NLSY79 Appendix 28: NLSY79 Employer History Roster

The Employer History Roster (EHR) has been available to public data users since 2013. This sizeable set of variables incorporates a substantial amount of information about each employer into a single record for each. See Appendix 13: Development of Questionnaires and Codebooks for a description of rosters as a data structure. Programs that created the EHR can be accessed in the Employer History Roster Creation Programs pdf file.

Introduction

Throughout NLSY79 history, it has been a common exercise for users to track employer records through multiple survey years. The traditional means of linking job records is somewhat involved, requiring multiple steps and identification of job numbers in current, previous and subsequent survey years (See Appendix 9: Linking Employers Through Survey Years).

The EHR is intended to alleviate if not eliminate the burden of linking jobs for users. In the EHR, employer records through survey years have already been linked into a single continuous cumulative record for each employer ever reported by a respondent. The roster currently contains data for up to 67 employers (up to 67 records). This means that at least one respondent has reported 67 jobs over the course of the NLSY79's 29 survey rounds, spanning 42 years. Most respondents of course will have a much smaller number of jobs reported since 1979 and only a single respondent has reported 67 jobs. 

While the EHR does not contain all information ever reported about each employer, it does contain a substantial amount of commonly used data items that are consistently found through the survey's history. Users can also more easily incorporate job-related data items that are not currently on the roster. Instructive examples of how the roster data can be used appear below. Variables that are part of the EHR are available through the NLS Investigator search and extraction tool. These variables can be found in Investigator by using the areas of interest that start with 'EMPLOYER HISTORY,' question names that start with 'EMPLOYERS_ALL' or reference numbers that start with 'E.' The variables in this roster are classified as 'XRND', meaning that they are not linked to any specific survey year (see the Glossary of NLS Terms for a full definition of XRND). Instead, they are updated each round with the most current cumulative data available.

Please note: Variables referenced in the following discussion are described in Table 1, which lists the contents of the Employer History Roster.

Roster Construction

The process of constructing the historical EHR began by establishing preliminary linkages similar to that described in Appendix 9. Employers were tracked between current, previous and subsequent survey years using multiple employer numbers in each survey year. After preliminary matches were established, linkages and discrepancies were examined in detail, often using in-house data not available publicly. This might include items such as occupation and industry descriptions and interviewer comments made during the course of an interview, in conjunction with employer names. Examining these items could in some cases serve to clarify connections between jobs reported in various years. Consequently, at least some employer records built in the EHR may be of higher quality than those that can be established using publicly available data. 

Employer Order

Employers are listed on the EHR in chronological order by the earliest start date reported for a job. For instance:

• Employer #1 will be the job with the earliest start date ever reported for any job.
• Employer #4 will be the job with the 4^th earliest start date. There will be three jobs with earlier start dates in the roster, numbered Employers #1, #2 and #3.

Please note that the order of an employer on the roster (from 1-67 possible) is based solely on the original start date (the earliest start date available) at the first report of that job. It is not based on the survey year in which the job was first reported. Consider the following example:

• A respondent has reported 3 jobs through the 1987 survey year. These jobs with original start dates ranging from 1978 to 1984, have the following pattern:
  • Employer first reported 1979, original start date = Jan. 1978
  • Employer first reported 1983, original start date = Nov. 1982
  • Employer first reported 1987, original start date = Jan. 1987

During the 1988 interview, the respondent reports a 4th employer that s/he failed to report at an earlier interview with an original start date of March 1985. While this job will be numbered by the 'unique identification number' as a job reported in the 1988 survey year, its order on the roster will precede the job reported in 1987. Although the 1987 job was reported first, it actually started later. In this simple example, the order of employers on the roster would appear as:

• Employer #1:  first reported 1979, original start date = Jan. 1978, unique id = 19790100
• Employer #2:  first reported 1983, original start date = Nov. 1982, unique id = 19830100
• Employer #3:  first reported 1988, original start date = Mar. 1985, unique id = 19880100
• Employer #4:  first reported 1987, original start date = Jan. 1987, unique id = 19870100

These types of reporting irregularities along with possible isolated errors in recording and data entry errors are not common but are recurrent throughout the life of the survey. As many inconsistencies as possible were resolved in the construction of the EHR.

Roster Content

The vast majority of variables currently included on the EHR are directly copied from responses collected during the survey or from data items already coded or created separately for the public release dataset. For example, start and stop dates, currently working status and class of worker are all responses collected during the interview. Occupation and industry variables are coded from responses collected during the interview. Start and stop weeks and array job numbers are variables created by the Work History programs (see Appendix 18: Work History Data).

Table 1 lists the contents of the Employer History Roster by area of interest and question name. The EHR currently contains data for up to 67 jobs ever reported.

