equipment will change employers' relative demands across skill categories. Workplace organization and job descriptions are clearly malleable over a long-term time horizon, and employers frequently make choices about whether or not to upgrade hiring requirements along educational or experience dimensions (e.g., Levy and Murnane, 1995).
Even in the short run, with a given set of jobs, employers face a variety of choices regarding needed skills. For instance, employers who have difficulty filling vacant jobs could attract more (and presumably better-qualified) applicants by choosing to pay higher wages.7 Alternatively, firms could invest more in recruiting, through advertising or the use of employment agencies (Holzer, 1987).
Finally, employers could generate more highly skilled employees by training the workers they hire, rather than demanding applicants who have certain skills ex ante. Clearly, the strategies of increasing wage levels or recruitment might only redistribute a fixed number of skilled workers among employers; the strategy of training workers would help to generate a higher overall level of skill in the work force.
Firms can choose how much training to invest in employees on the basis of prospective market returns for such training through, for example, higher employee productivity. If the skills employers hope to generate are completely general—that is, can be used in many types of work settings—they will generally choose not to bear the costs of such training since employee turnover may cause them to lose their investments. In this case, employers will provide such training only if they can transfer the costs to employees (by paying them lower wages) or can reduce turnover (through apprenticeships, etc.). As the skills needed become more specific to an industry, an occupation, and, especially, an individual firm, employers should be more willing to share in these costs (Becker, 1975).
But a firm's willingness to make these investments might be limited by a variety of market imperfections, such as wage rigidities, financial constraints, and short-term planning horizons.8 Furthermore, the provision of training might actually cause employers to raise, rather than lower, their ex ante skill requirements if they view certain personal skills as being complements to, rather than substitutes for, the ones they hope to provide through training. (See, e.g., Lynch, 1992, and Cappelli, 1996, for some mixed evidence on this issue.)
The training choices of firms might therefore reinforce gaps or mismatches
between their own skill needs and those of less educated workers, instead of helping to mitigate those gaps.
Inferring Applicants' Skills
Employers might have a clear sense of what tasks need to be performed on their jobs and what skills and personal characteristics are necessary for performing those tasks. But those skills are often not directly observable to an employer at the time of hiring. In other words, the employer will often not know very much about an applicant's prospective ability to perform well on the job. Therefore, employers look for a variety of personal credentials, such as level of education, previous job training or work experience, and references. During interviews, they also look for a variety of personal characteristics, such as social and verbal skills and attitude. Other screens, such as tests (either of cognitive abilities or specific job tasks), are sometimes used as well.
An applicant's personal characteristics or test results are used as signals or predictors of future productivity on the job, rather than considered to be indicators per se of the skills required for job performance. Of course, the ability of these credentials and characteristics to actually predict job performance may be quite limited (Bishop, 1993). Employers' perceptions of some of them, especially attitudes, are inherently subjective and could lead to discriminatory hiring outcomes as well. Indeed, at least some firms are aware of potential legal constraints on their ability to use certain screens that they cannot tie directly to job performance.9
More generally, the costs to employers of obtaining various kinds of information (such as school grades, transcripts, and criminal background checks) might outweigh the information's potential usefulness, thereby discouraging employers from seeking information that would better enable them to judge worker quality.10 Thus, mismatches between jobs and workers could result from employers' lacking information that would enable them to identify skilled applicants as well
as from low overall skill levels among applicants.11 Both of these problems can stem from "market failures" in private-sector labor markets, and a variety of policy interventions may be appropriate if evidence of these problems is found in labor market data.
General Economic Evidence
The above discussion suggests that, when the demand for certain skills grows relative to their supply in the labor market, those skills will generate higher returns (at least in the short run). Therefore, one way of making inferences about "skill gaps" is to review the general empirical evidence on labor market returns to various skills and how they have changed over time.
Returns to Education
The simplest and most easily observable measure of labor market skill is the number of years of schooling and educational degrees that an individual has obtained. On this dimension the evidence is strong and very clear: the returns to education have risen quite dramatically in recent years. For instance, in 1979 the average weekly earnings for young college graduates were about 45 percent of those for high school graduates; by the late 1980s the ratio had risen to about 85 percent (Katz and Murphy, 1992).
This rising gap in earnings between more and less educated workers has coincided with a dramatic decline in the real hourly earnings of less educated males, especially among the young. Thus, the real wages of male high school graduates between the ages of 25 and 34 declined by over 20 percent from the late 1970s to the late 1980s (Katz and Murphy, 1992); the declines for male high school dropouts were even larger. In contrast, real wages rose modestly for young males with college or higher degrees and rose substantially for college-educated young females (Bound and Holzer, 1995). (The gender gap in earnings declined at all levels of education during the 1980s, even while inequality was growing across other dimensions. See Blau and Kahn, 1994, for explanations of why this might have occurred.)
These changes in relative and real wages across groups parallel the changes that have occurred in employment rates (or annual hours worked). Basically, employment and labor force participation rates have risen for young females, especially the more educated, while they have declined quite substantially for male high school dropouts and blacks (Juhn, 1992; Bound and Holzer, 1995). The falling employment rates for young and less educated black males have also