printable version of page Printer-friendly page

Focus On Basics

Volume 2, Issue B ::: June 1998

The GED: Whom Does It Help?

Results from a new approach to studying the economic benefits of the GED

by John H. Tyler
Does acquiring the GED increase the earnings of drop outs? At least 13 different studies in the last decade have examined aspects of this question. Not one of these studies, however, was really able to separate the impact of the credential from the contributions that unobservable factors inherent in GED holders, such as motivation, might make. A quirk of policy enabled me and my colleagues at the Harvard Graduate School of Education, Richard J. Murnane and John B. Willett, to develop a unique approach to looking at this question. Our results differ considerably from those of our colleagues. We found that, unlike almost all previous studies, at least for young white drop outs, acquisition of a credential of General Educational Development (GED) can have a substantial impact on earnings.

Previous research into the economic benefits of the GED points to relatively inconsequential increases in hourly wages, annual earnings, or employment for GED holders relative to drop outs without a GED. In contrast, our study shows that young white GED holders receive a large boost in annual earnings if they acquire a GED. Our treatment group was drop outs age 16 to 21 who passed the GED with scores that were at or just above passing: what could be described as on the margin of passing. Our comparison group was drop outs age 16 to 21 who had the same marginal scores on the GED but, because of different passing requirements in their states, did not receive the credential. When we compare our treatment and comparison groups, we find that the annual earnings of the white treatment group of GED holders, five years after they received the credential, are ten to 20 percent higher than the annual earnings of the comparison group of drop outs who do not possess a GED. This is a very large percentage increase, but it represents an increase in annual earnings of only about $1,500, leaving the clear message that the GED cannot be counted upon as a sole ticket out of poverty.

We were able to conduct separate analyses for white and nonwhite drop outs, and we find no statistically significant differences between the annual earnings of the treatment and comparison groups of nonwhite drop outs. I will discuss this rather surprising and distressing finding later and offer possible explanations.

Different Methodology

As I stated above, these findings come from a study that uses a different methodology than has previously been employed in GED-related research. GED holders are a self-selected, rather than random, group. Given this, failure to account for factors that may cause some drop outs to pursue a GED while other, seemingly similar, drop outs do not results in estimates biased away from the truth. For example, if it is the most motivated drop outs who tend to pursue the GED, then failure to account for this will overstate the effect of the GED on drop outs. Our methodology accounts for this self-selection bias by starting with a data set of drop outs who have all chosen to attempt a GED. We then use the fact that different states have different GED passing standards to compare drop outs who have the same GED exam scores, but who do or do not have a GED depending on the state in which they attempted the exams. With this methodology, our treatment group individuals with a GED is composed of drop outs who are on the margin of passing, but have a GED because they are in a state with a lower passing standard. Meanwhile, our comparison group individuals without a GED is composed of drop outs who are on the margin of passing, but who do not have a GED because they are in a state with a higher passing standard. We are able to account for the fact that our treatment and comparison individuals come from different states and these states may have different labor markets, cost of living, etc.


The data we used to conduct this study are also unique to GED-related research. Past research relied on data sets such as High School and Beyond or the National Longitudinal Study of Youth, which do not have details on GED scores or attempts at passing. Our data were supplied by the GED Testing Service and the state Education Departments in Connecticut, New York, and Florida. These data contain basic demographic information and critical to our methodology GED test scores for drop outs who were age 16 to 21 in 1990, the year they last attempted the GED exams. We have data from most, but not all, states on these 1990 GED candidates. Notice that everyone in our data, passers as well as non-passers, has selected' themselves into the pool of drop outs who would like to have a GED, as indicated by the fact that they attempted the battery of GED exams. To obtain an outcome measure, we worked with programmers at the Social Security Administration (SSA) to merge these GED data with SSA annual earnings data, yielding a data set containing basic demographic information (including states where the GED was attempted), GED test scores, and annual Social Security-taxable earnings. To allow the GED time to take effect in the labor market, we measure annual earnings in 1995, five years after our sample last attempted the GED.


Understanding the mechanisms through which a GED might have an impact on the earnings of drop outs is necessary to interpret our results properly. There are three.

