[Note: This user note describes the consequences of a problem in the collection instrument for the 2004 SIPP panel that affected the data for wave 3 and later waves . For ease of presentation, this note is written with specific reference to wave 3, but it applies equally, with appropriate adjustments, to the waves following wave 3.]

Prior to each collection wave of the 2004 SIPP panel (except the first wave), the collection instrument is loaded with a "feedback" file that contains the responses for each person collected in the previous wave, or if the person was not interviewed in the previous wave, the responses that were collected in the wave preceding the previous wave (two waves ago). The source wave of these responses is called the “feedback wave.” The feedback-wave responses are referred to – or “fed back” – in selected questions of the current wave in an effort to help the respondent recall and describe changes in his or her status since the previous interview.

By mistake, the feedback file for the wave 3 instrument treated all persons as if they had not been interviewed in wave 2. This error meant that, regardless of what the respondent reported in wave 2, the wave 3 instrument always fed back the person’s responses from wave 1. The resulting misinformation potentially had an adverse effect on the way that the wave 3 instrument collected or interpreted data about previously-reported jobs. If the person had a job at the end of the feedback wave, the instrument was supposed to ask whether the person still had that job in the current wave. For wave 3, however, the instrument inadvertently ignored reported job changes between waves 1 and 2, and asked whether the person still held the job that they had at the end of wave 1, even if the person had ended that job in wave 2 and begun another job. So, instead of using wave 2 information as the starting point for questions about jobs held in wave 3, the instrument used possibly outdated information from wave 1. ¹

The error had the potential to make it appear that the person had a different job in wave 3 than he or she had in wave 2, when, in reality, the person was merely reporting that his or her wave 2 job had continued into wave 3. Such a mistake could have resulted in the collection of faulty data about the existence and characteristics of a person’s wave 3 jobs. Most damagingly, the error could have led to the collection of incorrect data for the person’s employer numbers and the starting and ending dates of his or her jobs, and, consequently, to errors in the measurements of the person’s overall employment status (such as the monthly employer status variable) that are based on these items. The employer number (EENO variable) is a unique and unchangeable number that is assigned to a job in each and every wave in which it is held; it is used to link the edited information for that job across waves. The dates that a job was held in a wave are used to derive whether the person was employed, unemployed, or not in labor force in any given week.

Upon discovering the feedback error, we attempted in the editing process to repair its possible damage. First, we identified all job records in wave 3 that were at-risk of being affected by the problem. We then applied an algorithm to each at-risk job to decide whether it was a continuation of a wave 2 job rather than a new job in wave 3. If the job was judged to be a continuation, we changed the wave 3 employer number of the job to be the same as that of the specified wave 2 job, and we made all other information, such as job start/end dates, agree with the corresponding information of the wave 2 job. Otherwise, we accepted at face value the reported or recorded information for the job (see columns 4 and 5 of the table below).

Table 1 below displays the theoretical consequences of the feedback error and the outcomes of our effort to address them. To make the exposition manageable, the table makes various simplifying assumptions, chiefly (which is true for the overwhelming majority of cases) that people had, at most, one job in any given week. We use “Job A “ to refer to the job that the person held in the last week of the reference period for wave 1; “Job B” to the last job the person held in the reference period of wave 2 (regardless of whether it was held in the last week of the reference period for wave 2) ; and “Job C” to the job the person reported as having in the first week of the reference period for wave 3. Note that Jobs A, B, and C may all refer to the same job with the same employer. The term “reported Job C” refers to the wave 3 job as it was characterized in the responses to the survey collection instrument, while the term “edited Job C “ refers to the Job C as it was characterized after the editing process described above. “R” is short for “respondent.”

The editing decisions for each of the situations in the above table have implications for the confidence that can be placed in the job records that were directly affected, and for the variables of the general labor force section whose values are based altogether or in part on this information.

To assist the user in identifying people whose records may have been adversely affected by the feedback problem, and to help the user evaluate or assess the potential ramifications of using data from these records, we created a flag variable – ACAROVR– that indicates the presence and the potential impact of the feedback problem. There is one such flag variable for each of the person’s jobs: ACAROVR1 is the flag for the person’s first job; ACAROVR2 for the second job (if any).