(a) Variables created for 'each job for each survey year' will have one variable picked up for each survey year for each job the respondent has reported. For instance, the EMPLOYERS_ALL_CURWK_[YEAR].[JOB#] set of variables will include the 'currently working' status for each job reported, in each survey year that a job was active. If a respondent reports employer #1, spanning from 1980-1989, a variable will be present for employer #1 for each survey year (EMPLOYERS_ALL_CURWK_1980.01, EMPLOYERS_ALL_CURWK_1981.01, EMPLOYERS_ALL_CURWK_1982.01...EMPLOYERS_ALL_CURWK_1989.01). If the respondent reports employer #2, spanning from 1985-1990, another set of those variables would be available for employer #2 for each of those survey years (EMPLOYERS_ALL_CURWK_1985.02, EMPLOYERS_ALL_CURWK_1986.02, etc.). Users should note that the EMPLOYERS_ALL_GOVJOB_[YEAR].[JOB#] set of variables only exist through 1987, when that question was dropped from the survey.

(b) Variables created for 'each job' will have one variable per employer. EMPLOYERS_ALL_UID.[JOB#], EMPLOYERS_ALL_STARTDATE_ORIGINAL.[JOB#]~[D/M/Y] and EMPLOYERS_ALL_STOPDATE_MOST_RECENT.[JOB#]~[D/M/Y] are all examples of variables that appear only once for each job. Each employer has only one unique identification number, one original start date and one most recent stop date. These variables will be present for each reported employer, regardless of the survey years in which they were reported.

(c) Once assigned, the unique id and original start date for an employer will not change except under two conditions: 1) an error is found in the reporting record that necessitates either a correction to an original start date or a reordering of employers, and/or 2) more employers are recovered from archives that necessitate a chronological reordering of employers for a respondent.

(d) The most recent stop date for an employer will be updated with each successive interview during which that employer is reported, until the respondent leaves the employer and does not return.

(e) Variables created for 'each job for each gap for each survey' are similar to the variables in group (b) above, but are present for up to 4 gaps reported within the tenure of each job. For example, if a respondent reports two gaps within the tenure for employer #2 and one gap within the tenure for employer #3 in 1986, gap variables would be present for each employer for that survey year. EMPLOYERS_ALL_WHYNOWK_[YEAR].[JOB#].[GAP#] variables in that case would be EMPLOYERS_ALL_WHYNOWK_1986.02.01 and EMPLOYERS_ALL_WHYNOWK_1986.02.02 (gaps within tenure with employer #2), and EMPLOYERS_ALL_WHYNOWK_1986.03.01 (gap within tenure with employer #3).

Roster Structure

The Employer History Roster is structured with a single 'record' for each employer. The information from survey year to survey year for the same employers has already been linked into this single record. This allows researchers to look at the record for an employer for the entire time the job was reported. Information for each specific job can be identified using the following search criteria:

• Question Name (enter search term) + starts with + 'EMPLOYERS_ALL'
• Word in Title (enter search term) + contains + 'JOB ##'

The NLS Investigator search captured in Figure 1 below will produce all variables on the roster pertaining to the first job a respondent ever reported, regardless of when that job was first reported. Some respondents may have reported their first job in 1979 while others may not have reported their first job until 1985.

Each successive survey year presents the opportunity for a respondent to report additional employers. Consider a respondent with the reporting record depicted in Table 2:

In the example in Table 2, the respondent has reported four employers at the initial 1979 interview, two additional employers at the 1980 interview, and three more new employers at the 1981 interview. 

• The original start date, most recent stop date and unique id variables fall into the category of variables in Table 1 created for 'each job'. They will be present for all jobs reported by a respondent. If by 1990, the respondent has reported 13 jobs, 13 original start dates, 13 original stop dates and 13 unique ids will be present in the EHR for that respondent.
• In the Table 2 example, those variables falling into the other categories in Table 1 (variables present for each job (and each gap if applicable) for each survey year in which the job is reported), will be present only for the survey years listed in column 6 ('Employer information reported in survey years'), as those are the survey years in which information on the specific jobs would have been collected. Examples of the variables that would be present for the respondent and jobs listed in Table 2 are in the final column ('Examples of variables present for each job').

Variables for all jobs will not be present for all respondents. For example:

• A respondent reporting only 5 jobs through 2010 will have no data at all for jobs 6 through 67.
• Likewise, a respondent reporting their 28^th employer starting in 1990 will have no data for that 28^th employer picked up from survey years 1979-1989. In this instance, variables EMPLOYERS_ALL_UID.28, EMPLOYERS_ALL_STARTDATE_ORIGINAL.28~[D/M/Y], and EMPLOYERS_ALL_MOST_RECENT_STARTDATE.28~[D/M/Y] will be present but variables from the specific survey years 1979-89 (for instance,  EMPLOYERS_ALL_OCC_1979.28, EMPLOYERS_ALL_OCC_1980.28, EMPLOYERS_ALL_OCC_1981.28, etc.) will not be present. Those variables will appear for 1990 when the job is first reported and be present for each subsequent survey year in which the job is reported. If this employer is reported in 1990, 1991 and 1992, EMPLOYERS_ALL_OCC_1990.28, EMPLOYERS_ALL_OCC_1991.28 and EMPLOYERS_ALL_OCC_1992.28 will be present for that respondent for job 28. Occupation codes for all other survey years prior and subsequent will be missing for that job for that respondent.

