Although the NLS has collected information on labor force behavior since its inception, only partial work histories can be constructed for respondents for certain survey years. The degree of completeness of the work history data varies by survey year.
For those wishing to measure labor force attachment over time, three approaches are available. One can examine (1) the amount of time in weeks that a respondent spent working, unemployed (looking for work), or out of the labor force; (2) the start and stop dates of each job a respondent has held (i.e., a continuous job history); or (3) the start and stop dates associated with each employer for whom a respondent worked (i.e., a continuous employer history).
In general, summary weeks data (i.e., information on the number of weeks working, weeks unemployed, and weeks out of the labor force) were collected during each interview for either the previous 12 months or the previous calendar year. The term "summary weeks data" refers to the respondent's answers (in weeks) to the following types of questions: "During the past 12 months, in how many different weeks did you do any work at all?" Respondents who worked 52 weeks were asked: "Did you lose any full weeks of work during the past 12 months because you were on layoff from a job or lost a job?" Respondents who worked less than 52 weeks were asked: "In any of the remaining weeks, were you looking for work or on layoff from a job?" Those responding "yes" were asked: "How many weeks?" Respondents who did not work during the past 12 months were asked if they had spent any time looking for work or on layoff and if they had, how many weeks. While placement and wording of the individual questions have varied, this core set of summary questions is always present in each interview.
Unfortunately, until the computer-assisted interviews, which started in 1995, data collection consistency did not occur in obtaining information to track all job and/or all employer changes. The gaps in information collected on weeks worked (see discussion below) are minor compared to the gaps in information on jobs held and employment spells. Due to the fact that personal and telephone interviews used different time reference periods, it is only possible to construct a complete job and/or employer record for the later years of the survey.
There are three different ways to construct a summary measure for number of weeks worked, seeking work, or out of the labor force. Users can examine the start and stop dates associated with each job, especially in the personal interview years, when the questionnaire included a detailed work history in a column format. (The titles for these variables can be found on the data file by searching for the words "Most Recent Job.") When the information about start and stop dates is combined, a fairly complete picture of total number of weeks in the labor force can be pieced together. This is the usual procedure that has been used by survey staff to create the *KEY* weeks variables. Users attempting to create number of weeks worked themselves instead of using the created *KEY* variables need to pay close attention to the skip patterns followed in the early survey years. Many check items send respondents to different parts of the questionnaire to respond to questions worded specifically for their particular situations. When constructing number of weeks worked, users should pay particular attention to the dates in the detailed work history section. During the early survey years, the Census Bureau truncated the date the respondent started the job to the preceding interview date if it started before then, so the actual starting date may not be available; in the later years, when an interviewer inadvertently went back before the date of the last interview and gathered information before that date, this information was sometimes left on the data file instead of being blanked out and eliminated.
Two alternatives to this time-consuming procedure of piecing the record together from start and stop dates include (1) use of information from the summary weeks questions present in the questionnaire for all years through 1993 or (2) a combination of data from (a) the *KEY* summary weeks variables for those years in which they were constructed and (b) information from the summary weeks questions for those years in which no *KEY* variable is available. The *KEY* variables (e.g., those variables with titles of '# OF WEEKS WORKED [reference period] *KEY*,' '# OF WEEKS UNEMPLOYED [reference period] *KEY*,' and '# OF WEEKS OLF [reference period] *KEY*') were created for those survey years in which respondents were personally interviewed. Care should be taken to check that the number of cases on the summary weeks variables is reasonably close to the number of respondents interviewed (since all respondents should have a value on these variables). If this is not the case, the user needs to make sure that the desired information is not present in another part of the questionnaire or to adjust for the fact that in some years respondents who had not worked since the last interview are assigned to " missing" instead of being assigned a "zero" for zero weeks of work, as one would expect.
Gaps in the reference periods for the summary week variables occur in the early 1970s when the project phased in an alternating personal and telephone interview pattern. The regularly fielded personal interviews conducted during the early survey years gave way to a 2-2-1 interview pattern (i.e., two telephone interviews occurring two years apart followed by a personal interview at the end of the five-year period). The intent of the telephone interview was to obtain a brief update of information on each respondent and to maintain sufficient contact such that the lengthier personal interview could be completed. Due to the fact that the reference period for the summary weeks questions within a telephone interview was the previous 12 months and that no interview was conducted the year before each telephone survey, gaps in the summary weeks record occurred.
The discussion below reviews the types of summary weeks information that are available from the questionnaire. Included is information on changes in the reference periods for which these data were collected. The weeks worked accounting is not completely accurate due to the slight over- or under-counting of weeks that occurs when a respondent is not interviewed exactly one year from the date of the last interview. If the respondent accurately answered the question on how many weeks in the last 12 months she worked and it had been 13 months since the last interview, the summary weeks variables would miss four weeks of employment status information. Census was asked in the early years to interview each respondent as close as possible to the date of the previous interview; the actual dates of interview can and should be checked.
The 1968 survey collected information from respondents not currently working on the specific year that they last worked. Responses were coded into the following categories: "never worked at all," "never worked two or more weeks," the (specific) month and year if the date was 1963 or later, or a residual category indicating that the last time worked was before 1963. The current or last job is that job held after January 1, 1963. All respondents were asked the summary weeks questions on number of weeks worked, weeks unemployed, and weeks out of the labor force for the previous calendar year (i.e., 1967). If the respondent was not enrolled in school or was working 35 hours or more a week, she was asked about the first job she had held for at least one month after she stopped attending school full-time.
In 1969, those respondents who were currently working or who had held a job since January 15, 1968, were asked about that job; summary weeks questions refer to the last 12-month period. Also, respondents were asked for information on any intervening job (or the longest intervening job, if more than one).
