remains here even after these differences are controlled for. While the test scores of blacks have risen in recent years relative to those of whites (Grissmer et al., 1994), the economic returns to test scores appear to have risen even more rapidly, thus preventing any relative improvement in black-white wage rates (Bound and Freeman, 1992).
A similar story can be told about overall test scores in the population: they rebounded somewhat from their lows in the 1970s (Bishop, 1992), especially among the lowest-scoring groups. But this did not occur rapidly enough to offset the rising returns and resulting increases in inequality associated with these differences. Summary data on the National Assessment of Educational Progress test scores still indicate large fractions of the population scoring at very low levels of reading or mathematical competence (Barton and Kirsch, 1990). The lower means and higher variances in test scores among less educated workers in the United States than in several other industrial countries have also been linked to the greater relative declines in earnings that have occurred here in recent years (Bishop, 1989; Nickell, 1997) as U.S. workers have appeared to have more difficulty adjusting to recent shifts in labor demands.
A few caveats are in order here. The magnitudes of estimated relationships between test scores and earnings or job performance are not enormous—test scores generally account for little more than 10 to 20 percent of the variance in earnings or job performance (Bishop, 1995; Hunt, 1995). Some questions remain as to what the scores really measure and whether their correlations with measured job performance are real or spurious. For blacks these questions are even greater than for whites.15
Nevertheless, employer demand for cognitive abilities seems to be growing even within educational categories and is apparently growing more quickly than the supply of these skills can respond. Thus, the economic returns to these abilities appear to be rising as well.
Employers' Skill Demands: Survey Evidence
The Multi-City Employer Survey
While the evidence cited above clearly indicates that relative employer demand has shifted to workers with more education and better cognitive abilities, many questions remain unanswered. Exactly what skills do employers currently need among their less educated (i.e., noncollege) workers? How do employers seek these skills—that is, how do they screen for them, and what observable credentials and characteristics do they regard as signals of potential