ACAROVR1 NUMBER(2),

• C Job Variables
• S JB: Flag indicating relationship of job to a previous-wave job.
• Y E
• I W
• :L:
• Flag indicating the relationship of the job to a job held in
• previous wave and the possible impact of the relationship on
• Labor Force variables (see User Note).
• :L:
• V 0 Not applicable
• V 1 Reported as new job but converted to the continuation
• V of a job held at end of previous wave
• V 2 Reported and retained as new job, but possible continuation
• V of a job held at end of previous wave
• V 3 Category 1 or 2 regenerated job
• V 4 Category 3 or 4 regenerated job
• V 5 Category 5 or 6 regenerated job
• V 6 Category 7 regenerated job
• V 7 Category 8 regenerated job
• V 8 Category 9 regenerated job
• V 9 Category 10 or 11 regenerated job
• U Universe: All persons 15+ at end of reference period
• U who had a job during the reference period
• U EPOPSTAT = 1 and EPDJBTHN = 1 and
• U (EJOBCNTR > 0 or ECFLAG =1)

-----------------------

ACAROVR2 NUMBER(2),

• C Job Variables
• S JB: Flag indicating relationship of job to a previous-wave job.
• Y E
• I W
• :L:
• Flag indicating the relationship of the job to a job held in
• previous wave and the possible impact of the relationship on
• Labor Force variables (see User Note).
• :L:
• V 0 Not applicable
• V 1 Reported as new job but converted to the continuation
• V of a job held at end of previous wave
• V 2 Reported and retained as new job, but possible continuation
• V of a job held at end of previous wave
• V 3 Category 1 or 2 regenerated job
• V 4 Category 3 or 4 regenerated job
• V 5 Category 5 or 6 regenerated job
• V 6 Category 7 regenerated job
• V 7 Category 8 regenerated job
• V 8 Category 9 regenerated job
• V 9 Category 10 or 11 regenerated job
• U Universe: All persons 15+ at end of reference period who
• U had two or more jobs during the reference period
• U EPOPSTAT = 1 and EPDJBTHN = 1 and
• U (EJOBCNTR > 1 or ECFLAG =1) 

The values of these variables correspond to the set of conditions in the rows and columns of the above Table 1 that apply to the associated job, as shown in the following Table 2:

Categories of ACAROVR with values 3 through 9 are used to classify people with possibly flawed job information by the characteristics of their overall employment situation. This classification may be useful in assessing the likelihood and potential extent of the harm of the flawed data on the person’s general labor force measurements. The categories use the term “regenerated job,” a reminder that the job was supposedly ended in wave 2, but was brought back to life (regenerated), as far as SIPP is concerned, in wave 3 by the feedback mechanism, possibly in error. The categories of regenerated jobs are defined as follows:

A. Regenerated-Job (RJ) Categories 1 through 5:

The job associated with the ACAROVR value ended in week n (where n= 1 to (17 or 18) ) of the reference period for Wave 3, and the person had another job (job2) in wave 3 that was held in a week k, where k le n, and, letting p represent the last week in the reference period that the other job (job2) was held:

Category 1 RJ : p is greater than or equal to( ge) n and job2 was held in the first week of the wave 3 reference period;

Category 2 RJ: else, p ge n and job2 was not held in the first week of the wave 3 reference period;

Category 3 RJ : else, p is less than or equal to (lt) n and job2 was held in the first week of the wave 3 reference period;

Category 4 RJ : else, p lt n and job2 was not held in the first week of the wave 3 reference period;

Category 5 RJ : else, the person had another job in wave 3 that was held in week n+1;

Category 6 RJ : else, the person had another job in wave 3 that was held after week n+1;

Category 7 RJ : else, the person did not have another job in wave 3.

B. Regenerated-Job (RJ) Categories 8 through 11:

The job associated with the ACAROVR value did not end in the reference period for Wave 3, and, the person had another job in wave 3 that was held in week 1 of the reference period, and:

Category 8 RJ :the person still had that other job at the end of the wave 3 reference period;

Category 9 RJ :else, the other job ended in the wave 3 reference period;

Category 10 RJ : else, the person had another job in wave 3 that was not held in week 1 of the reference period;

Category 11 RJ : else, the person did not have another job in wave 3.

Example for Category 1 RJ:

Let's say that we have a respondent named Bob. The information in the ACAROVR1 variable refers to Bob's job with BIG Company. The reference period for Bob in Wave 3 began in June 2004 and ended in September 2004. There were 18 weeks in this reference period, so n, which represents the nth week in the reference period and designates the week that Bob ended his job with BIG Company, lies between 1 and 18. Bob quit his job with BIG Company on July 16, 2004, which means that he quit the job in the seventh week of the reference period, and so n=7. Bob also had a job for Small Company in Wave 3. He started this second job on June 1, 2004, and quit it on August 6, 2004. This means that he held the job with Small Company in weeks 1 through 10, so k can take on a value between 1 and 10, and there is at least one value of k that is less than or equal to n: in other words, Bob held this second job in at least one week that was earlier or the same as the week in which he quit his job for BIG Company (that is, k= (1,2,3,4,5,6, or 7) is <= n). For Bob's second job, p=10, because he quit the job with Small Company in week 10. Since (p=10) is greater than (n=7), and since Bob started to work for Small Company in week 1, his job with BIG Company is classified as a Category 1 RJ.