In addition, different series of variables may have missing items, depending on the reporting pattern. For instance:

• Users may find that there is data for jobs 1-33 and jobs 36-40 in 1994, but no data for jobs 34-35. Such a pattern would indicate that no respondents reported what would be their 34^th and 35^th job during the 1994 interview.
• There are also instances in which specific variable series have similar gaps. For instance, in a specific year, one might find industry and occupation codes for jobs 1-33 and 36-40, but again not for the 34^th and 35^th job. Such a pattern could result because no one reported their 34^th and 35^th job that year.  Alternatively if several respondents reported a 34^th/35^th job that survey year, but those jobs did not meet the criteria (based on hours worked and length of time) that would have led to collection of industry and occupation descriptions, there would be no industry and occupation data from that survey year.

Linking to Non-Roster Employer Variables

While a large number of commonly used employer-related variables have been incorporated into the EHR, there are some variables in various survey years that have not. Users can link employer variables from a specific survey year that are not currently on the EHR to the cumulative record for an employer on the EHR by using the EMPLOYERS_ALL_ID_[YEAR].[JOB#] variables. These variables contain the identification number of a job in a specific survey year in which it was reported. The following paragraphs describe several examples of linking scenarios.

Imagine that a researcher wants to link data on promotions asked in 1989 to the cumulative employer record found in the EHR. Using the EMPLOYERS_ALL_ID_1989.[JOB#] variables, one can link the promotions data for the correct employer in 1989 to the cumulative employer record in the EHR. To accomplish this, the value of EMPLOYERS_ALL_ID_1989.[JOB#] would be checked for each employer on the EHR. If that value is '1,' the promotion variables for job #1 in 1989 would be linked. If the value of EMPLOYERS_ALL_UID_1989.[JOB#] is '2,' promotion variables for job #2 in 1989 would be linked, and so on.

In another example, one might want to link the EHR cumulative employer record with the information on the respondent's employee status (regular, consultant, contractor, temp worker) asked in 1994, 1996 and 1998. One would again use the EMPLOYERS_ALL_ID_1994.[JOB#], EMPLOYERS_ALL_ID_1996.[JOB#] and EMPLOYERS_ALL_ID_1998.[JOB#] variables in the same way described above to link to the appropriate employee status variables in each year. If the value of EMPLOYERS_ALL_ID_1994.24 is '1' then the employee status variables for employer #1 in 1994 would be linked to the cumulative record for job #24 on the EHR.  If the value of EMPLOYERS_ALL_ID_1994.30 is '1' then those variables would be linked to the cumulative record for job #30 on the EHR. Following a similar procedure, the employee status variables for 1996 and 1998 could be linked to the appropriate cumulative job record on the EHR.

In addition, it is now possible to identify which cumulative job records on the EHR link to specific job numbers on the Work History Labor Force Status and Dual Jobs arrays. The EHR includes a set of variables called EMPLOYERS_ALL_NUM_ARRAY_[YEAR].[JOB#]. These variables contain the job number assigned by the Work History programs to each job for each survey year. The job number assigned by the Work History programs consists of the survey round and the number of the job for that survey year. So for instance, the 2^nd job reported in 1993 (round 15 of the survey) would appear as job #1502 on the Labor Force Status and/or Dual Job arrays.  If this same job was reported as job #1 in 1994 (round 16 of the survey), it would appear as job #1601 in the Labor Force Status array. The EMPLOYERS_ALL_NUM_ARRAY_[YEAR].[JOB#] variables allow users to identify and track a specific employer through the Work History job arrays more easily. For instance, if a user wants additional information on the first job a respondent reported after a gap, and the Work History array job number is 1701, one would want to identify which of the employers in the EHR (1-59) contains 1701 in the EMPLOYERS_ALL_NUM_ARRAY_1996.[JOB#] variable. Establishing this link to the EHR then provides an expanded series of variables depicting the entire history of the employer.

User Notes

The Employer History Roster was made available after a considerable effort and was a long time in the making. Improvements will continue to be made as time allows. Some possible areas of improvement and expansion are discussed below:

Each survey year there are a small number of cases that report more than five employers. Prior to 1993, jobs 6+ were provided and stored on separate media. With the many technological transformations that have taken place in the past two decades, much of the information on this small set of jobs in each older survey year has become very difficult to recover from original sources, and is generally not included in the EHR. Reporting of more than five jobs per survey year is not common.

As mentioned earlier, the EHR includes a great deal, but not all, information collected on employers in each survey. For instance, some information on employee status (regular, temp worker, contractor, etc.), starting wages and hours of a job, fringe benefits applying to a job, etc. have been collected in various survey years but are not included on the EHR. Users can access this additional data by linking to individual Employer Supplements in the appropriate survey years.