This page is located at:

Cognitive Skills Matter in the Labor Market, Even for School Dropouts

John H. Tyler, Richard J. Murnane, and John B. Willett 
Harvard School of Public Health

April 2000

Between 1979 and 1996 the median earnings of 25-34 year-old males who left school before obtaining a high school diploma fell by 30 percent; the corresponding figure for female dropouts is a 21 percent decline. Over this same period the earnings premium four-year college graduates received over the earnings of male school dropouts increased from 60 percent to 133 percent. The primary explanation for these patterns is that the demand for unskilled workers declined relative to the demand for skilled workers. While these between-group changes have been well analyzed, little research has examined what is taking place among the least educated in the workforce. Do skills matter among dropouts in this information age economy?

While the average cognitive skill level of school dropouts is quite low, there is considerable variation among dropouts in cognitive skill levels. One could argue that, in an economy in which basic cognitive skills* are increasingly valued, differences in skills would translate into earnings differences for dropouts just as they do for workers with greater educational attainments. On the other hand, the economic trends that have depressed the average earnings of the less skilled may have relegated most young dropouts to entry level jobs where skills matter very little and consequently are not rewarded. This could be especially true for minority male dropouts, whose earnings in 1996 averaged 28 percent less than those of white male dropouts. This report presents evidence on the labor market payoff to cognitive skills for school dropouts, and whether the payoff differs by gender and race/ethnicity.

We analyze data containing information on the universe of school dropouts in New York and Florida who took the GED exams between 1986 and 1990 and were aged 16 to 21 when they last took these exams. The data released to us by the departments of education in these states contain basic demographic information and GED test scores. To obtain an outcome measure, we worked with the Social Security Administration (SSA) to merge the state GED data with Social Security taxable annual earnings via Social Security numbers. These data allow us to address several issues that generally hamper attempts to explore the impact of cognitive skills on labor market performance.

The first problem is that since the test scores that provide the measures of cognitive skills in most data sets are of no consequence to the test-takers, the scores may provide underestimates of true cognitive skills, especially for individuals who find test-taking distasteful. Our measure of cognitive skills is an individual’s scores on the GED exams. Since the GED is a high stakes test for the dropouts in our data, we anticipate that these scores more accurately measure the cognitive skills of the dropouts in our data.

Second, test scores may be correlated with unmeasured variables such as motivation levels which affect labor market performance. Consequently, correlations between test scores and labor market earnings may reflect the importance of unmeasured motivation, rather than the causal impact of cognitive skills. All individuals in our data have exhibited a desire to obtain a GED, as evidenced by their willingness to prepare for and attempt the lengthy battery of tests. Consequently, there is probably less unmeasured variation in motivation among individuals in this data set than is the case among participants in other surveys.

Third, many data sets only have information on wages. As a result individuals who are not working are often excluded from the analysis, giving concern about selection effects associated with who is and is not working. We measure labor market performance by annual earnings. This allows us to include dropouts with zero earnings (those who did not work within the year) in our analysis sample.

Finally, we measure labor market earnings five years after dropouts last attempted the GED exams. This reduces the problem of assessing the direction of causation of the correlation between test scores and earnings.

Our results, based on the earnings of 21-26 year-old dropouts who attempted the GED exams in Florida and New York in the years 1986 to 1990, have both discouraging and encouraging elements as we think about the labor market prospects of young dropouts in the 1990s.

First, our data show that the average annual earnings levels of young dropouts are quite low. Average unconditional earnings for males range from a low of $9,394 in New York in 1995 to a high of $10,869 in Florida in 1995. The comparable figures for females are a low of $6,886 in New York in 1995 and a high of $7,955 in Florida in 1994. Since these averages contain zero earnings for individuals who did not work in a year, they represent the combined effect of wages and labor supply on earnings.

More encouraging news is that for whites and minority-group members, males and females, skills are an important determinant of earnings. Our results show that in the labor market of the early 1990s, young high school dropouts could expect higher annual earnings if they had higher levels of basic cognitive skills. Our results also show inter-group variation in the returns to skills, with females generally having higher returns than males, and minority group members, especially those with relatively high skills, having greater returns than white dropouts. For example, among dropouts who scored high enough to be awarded a GED, male and female minority-group members who scored in the upper ranges of the GED had annual earnings in 1995 that were nearly $1,300 to $1,400 dollars higher than lower scoring minority-group members who had a GED. The comparable figure for white females was about $950, and only about $150 (and statistically insignificant) for white males. We also found returns to skills among dropouts whose skills were so low that they were unable to pass the GED exams. Among these lower skilled dropouts the premium for higher GED scores (even though that score was too low to meet the GED passing standard) ranged from about $2,000 for white and minority males to about $3,000 for white and minority females.

The significant economic return to modest skill differences among minority-group male dropouts sheds light on a puzzle in the random-assignment evaluation of the Perry Pre-School Program. As Barnett (1996) has described, black males who participated in the early childhood intervention program did not have greater educational attainments, on average, then did black males in the control group.  Yet, the black males in the treatment group did have somewhat higher average earnings at age 27 than did black males in the control group. Barnett speculated that this may have stemmed from the higher average cognitive skills of the treatment group, as measured by test scores during their schooling years. The evidence presented in this paper on the economic returns to skill differences for minority male dropouts supports Barnett’s hypothesis.

The research results in this report contain important policy implications. Welfare reform is pushing many low skilled individuals into a labor market where skills increasingly matter. We find that higher cognitive skills mean higher subsequent earnings for dropouts. This is important information for teachers in GED preparation programs who often face pressure from students to teach to the test at the expense of time spent on skill formation. The message they should impart to there students is passing the GED exam is important, but increasing your skills along the way is also important. However, while our estimates of the returns to skills for many groups are large in percentage terms, we caution that this is largely because these young dropouts have such low average annual earnings to begin with. This is the bad news, and it is a strong argument against a decision to drop out in the first place.

* Basic cognitive skills are defined here as the reading, writing, and math skills, along with vocabulary and background knowledge, measured by tests such as the GED.Other cognitive skills such as problem solving and oral communication are not part of this analysis.


Full Report Available

To receive a copy of the full report, send a request indicating the report number or title
along with a check or money order in the amount of $10, payable to World Education, to:

Caye Caplan
NCSALL Reports
World Education
44 Farnsworth Street
Boston, MA 02210
or call: (617) 482-9485