· If preparation for the GED tests tends to increase cognitive skills, and if we assume that higher levels of cognitive skills lead to increased earnings, then there is a human capital component' to the GED.

· Many post-secondary education and training programs are denied to uncredentialed drop outs, but open to GED holders. To the extent that post-secondary education and training lead to increased earnings, then the GED's function as a gateway' to these programs would result in higher earnings for GED holders.

· Gaining information about the future productivity potential of job applicants can be a difficult and expensive enterprise. Employers may value the GED as a signal of unobservable or costly to observe productive attributes. If so, then drop outs who use the GED to signal' higher levels of motivation, maturity, commitment to work, or other productive attributes would tend to have higher earnings than drop outs who lacked the signal.

As a result of our research design, our estimates measure only the value of the GED as a labor market signal. Two factors lead us to this conclusion. First, since our treatment and comparison groups have the same GED test scores, the two groups are balanced on the human capital dimension: on average, the treatment and comparison groups have the same skill levels as measured by the GED exams. Thus, any observed differences in earnings cannot be the result of differences in underlying skills of the two groups; hence, there is no human capital component in our estimates of the effect of the GED on earnings.

Second, since other research we have conducted indicates that the lowest scoring GED holders those who make up our GED treatment group acquire very little post-secondary education or training, our estimates have essentially no gateway component. This leaves only labor market signaling as an explanation for the earnings differences we find between GED holders and uncredentialed drop outs. Thus, our results are correctly interpreted as the labor market signaling effect of the GED on the earnings of young drop outs who choose to acquire a GED and whose skills place them on the margins of passing.


This study has certain limitations that result from SSA confidentiality requirements and the methodology we employ. As a result of federal guidelines designed to protect the confidentiality of individuals, the data released to us by the SSA impose three constraints on our study. First, we have to group all individuals who are not white into a single category, thus destroying the ability to examine whether the GED affects the earnings of African-Americans, Hispanics, Asians, and other minority groups differently. We can only speak to the overall average effect of the GED on this nonwhite group. Second, we cannot explore potential gender differences in the effects of the GED on earnings. And third, we cannot examine the impact of the GED for older GED holders. In future work, using different data, we will be able to retain our methodology and explore these important racial-ethnic and gender issues.

Our methodology, which allows us to address heretofore intractable selectivity-bias issues, also imposes some limitations on what we can say. As a direct result of our methodology, we can do no more than speculate about the following questions that are important to a better understanding of how the GED works in the labor market:

·  How large are the average human capital or gateway components of the GED?

· What is the effect of the GED on the earnings of the random drop out, a sample that includes drop outs who would never voluntarily select into the GED pool?

· What is the effect of the GED on higher scoring GED holders?

While it is important to point out the limitations to this study, a discussion of what we cannot say should not overshadow what we can say with this research. Namely, that we have very credible findings indicating that, at least for young white drop outs, there is a substantial payoff for individuals who chose to pursue acquiring a GED in 1990 and whose skills place them on the margin of passing.

Exploring Results

Given the interpretation of our results, we have to ask why employers would appear to value the GED as a signal of productive attributes for young, relatively low-skilled white drop outs, but not value it as a signal of the potential productivity of similar nonwhite drop outs? One possible answer to this question is that for young nonwhite drop outs, employers may place a higher value on other signals, such as language or residential address, than on the GED signal. To be explicit, consider this hypothetical situation. Two young, nonwhite drop outs apply for the same entry level job. One has a GED, the other does not. All things observable to the employer being equal, we might expect the GED-holder to have an edge. In this example, however, I assume that all things are not equal. The GED-holder in this hypothetical situation speaks English as a second language (this is observable to the employer), and as a result the employer gives the job to the uncredentialed native speaker. This type of behavior on the part of employers could lead to the results we find. A parallel example would adhere for residential address. Another (and not mutually exclusive) explanation for our different white / nonwhite results contrasts the signaling effect of the credential for two different types' of GED holders. According to this hypothesis, some individuals actively seek to obtain a GED to convey a level of maturity or commitment to work, and some GED holders tend to acquire the credential primarily as a quasicompulsory' part of some larger program such as Job Training Partnership Act (JTPA) training programs, or Job Corps activities. It may be that the GED conveys very different information when garnered in these two different ways, with employers discounting the GED signal' when it is coupled with public assistance programs. If this hypothesis were true, and if substantially more nonwhite than white GED holders obtained their credential in a quasi-compulsory manner, then this could explain our results. Our best estimates for the percentage of 1990-minted GED holders who may have acquired their credential in conjunction with a public assistance program are 44 percent for nonwhites and only 11 percent for whites. While these numbers do not prove the hypothesis, they at least work in a direction that lends credence to this explanation.

