'Second-chance' strategies for women who drop out of school

Monthly Labor Review

Volumn 123, Issue 12, 2000, Pages 19-31

  Boudett, Kathryn Parker ^a   Murnane, Richard J ^b   Willett, John B ^b  

^a Harvard Children's Initiative   (United States)

^b HARVARD GRADUATE SCHOOL OF EDUCATION   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0012638164     PISSN: 00981818     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Article

References (35)

1. Types of training are summarized in this way because inconsistencies in the collection of data make it difficult to distinguish reliably between the various types of training in all survey years. For example, funding for the National Longitudinal Survey of Youth (NLSY) shifted in 1987 from the Employment and Training Administration to the Bureau of Labor Statistics, resulting in a decreased emphasis on distinguishing between government and other types of training. (See NLSY User's Guide (Columbus, OH, Ohio State University, Center for Human Resource Research, 1995), chapters 1.14, 1.35.)
2. The learning of skills on the job as part of the experience of working, but not as part of a formal training program, is not deemed on-the-job training.
3. G. S. Becker, Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education (New York: National Bureau of Economic Research, 1964).
4. M. Spence, "Job Market Signaling," Quarterly Journal of Economics, August 1973, pp. 355-74.
5. S. V. Cameron and J. J. Heckman, "The Nonequivalence of High School Equivalents," Journal of Labor Economics, vol. 11, no. 1, pp. 1-47.
6. S. V. Cameron, "Assessing High School Certification for Women Who Drop Out," paper presented at the annual meeting of the American Economics Association, Washington, DC, January 1995.
7. Cameron also conducted a longitudinal analysis in which he included dummy variables indicating whether a particular wage observation pertained to an individual's pre-GED or post-GED period. When he averaged the wages earned during all periods before the individual obtained a GED and compared the figure with the average wages earned during all periods after the student received the GED, Cameron found that the difference was not statistically different from zero. However, when he broke down the pre-and post-GED periods more finely, Cameron found that the longer the interval after GED receipt, the larger was the wage gap. In fact, the difference between wages observed 5 or more years after attainment of the GED and wages observed before attainment of the GED was nearly 6 percent, a finding that is consistent with the estimates of the impact of the GED on wages 5 years after certification presented later in the article. Cameron found that he could not reject the hypothesis that, as a group, the coefficients on the post-GED attainment periods were different from those on the pre-GED attainment periods. This finding is consistent with the hypothesis that it takes several years for GED certification to affect wages.
8. J. J. Heckman, J. Hsse, and Y. Rubinstein, "The GED is a 'Mixed Signal': The Effect of Cognitive and Non-cognitive Skills on Human Capital and Labor Market Outcomes," paper presented at the National Bureau of Economic Research Summer Labor Economics Workshop, Cambridge, MA, July 2000.
9. J. Cao, E. Stromsdorfer, and G. Weeks, "The Human Capital Effect of General Education Development Certificates on Low Income Women," Journal of Human Resources, vol. 31, no. 1, 1996, pp. 206-28.
10. R. J. Murnane, J. B. Willett, and J. H. Tyler, "Who Benefits from Obtaining a GED? Evidence from High School and Beyond," Review of Economics and Statistics, February 2000, pp. 23-37.
11. J. H. Tyler, R. J. Murnane, and J. B. Willett, "Estimating the Labor Market Signaling Value of the GED," Quarterly Journal of Economics, May 2000, pp. 431-68.
12. For two reasons, analyses of formal adult education programs might be more likely than analyses of national data sets to find a positive impact of the attainment of a GED on earnings. First, in preparing students to take the exam, formal programs may actually help participants improve their human capital; national surveys include GED holders who took the exam with little preparation. Second, evaluations of programs tend to model the impact of GED certification on labor market outcomes as occurring over time, with a period of years allowed for follow-up. This is not the case in some studies based on national data sets.
13. J. M. Bos, The Labor Market Value of Remedial Education: Evidence from Time Series Data on an Experimental Program for School Dropouts, Ph.D. dissertation, New York University, 1995.
14. D. Friedlander, P. K. Robins, and J. M. Bos, The Labor Market Value of GED Attainment and Higher TALS Scores for AFDC Recipients in GAIN, research paper supported by the U.S. Department of Education, New York, Manpower Demonstration Research Corporation, 1995.
15. See R. B. Freeman, "Occupational Training in Proprietary Schools and in Technical Institutes," Review of Economics and Statistics," August 1974, pp. 310-18; T. J. Kane and C. E. Rouse, "Labor Market Returns to Two- and Four-Year College," American Economic Review, June 1995, pp. 600-14; and L. M. Lynch, "Private-Sector Training and the Income of Young Workers," American Economic Review, March 1992, pp. 299-312.