In 1970, the detailed work history column section asked respondents who were currently working or who had held a job since January 1, 1969, about that job (current or last) and about all other jobs. An expanded set of summary weeks questions is present, with a reference period of January 1, 1969.
The 1971 interview repeated the 1970 pattern, with the work history section referring to the date of the previous interview. The 1972 and 1973 surveys repeated the 1971 pattern. Except for respondents who were not interviewed in all years, fairly accurate total number of weeks worked, unemployed, or out of the labor force variables can be created for 1968-73.
The gaps in the summary weeks information began with the 1975 telephone interview. The current or last job questions refer back to the date of the last interview; the summary weeks questions only asked about the last 12 months. The 1977 telephone interview followed the 1975 pattern.
The 1978 personal interview collected data for respondents who had worked since the date of the 1977 interview (or January 3, 1977, if the respondent was not interviewed in 1977). The rest of the survey follows the 1970 pattern. Respondents were also asked for information about the longest job held since January 1973 and for the number of years, out of the past five, that they worked for at least 6 months.
The 1980 telephone interview referred to the date of the 1978 interview (or to January 2, 1978, if the respondent was not interviewed in 1978) for the current or last job and to the previous 12 months for the summary weeks questions. Item 19c obtained information on the number of weeks worked for the 12-month period previous to the last 12 months. Answer categories were "1" through "4" with "1" meaning that the respondent worked most of the year (46-52 weeks), "2" meaning that she had worked more than half a year (26-45 weeks), "3" meaning that she had worked less than half a year (1-25 weeks), and "4" meaning she had not worked at all. By using the midpoint and assigning zero weeks to those respondents who did not work at all, users can approximate the number of weeks worked, although one cannot distinguish between those unemployed and those out of the labor force. The 1982 telephone interview repeats the 1980 telephone pattern using the date of the last interview or January 2, 1980.
The 1983 personal interview collected data for those respondents who had worked since the date of the 1982 interview (or since January 1, 1982, if the respondent was not interviewed in 1982). Respondents were asked for information on their current or last job and on all other jobs held since 1982. The summary weeks questions were asked of all respondents; however, the pattern was slightly different from that used in 1978. If the *KEY* variables are not being used, the user will need to pick up the inputs from different places in the questionnaire in order to create one variable for all respondents.
The 1985 telephone interview referred to the date of the 1983 interview (or to January 2, 1983, if the respondent was not interviewed in 1983) for the current or last job and to the last 12 months for the regular summary weeks questions. The information obtained on weeks worked in the 12 month period prior to the previous 12 months is coded in actual weeks, rather than in a range as in 1980 and 1982. However, it is not possible to distinguish between those respondents who are unemployed and those out of the labor force for the intervening year (i.e., 1983 to 1984). The 1987 telephone interview repeated the 1985 pattern, using the date of the previous interview or January 2, 1985, as the reference point.
The 1988 personal interview collected data for those respondents who had worked since the date of the 1987 interview (or since January 1, 1987, if the respondent was not interviewed in 1987). Respondents were asked for information on their current or last job and all employers (not jobs) for whom they had worked since the 1987 interview. The focus of the work history questions shifted from jobs to employers for whom the respondent had worked three or more consecutive months.
The 1991 interview was conducted in person rather than by telephone, due to the BLS decision to eliminate the 2-2-1 interview pattern and field a personal interview every other year. (The next personal interview was scheduled for 1990, but the survey was delayed a year due to the demands of the 1990 decennial census.) This interview asked respondents about their current or last job and about all employers (not jobs) for whom they had worked since the date of the 1988 interview (or the most recent interview if the respondent was not interviewed in 1988). Due to the fact that this change in the reference date back to the last interview coincided with changes in rules about dropping respondents after two years of noninterview, Census interviewed some respondents whose last interview took place in the mid-1980s. Certain respondents will consequently have work histories that go back past 1988. The summary weeks questions cover the three-year gap in one-year increments. The 1993 interview repeated the 1991 pattern, except that there was only a two-year gap.
The 1995-2003 personal interviews asked respondents about the start and stop dates of their current/last job and any intervening jobs. These start and stop dates were used--in conjunction with their reason for not working--to create summary weeks variables.
Survey Instruments: The work experience data are collected in the "Work History," "Employment," "Work Experience," "On Jobs," or "Employer Supplement" questionnaire sections in various surveys.
Created Work History Variables
The 1999-2003 data releases include a new set of week-by-week employment status variables for the CAPI/CATI interview years. Beginning with the first week of 1994 and continuing through the respondent's most recent interview date, a variable for each week indicates whether the respondent was working (coded "1") or not working (coded "0") that week. A summary variable for each year totals the number of weeks that the respondent worked. These variables can be located on the data file by searching for their question names as follows:
NCV-WORK-xx-01 to NCV-WORK-xx-52 (working/not working each week of year 19xx) NCV-WORKxx (total weeks working in year 19xx)
Missing data are treated in the following manner: If the job start or end year is provided, an unknown or missing day is set to 15, and an unknown or missing month is set to 1 (January). Missing years are not imputed. If days provided are inconsistent for a given month (e.g., April 31), the day is reset to the closest consistent day (April 30). More information is available in Appendix 41 (PDF) in the Young Women Codebook Supplement.
Due to an instrument problem, there were 343 respondents for whom one or more employers were inadvertently deleted from the employer roster in 2001, resulting in some information missing for these jobs. Information for these jobs was back-collected for all but 24 of these respondents. Researchers looking across years may thus see an under-reported number of employers in 2001. For a list of the 398 respondents for whom some employer information may be missing, see Appendix 42 (PDF) in the Codebook Supplement.