A r c h i v e d I n f o r m a t i o n

Educational and Labor Market Performance of GED Recipients - February 1998

Appendix C: Technical Issues

Does the GED credential improve the labor market opportunities of those who hold it? In comparison to what? To what would have been the case if they had not prepared for, taken, and passed the GED Tests? To what would have been the case if they had graduated from high school? To show a GED effect, we need evidence of how well GED certificate holders are doing in the labor market compared to how they would have done had they not received the GED. To see whether any effects are equivalent to those of a high school diploma, we need to know how the GEDs would have done had they graduated from high school. These hypothetical conditions are termed "counter-factual," because they describe events that did not happen. The challenge for empirical research in dealing with such questions is to approximate the counter-factual condition. Possibilities include (1) comparing GED recipients to other individuals with similar characteristics who did not get the GED or who graduated from high school (cross-sectional evidence), and (2) the comparing the same individuals (and comparison groups) before and after they received the GED (longitudinal evidence). This study examines both cross-sectional and longitudinal evidence from large-scale national surveys.

Four types of survey sample data have been used to study the relationship between passing the GED Tests and labor market outcomes.

1. Follow-up Surveys of GED Test-takers or Passers

These might be called one-step longitudinal studies. There is usually no baseline survey, other than the administration of the test and collection of a modicum of background data at that time. A follow-up survey is conducted some time later. There are usually no additional follow-up surveys after the first one. The strength of these surveys is that they focus on the group of interest. For example, it is common for surveys to follow up individuals who took the GED Tests. However, most follow-up surveys do not include a comparison group. Therefore, inferring the effect of taking the GED can only be done by comparing the GED examinee's status before and after taking the GED. The problem with this approach is that it is normal for individuals to get promotions and salary increases. There is no benchmark or basis for estimating how well the GED test-takers would have done if they had not taken the GED. A variation is to ask the recipients to provide their own judgments about whether getting the GED caused an improvement in employment. This approach is problematic because GED test-takers may not be a reliable source for this judgment.

2. Panel Surveys

These include (1) the National Longitudinal Survey of the Labor Market Experiences of Youth (NLSY), (2) High School and Beyond (HS&B), (3) the National Education Longitudinal Study of Eighth Graders in 1988 (NELS), (4) the Beginning Postsecondary Student Longitudinal Study (BPS), and (5) the October Current Population Survey.

These surveys provide good data for before-and-after comparisons and they avoid recall bias. Respondents are asked about items such as employment, hours worked, weeks worked, salaries, and earnings for the current week or the past year and then surveyed again in one, two, or more years and asked the same questions again. Most longitudinal surveys also provide a rich array of control variables such as measures of academic achievement, family background information, labor force experience, and so forth. They offer as comparison groups both high school graduates and individuals who dropped out of high school and did not get a GED. As was noted for follow-up surveys, labor market rewards typically improve with age and experience. To isolate the effects of the GED from the effects of age and experience requires that they be modeled correctly. An alternative comparison group is high school graduates. This group includes many who went on to postsecondary education and may differ from GEDs in ways related to postsecondary education. Therefore in making comparisons, it is important to exclude those with college experience or control for level of education after high school. Some longitudinal studies, such as the NLSY, follow individuals for 12 or more years. This allows time to measure such differences between GED recipients and other dropouts in, for example, the growth of wage rates.

3. Cross-Sectional Surveys

The strength of cross-sectional surveys is that they typically offer a wide age distribution and a wide range of post-GED periods of time in the labor market. The National Adult Literacy Survey (NALS), for example, is a sample of individuals 16- to 64-years-old who may have taken the GED over a period as long as 40 years. A cross-sectional study provides comparison groups similar to longitudinal studies. However, labor market outcomes are available at only one point in time, and of course, people's labor market experiences vary over time.

4. Experiments

Studies that use any of the three types surveys discussed above, including longitudinal studies, are plagued by the fact that there may be differences, that are important for labor market outcomes, between those who take and pass the GED Tests and those who do not and that are not reflected in the long list of variables available to the analyst. Experiments through randomization provide a means of controlling for the unmeasurable differences between those who get a GED and those who do not. Social experiments are usually difficult to conduct, because researchers can neither force individuals to participate who do not want to nor can they prevent individuals from participating who do want to (and are eligible). However, researchers have taken advantage of the fact that there are sometimes more applicants than a program can accommodate, and a fair way to decide who gets in is through random selection. This makes it possible to compare the experiences of program participants with other similar people who also wanted to participate but who were denied access just by chance. Another way of conducting an experiment, in this case a "natural experiment," is to compare outcomes for similar categories of individuals who can be separated into treatment and control groups by virtue of systematic coincidences, such as laws and rules of states they happen to live in. One experiment that considered GED outcomes was based on random selection among applicants (Cave and Bos 1995). Another compared outcomes for GED test-takers in states with different standards for passing the test (Tyler et al. 1997).
-###-

[Appendix B: Table 7]

[Appendix D: GED Follow-up Surveys]