16. See R. B. Freeman, "Occupational Training in Proprietary Schools and in Technical Institutes," Review of Economics and Statistics," August 1974, pp. 310-18; T. J. Kane and C. E. Rouse, "Labor Market Returns to Two- and Four-Year College," American Economic Review, June 1995, pp. 600-14; and L. M. Lynch, "Private-Sector Training and the Income of Young Workers," American Economic Review, March 1992, pp. 299-312.
17. See R. B. Freeman, "Occupational Training in Proprietary Schools and in Technical Institutes," Review of Economics and Statistics," August 1974, pp. 310-18; T. J. Kane and C. E. Rouse, "Labor Market Returns to Two- and Four-Year College," American Economic Review, June 1995, pp. 600-14; and L. M. Lynch, "Private-Sector Training and the Income of Young Workers," American Economic Review, March 1992, pp. 299-312.
18. Data in this analysis are not weighted, because weights generated for use on the full sample of 12,686 respondents would not necessarily be applicable to the selected sample of 689 female dropouts.
19. In 1980, 94 percent of the original 1979 NLSY sample took the Armed Services Vocational Aptitude Battery (ASVAB), a series of multiple-choice tests. A composite score derived from four subsections of the test is used to construct the AFQT score. The scores used in this article are based on the 1989 calculation of the composite ASVAB score. (See NLSY User's Guide, p. 52.) All scores are corrected for age on the basis of the full NLSY sample.
20. We thank Nachum Sicherman and Ann Bartel for sharing their code for calculating training in the NLSY. Training and college data are collected differently: training questions allow for a calculation of the number of hours of training obtained, whereas college questions tend to focus on the distinction between number of years of education attended versus number of years completed.
21. See, for example, R. J. Murnane, J. B. Willett, and F. Levy, "The Growing Importance of Skills in Wage Determination," Review of Economics and Statistics, May 1995, pp. 251-66.
22. R. J. Murnane, J. B. Willett, and K. P. Boudett, "Does Acquisition of a GED Lead to More Training, Post-Secondary Education and Military Service for School Dropouts?" Industrial and Labor Relations Review, October 1997, pp. 100-16.
23. Only after 1987, however, did NLSY surveys begin asking respondents about training programs of less than 1 month's duration. This may explain the pattern of shorter training spells for older women.
24. One of the criteria used in this article to accurately identify dropouts in the NLSY is that the individual not have attended college. Among the very small number of individuals who did not have a GED, but who did report having attended college, most appeared to be traditional high school graduates who had missing data on the question about whether they had a high school diploma or a GED credential. Because most colleges require a high school diploma or a GED certificate for admission, the restriction was imposed that women who reported college have one of these credentials.
25. Hypothesis tests indicated that interactions allowing the earnings growth trajectory to be different for Hispanic women were not needed in any of the models used.
26. Many alternative specifications for the effect of attaining a GED on earnings were explored; the model presented here was found to be the best. There was no evidence either of a temporary dip in earnings immediately preceding the acquisition of the GED (a possibility discussed in O. Ashenfelter, "Estimating the Effect of Training Programs on Income," Review of Economics and Statistics, February 1978, pp. 47-57) or of a difference in the impact of the GED on the earnings of women with different initial AFQT scores. The possibility that women with different skill levels would be affected differentially by their acquisition of a GED was explored; no statistically significant pattern was found. This result differs from that obtained with a similar model applied to men: wages of men whose initial skill levels were weakest were affected the most by their acquisition of the GED. (See Murnane, Willett, and Boudett, "Do Male Dropouts Benefit from Obtaining a GED, Postsecondary Education, and Training?" Evaluation Review, October 1999, pp. 475-502.) The result is also different from what Tyler, Murnane, and Willett ("Labor Market Signaling Value of the GED") found regarding women in the High School and Beyond database. In addition, all of the models examined were refitted with a random-effects specification. The fixed-effects specifications tended to produce larger impacts from GED and training, but smaller impacts from college, than the random-effects specifications generated. The fixed-effects results are emphasized because a Hausman test indicated that the random-effects specification is not justified. Results from these alternative specifications are available from the first author upon request.
