One of the key differences between the two estimates is that the GAO number reflects illegal employment at any point during the year, rather than during an average week. As done here with the March, 1997 CPS for employment throughout 1996, the GAO study used the retrospective questions from March, 1989 in which respondents reported on any employment that took place during 1988. When those retrospective data are adjusted to reflect illegal employment in an average week in 1988 (by weighting the data by the proportion of reported weeks worked in 1988), the GAO estimate of 166,000 is reduced to 86,800, or by almost half.

Both numbers - those illegally employed in an average week and those illegally employed at any point during a year - are meaningful measures of illegal employment, and are included in Table 1. Even accounting for the difference in type of measure, however, the derived GAO estimate of 86,800 15-year-olds illegally employed in an average week is almost twice as high as this study's estimate. The remaining discrepancy is accounted for by methodological differences in defining apparent illegal employment. Those differences, and their relative importance, are:

1. The GAO estimate overcounts illegal employment in several important industries. It counted 15-year-olds as working illegally if their principal job was in a firm in the manufacturing, construction, mining, public utilities, communication, or transportation industries. As that study notes (1991: 41), this will overcount those working illegally because non-manufacturing and non-mining occupations within manufacturing and mining firms are legal for 15-year-olds, as are office and sales jobs in the other industries provided the work is away from transportation equipment and the actual construction site. In contrast, the current study counts illegal employment based on restricted occupations rather than industries.{1} When this studyÕs codes are used to re-classify 1988 employment, 55,100 is subtracted from the GAOÕs estimate of 166,000 illegally employed in a year, and 38,100 is subtracted from the derived GAO estimate of 86,800 illegally employed in an average week.

2. The GAO estimate overcounts illegal employment by including youths who are self-employed or working for family businesses. The former are not covered by the FLSA (which applies to employers), while youths working for family businesses are covered by the hazardous occupation restrictions but not by hours restrictions. When the adjustments are made, 9,500 is subtracted from the GAOÕs estimate of 166,000 illegally employed in a year, and 4,700 is subtracted from the derived GAO estimate of 86,800 illegally employed in an average week.

3. The GAO estimate potentially overcounts hours violations with its procedure for imputing hours violations during the school year. The March CPS asks respondents for the number of weeks worked and the usual hours worked during those weeks. The GAO study counts youths as employed illegally if they reported more than 40 hours usually worked per week, since that is not allowed at any time of the year for 15-year-olds. It also counts youths as employed illegally if they reported more than 18 hours usually worked per week and 16 or more weeks worked last year, since those working that many weeks would have worked some during the school year. A problem is that youths may not have worked more than 18 hours during a school week but still reported "usually" working more than 18 hours since that was the summer schedule. If one excludes those youths who report usually working 19-39 hours for 32 or fewer weeks (since half or more of those weeks were probably worked during non-school weeks, leading the respondent to refer to non-school weeks for "usual" hours), 43,200 is subtracted from the GAO's estimate of 166,000 illegally employed in a year, and 20,600 is subtracted from the derived GAO estimate of 86,800 illegally employed in an average week.

When the adjustments for potential overcounts described in 1-3 above are combined (eliminating overlap), the result is that 74,300 15-year-olds are estimated to have been employed in violation of federal law at some point in 1988, and 36,200 in an average week. This estimate does not include those employed in violation of state laws, most of which apply federal or higher standards.{2} When the latter number is adjusted upward to include all those working in apparent violation of federal or state laws in an average week, the estimate becomes 41,200. While one would not expect this number to exactly match the 1995-97 estimate of 47,400 employed illegally in an average week (due to time differences and problems in comparing retrospective and current measures), the numbers are clearly of the same magnitude, indicating that this decomposition has accounted for the major differences between the GAO estimate and this studyÕs estimate.

In sum, the difference between the two estimates can be traced both to differences in data presentation (illegal employment in an average week vs. at any point during a year) and in how illegal employment was coded. The coding differences are clearly important, since the elimination of apparent overcount in the GAO study reduces that estimate by about half. While the methods in this study were used in order to produce the most accurate estimates, there clearly remain a number of ways in which the survey-based data can undercount illegal youth employment (discussed both here and in GAO, 1991). While some further knowledge may be gained by using different methods and assumptions on existing datasets, a more important project is to generate new high-quality data on youth employment.

FOOTNOTES

{1} The exceptions are that the agricultural industry is excluded from the CPS estimates, and several industry codes are used to denote illegal employment for minors under state laws (generally based on whether the business sells or serves alcohol).

{2} This and the other estimates reported in this Appendix have been adjusted to reflect FLSA coverage using the ratio from the GAO study.