This note discusses a previously known issue with reported business profits and how the problem was addressed in the 2022 SIPP. We first give an overview of the issue. We then give an account of how profits were assigned and propose a methodology for handling the issue in 2014 through 2021 data. We conclude with the 2022 SIPP change.
The SEHSD Labor Force Statistics Branch discovered an error with the collection and processing of business profits (TJB(n)_PRFTB). Since Wave 1 of the 2014 SIPP, the profit question asks:
For [ABC Incorporated], what was the dollar amount of [your] share of the [net profit] for the year [of 20XX]?
(The survey instrument dynamically fills a job’s name, the reference year, and other relevant information. So, for clarity we have substituted in example phrasing).
The question is intended to collect net business profits for the entire year, but it is possible for the question to be asked twice for a given job because the SIPP allows respondents to report up to two spells of employment for each job. As a result, in each spell, respondents are asked to report their net business profits for the reference year.
Two problems arise from asking the question twice:
1) Respondents can contradict themselves and provide different answers each time the question is asked.
1. In some cases, respondents may be reporting spell-level profits. For example, suppose a respondent worked at ABC Incorporated from January to June (spell 1) and then from August to December (spell 2). When asked about net business profits, they report $80,000 of profits on the first spell and $70,000 of profits on the second. This reporting pattern could be interpreted as respondents reporting spell-level profits: reporting the net profits for January to June and then for August to December, such that net business profits for the reference year would sum to $150,000.
2. Another possibility is that a respondent doesn’t recall the value they reported in the first spell and reports a different value in the second spell. For example, suppose the respondent earned $80,000 during the reference year. When interviewed, the respondent correctly reports earning $80,000 of profits in the first spell. But then, when asked a second time in spell 2, they forget their first response and incorrectly report earning $70,000 of profits during the reference year. (Note, that as a data user, we would not have information on which of the two responses is “correct”.)
2) Internal processing treats business profits as a spell-level question even though it was intended to collect annual data, affecting variables both directly and indirectly.
1. TPEARN and TPEARN_ALT are directly affected because the first spell amount is equally divided across the weeks worked in the first spell and the second spell amount is equally divided across the weeks worked in the second spell. For example, suppose the respondent earned $80,000 of profits during the reference year across 44 weeks (24 weeks in spell 1 and 20 weeks in spell 2). Weekly earnings would be $1,818.18. If the respondent reports $80,000 of profits in the first spell, weekly earnings in spell 1 would be $3,333.33; and if they report $70,000 of profits in the second spell, weekly earnings in spell 2 would be $3,500.
2. Indirectly, this problem affects earnings variables imputed using SIPP’s Sequential Regression Multivariate Imputation (SRMI) for individuals who had two spells on a self-employed job. Because profits are an input in SIPP’s SRMI imputation model, the factor used to scale hot-deck imputed earnings takes both reported and imputed profits into account. We discuss how this assumption affects imputed profits more in the next section. For a more detailed discussion of SRMI Imputed Earnings see the User Note “2019 SIPP: User Note for Changes to Imputed Earnings”.
The problems associated with the conceptual discrepancy between how the business profits question text is phrased and how it is processed leads to different outcomes. This section accounts for each possibility.
The following is a proposed methodology to annualize profits for self-employed jobs with two spells. The goal is to provide a methodology that follows as closely as possible the methodology that will be applied in the 2022 SIPP by treating profits as a year-level variable.
The challenge in implementing the proposed methodology is that prior to the 2020 SIPP, users were not provided the EJB(n)_PFTBCHK variable. EJB(n)_PFTBCHK indicates whether a self-employed job’s earnings were included in the reported profits. EJB(n)_PFTBCHK is necessary to calculate the profit value reported at the time of interview, EJB(n)_PRFTB. The profits provided on the public-use file, TJB(n)_PRFTB, subtract off earnings if EJB(n)_PFTBCHK equals 2 to provide net profits. For the 2020 and 2021 SIPP, users will then be able to use EJB(n)_PFTBCHK to calculate EJB(n)_PRFTB and closely follow the methodology applied in the 2022 SIPP. For the SIPP 2014 Panel to SIPP 2019, users will not have access to EJB(n)_PFTBCHK and for some respondents will need to infer or assume a value for EJB(n)_PFTBCHK to calculate EJB(n)_PRFTB.
First, users will need to define the variables below from the data.
Second, users will want to construct EJB(n)_PRFTB, which is the profit value reported at the interview. There are two scenarios to account for:
1. TJB(n)_PRFTB equals EJB(n)_PRFTB: If EJB(n)_PFTBCHK is not equal to 2 on a self-employed job spell, then there is no need to subtract off earnings to calculate a business’ net profits. Users will then want to define EJB(n)_PRFTB as TJB(n)_PRFTB.
2. TJB(n)_PRFTB is not equal to EJB(n)_PRFTB: If EJB(n)_PFTBCHK is equal to 2 on a self-employed job spell, then earnings must be subtracted from EJB(n)_PRFTB to calculate a business’ net profits (TJB(n)_PRFTB). Users will then want to define EJB(n)_PRFTB as TJB(n)_PRFTB plus spell earnings. Or as follows:
a. If EJB(n)_PFTBCHK equals 2 on the first spell, then set EJB(n)_PRFTB on the first spell equal to the first spell TJB(n)_PRFTB + first spell earnings
b. If EJB(n)_PFTBCHK equals 2 on the second spell, then set EJB(n)_PRFTB on the second spell equal to the second spell TJB(n)_PRFTB + second spell earnings
Third, with the spell-level variable EJB(n)_PRFTB defined, users will determine which report to use so that profits are treated as a year-level variable. Our proposed methodology prioritizes using reported data over imputed data, hot-deck imputed data over SRMI imputed data, and any imputation method over cold-deck imputation. Below are the transformations for each two-spell scenario.
Fourth, recalculate TJB(n)_PRFTB for self-employed jobs with two spells.
For survey years prior to 2020, we recommend that users follow an identical approach. But, because EJB(n)_PFTBCHK is not present on the files, users will either need to assume or infer a value. In deciding their approach users should note the universe of EJB(n)_PFTBCHK: all incorporated self-employed jobs (EJB(n)_JBORSE = 2 and EJB(n)_INCPB = 1) and unincorporated self-employed jobs with earnings (EJB(n)_JBORSE = 2 and EJB(n)_INCPB = 2 and EJB(n)_BSLRYB = 1). Therefore, TJB(n)_PRFTB equals EJB(n)_PRFTB for all unincorporated self-employed jobs without earnings (EJB(n)_JBORSE = 2 and EJB(n)_INCPB = 2 and EJB(n)_BSLRYB = 2).
Our recommendation for an assumed value is EJB(n)_PFTBCHK = 1. In the 2020 SIPP, there were a total of 55 jobs with two spells and only 2 of those jobs had EJB(n)_PFTBCHK = 2.
Users may also be able to infer the value of EJB(n)_PFTBCHK by comparing the values of TJB(n)_PRFTB by examining the reported earnings or profits. For example, if TJB(n)_PRFTB is equal on both spells and first spell earnings does not equal second spell earnings, then it is likely that EJB(n)_PFTBCHK equals one.
The issue described above is present from the first wave of the 2014 Panel through the 2021 SIPP. Starting with 2022 SIPP the profits question are only asked once for each job, and the processing of business profits changed to treat the profits variable as a year-level variable.