27. Many alternative specifications for the effect of attaining a GED on earnings were explored; the model presented here was found to be the best. There was no evidence either of a temporary dip in earnings immediately preceding the acquisition of the GED (a possibility discussed in O. Ashenfelter, "Estimating the Effect of Training Programs on Income," Review of Economics and Statistics, February 1978, pp. 47-57) or of a difference in the impact of the GED on the earnings of women with different initial AFQT scores. The possibility that women with different skill levels would be affected differentially by their acquisition of a GED was explored; no statistically significant pattern was found. This result differs from that obtained with a similar model applied to men: wages of men whose initial skill levels were weakest were affected the most by their acquisition of the GED. (See Murnane, Willett, and Boudett, "Do Male Dropouts Benefit from Obtaining a GED, Postsecondary Education, and Training?" Evaluation Review, October 1999, pp. 475-502.) The result is also different from what Tyler, Murnane, and Willett ("Labor Market Signaling Value of the GED") found regarding women in the High School and Beyond database. In addition, all of the models examined were refitted with a random-effects specification. The fixed-effects specifications tended to produce larger impacts from GED and training, but smaller impacts from college, than the random-effects specifications generated. The fixed-effects results are emphasized because a Hausman test indicated that the random-effects specification is not justified. Results from these alternative specifications are available from the first author upon request.
28. To make interpreting the regression table and figures easier, the unemployment rate is centered so that it is equal to the actual rate minus 7.78, the average local unemployment rate in the sample.
29. Participating in training and attending college are modeled differently from obtaining a GED. Because obtaining a GED is a one-time event, it is possible to model earnings after receipt of the GED as a smooth quadratic function. By contrast, participating in training and attending college can occur in multiple bursts of unequal length and duration, making a quadratic function inappropriate for those predictors.
30. Very few women, however, report completing a full year of on-the-job training during any given year. Among women with positive values for CurrentOnJobTraining, the median number of hours of training received was 3 percent of a year (34 hours). This figure corresponds to an earnings boost of $11,731 × 0.03 = $352 during the year of training.
31. Notice that the sample sizes differ for each dependent variable. The annual-earnings analysis is based on the full sample of 689 women and includes years in which a woman reported zero earnings. The probability-of-employment analysis, which is estimated by using a conditional logit function (following G. Chamberlain, "Analysis of Covariance with Qualitative Data," Review of Economic Studies, January 1980, pp. 225-38), is based on a subsample of the full sample of 689 women for technical reasons. (The conditional logit analysis removes person-specific heterogeneity by conditioning on the number of times that the dependent variable equals unity for each individual. Because the model is identified only for individuals with variation in the dependent variable, to estimate the probability of employment using a fixed-effects strategy, 148 women (21 percent of the sample) who worked in none or all of the observed periods must be dropped from the sample.) Finally, the annual-hours-worked and "log"-hourly-wage analyses include only observations whose outcomes are nonzero; women who report zero values of the outcome in all periods are dropped from the sample.
32. Using data only through the 1983 NLSY survey, Lynch was able to isolate off-the-job training programs that were not provided by a government agency. Lynch's analysis found that training in such programs was associated with higher hourly wages for individuals without high school diplomas. During the early 1990s, the Federal Government closed approximately 1,000 proprietary schools that had made fraudulent use of student financial aid programs. It is possible that the economic pay-off to students of the remaining 3,000 or so proprietary schools is higher than the estimate presented in the current article, which is based on data from the 1980s. (See Lynch, "Private-Sector Training.")
33. One reason the estimate presented here of the impact of a year of college on wages is not as large as that of other researchers (see, for example, Kane and Rouse, "Labor Market Returns," who found the impact of attending college on wages to be 5 percent) could be that the sample studied was a particularly disadvantaged group of women who may not have had access to the highest-quality college programs.
34. Statistical Abstract of the United States (U.S. Bureau of the Census, 1999), table 762 (using a 1990 threshold, because all earnings mentioned in this article are reported in constant 1990 dollars).
35. See Measures and Methods for the National Reporting System for Adult Education: Implementation Guidelines (U.S. Department of Education, June 2000).