Attachment 1: Census Industrial & Occupational Classification Codes

Attachment 1: Census Industrial & Occupational Classification Codes

1990 and 2002 Industry and Occupation Codes

Through Round 5, respondent industry and occupation were coded using the 1990 census code frames. As part of the 2000 Decennial Census, the Census Bureau and the Bureau of Labor Statistics revised the industry and occupation code frames. Beginning with the Round 7 data release, 1990 codes are available for Rounds 1-5 jobs. Codes from the 2002 Census code frames are available for all jobs. In the future, only 2002 codes will be available for jobs that the youths report. The move to the new frame makes sense in part to maintain comparability with the Current Population Survey (CPS) and other federal surveys, which have adopted the new frame. In addition, using historical code frames becomes increasingly problematic over time as new industries and occupations arise that are inadequately handled in the old frame. At the same time, changing frames can introduce disruption into the longitudinal record; it may not always be appropriate to change frames whenever updates occur. The NLSY79 data continue to support codes to multiple historical frames in order to retain longitudinal comparability.

A second change underlies the shift to the 2002 code frame. While NORC had previously performed all coding to the 1990 code frames, the industry and occupation coding tasks were completed by the Census Bureau for the 2002 frames. There are documented differences in coding practices across the two organizations, so house effects in coding are likely to surface in the data. The process followed by the Bureau includes use of the respondents' reports of usual duties, title, etc., and contextual information on respondent income and education. Problem cases are selected for manual review by an experienced coder or coding supervisor.  Rates of manual review are similar for the NLSY97 as for the CPS. Wherever possible the Census Bureau has attempted to implement for this survey the coding procedures in place for the CPS.

Users should note that the 2002 frame differs considerably from the 1990 frame in organization, level of granularity, and other characteristics (see the Census Bureau website at for more information). Analyses of NLSY97 as well as CPS data indicate that jobs within a single category of one frame may disperse broadly to a variety of codes in the other frame.

Military Industries and Occupations

Researchers should be aware of several issues regarding military industries and occupations. The questionnaire is designed so that respondents in the active Armed Forces should not be asked to report the industry and occupation of that job in the same questions that collect information about civilian jobs. Rather, respondents report their military occupational specialty as part of a series tailored to military jobs that also asks about the branch of the Armed Forces, the respondent's dates of enlistment, military pay grade, etc. This means that the military industry and occupation codes should appear very rarely in the NLSY97 data, occurring only if a respondent has a civilian job but is associated with the military in some way. However, some respondents in the active military answered the civilian jobs questions instead of the military questions in rounds 2 and 3, causing military respondents to be included in the industry and occupation variables as well as other regular employment questions like hourly rate of pay. This occurred for two reasons.

First, the question that identifies military jobs is YEMP-58500, the class of worker question, which asks whether the respondent's employer was the government, a private employer, a non-profit organization, a family business or farm, or the active Armed Forces. Respondents in the military should respond that they are in the active Armed Forces, and they are then asked the series of questions tailored to military jobs. However, some respondents with military jobs appear to have answered this question incorrectly; most of these respondents reported that they worked for the government. These respondents were then asked the set of questions about regular civilian employee jobs.

Second, some military respondents in round 3 were incorrectly routed through the civilian jobs section due to a programming error. If a respondent states that he or she is still employed with an employer reported in a previous round, the class of worker question is not asked. Instead, the respondent is routed to the correct series of questions based on a preset flag that identifies the job as military or civilian. However, this flag was not set correctly for respondents who reported military employment in round 2, and so these respondents were routed to the civilian series of questions in round 3.

These problems affected a very small number of respondents in rounds 2 and 3.However, researchers interested in civilian employment may want to remove these respondents from their analyses; conversely, users investigating military employment may want to include these respondents even though they did not answer the military series of questions. Researchers can identify respondents in the active Armed Forces by examining the variables MILFLAG and MILCODE on the employer (YEMP) roster. These roster items were edited to identify military jobs regardless of whether they were reported in the military or civilian section of the instrument. Virtually all military jobs appear as job #01 on the roster. Information about using roster variables in analyses is provided in Appendix 8 in this Codebook Supplement.