability (Altonji, 1992) are much smaller than one would expect, given other evidence of positive effects of ability on wages and earnings (Crouse, 1979). Second, estimates obtained with controls for test scores after course work occurred were largely negative (Kang and Bishop, 1986, 1989), a finding that is difficult to explain unless it reflects negative selection. (Note that Kang and Bishop's models included controls for vocational courses, so the negative effects of academic courses do not simply reflect a foregoing of vocational courses. They may, however, reflect a foregoing of extra-curricular activities or non-school activities, such as part-time jobs.) Third, statistical corrections for labor force participation tend to enhance measured effects, or at least not to depress them (Hotchkiss and Dorsten, 1987). This pattern obviously reflects the greater likelihood of students who take many academic courses to enter postsecondary education rather than the labor force, but it may also reflect a negative selection bias for work outcomes among students who take many academic courses but who do not pursue further schooling. (In contrast, adjustment for labor force participation reduces the observed impact of cognitive ability at ages 25 and 28, [Cameron and Heckman, 1993], because more able persons are more likely to work once they leave school.)

Qualitative work adds depth to the tentative conclusion that the consequences of academic work in high school are more salient for finding a job and rising in one's occupation than for initial wages. Interviews with employers suggested that a minimum level of skills is required to compete for jobs, but beyond this threshold, cognitive skills make little difference at entry level. No quantitative analysis specified this type of non-linear model. Once a job is obtained, prospects for advancement depend in part on prior cognitive preparation. As one employer explained to Rosenbaum and Binder (1994, p. 9):

Research on course work and achievement suggests that relevant skills are enhanced by academic courses taken in high school.

[Indirect Evidence: Effects of Academic Skills]

[Conclusions]