Finally, other work we have done suggests a third explanation (see page 22). Using data on GED candidates from Connecticut and Florida, we find that a substantially larger proportion of young white GED candidates pass on the first attempt than do African-American or Hispanic candidates. In these data, about 75 percent of white drop outs pass on the first attempt, while only about 60 percent of the Hispanic and 45 percent of the African-American candidates passed on the first attempt. We also find that regardless of race-ethnicity, about the same percentage of first-time failers attempt a second or third time. If we believe that some unknown proportion of these multiple-attempters would pass as the result of chance high scores, and that this proportion is the same across racial-ethnic groups, then the result would be a higher proportion of nonwhite candidates who have a GED as a result of chance, relative to white drop outs. Furthermore, it is logical that most of these false positive' GED holders (drop outs who have a GED as a result of chance high scores) have scores that place them in the margin-of-passing' zone that we use to construct our treatment and comparison groups. Under this scenario, it is plausible that over time employers might tend to discount the signaling value of the GED for nonwhites whose skills are relatively low, which could explain the white-nonwhite differences we find in the data.

‘’ Reality

A logical question is: How do these results fit my experience? The important point to keep in mind is that any one person's particular experiences would only represent a tiny fraction of our data. That is, our estimates represent the average impact of the GED over the nation. This average could represent a world where the impact of the GED is about the same for everyone in the sample; it could represent a world where half of the individuals in the sample get a big boost out of the GED, while the other half get virtually no benefit; or, it could represent a world where there is a complete range of impacts associated with GED attainment. We can only present the average effect for young white drop outs and the average effect for young nonwhite drop outs. As a result, any one piece of anecdotal evidence as to how the GED works in someone's community may or may not fit the story that our estimates present. This is the limitation of quantitative research: we cannot say what will happen to any one individual. This compares to the limitation of qualitative research, which is the inability to generalize findings to the population of interest. Thus, each type of research has its own strengths and weaknesses. Ideally, both types of research are used to inform policy and practice, keeping in mind what each can and cannot say.

Policy Implications

Our research finds that employers value and use the GED as a signal of skills and attitudes they consider to be important in jobs. The message for policy here is that, at least for young white drop outs, the GED is serving an important function for both employers and drop outs. It is a relatively easily accessible and inexpensive way for drop outs with certain attributes to signal to employers that they are a good employment risk. Put another way, in the absence of more complete information, using the GED as a signal is a cost-effective way for employers to chose among drop out job applicants. The puzzling lack of a signaling effect we find for young nonwhite GED holders is a critical question. We cannot provide further answers to this question with our current data, but we are already working to secure more appropriate data that will be used to address this line of inquiry.

Even for white drop outs, however, the elements of good news contained in our findings must be leavened with the fact that the large percentage effects we find translate into relatively small real earnings gains of only $1,500 per year. Young GED holders have very low average annual earnings to start with, and so a $1,500 per year increase appears as a large percentage gain. Thus, we should remember that while the GED can lead to important earnings gains, by itself the credential is not a route out of poverty.

About the Author

John H. Tyler is Assistant Professor of Education, Economics, and Public Policy at Brown University. He completed this work for NCSALL while finishing his doctorate at the Harvard Graduate School of Education. Tyler taught middle school mathematics for eight years.

Full Report Available The research report upon which this article is based is available from NCSALL for $10. To order, write to NCSALL Reports, World Education, 44 Farnsworth Street, Boston, MA 02210-1211; or e-mail

Updated 7/27/07 :: Copyright © 2005 NCSALL