Here, I show that the fluctuation in the PTE stock comes from changes in the flow-probabilities of employed workers to and from PTE rather than changes in the flows to or from non-employment. Looking at the reasons reported in the CPS for why a worker may be at work part-time highlights the two different ways in which workers may become underemployed. Workers can be classified as PT E due to slack work or business conditions, or they may be unable to find a full-time job. The first reason would indicate that firm-level demand for currently employed workers drives changes in PTE, and would coincide with FT o PTE (within employment) flows. The second reason of failure to find full-time work would coincide with U o PTE flows (across employment/nonemployment states). The fluctuation in the within- employment flows and the lack of any fluctuation in the flows across PTE and U/N indicate that firm-level demand and the “slack work” reason are the primary source of cyclical variation in part-time employment for economic reasons.
To establish that within-employment flows fluctuate, examine the cyclical properties of flows between FT and PTE in Column 1 of Table 1.3. The flows of full-time employed to PTE are strongly countercyclical and 6.4 times as volatile as aggregate output. Meanwhile, flows from PTE to FT are strongly procyclical. In contrast with the cyclical movements in the FT o PTE flows, the across- employment/nonemployment flows of workers between PTE and U in Column 2 of Table 1.3 have very little correlation with aggregate output. The difference in the cyclical variation in the within-employment and across-employment/non-employment flows is obvious if we plot the probabilities of unemployed workers to both full-time and part-time for economic reasons. Figure 1.3 shows the variation in the U FT flow (the full-time job finding probability) in comparison with the nearly constant
U PTE flow probability. If we suspected the popular narrative of the PTE worker as someone who failed to find a full-time job to be the source of the fluctuations in PTE, we would expect that in recessions, a higher fraction of the unemployed would move to PTE. While it is true that more workers in total arrive to PTE from U in recessions, this increase is entirely due to the size of the unemployment stock rather than an increase in the probability of a worker moving to PTE.
To establish that the cyclical movements in the PTE rate are not just within- employment but due to within-job hours changes, I use the dependent interview structure of the post-1994 CPS to classify which transitions in full-time and part-time
status coincide with job changes. In months 2-4 and 6-8 of a respondent’s survey, the interviewer asks if the individual is still employed with the employer named in the previous month. As shown in Fallick and Fleischman (2004), this dependent interview makes it possible to track job-to-job transitions for workers who remain employed in consecutive months. I note which transitions between F T , P T N , and P T E correspond with a change in job. Especially with part-time and full-time status changes, one relevant concern is that the gain or loss of jobs for multiple job holders could be responsible for a substantial fraction of the F T – P T E flow. Therefore, I include any change in the listed primary employer or any change in the number of jobs held for multiple job holders as my measure of a job transition. Figure 1.4 plots the separate transition probabilities of workers from F T to P T E while staying with the same employer for transitions to PT E while experiencing a job transition. Comparing the flow probabilities from F T to these states, it is clear that the ma jority of the F T – P T E flow and its cyclical movements come from within-job changes in hours worked rather than from job-to-job transitions. Flow probabilities from P T E back to F T via the same job or a job transition (either a job-to-job transition or an increase to multiple jobs) are also similarly plotted in Figure 1.5.
It is also useful to observe the fraction of transitions which correspond with a job change. Looking at flows from F T to F T, P T N, and PT E in Figure 1.6, it is interesting to note that until the recovery from the Great Recession, the share of FT – P T E transitions which come from a job change is declining. This decline is especially steep during the last recession. Looking at Figure 1.7, the fraction of flows out of P T E that are due to job changes is also relatively small and declining. 80 – 90% of transitions from PT E back to full-time employment come from within-job. These figures also show that the majority of transitions into and out of P T E by individuals who remain in employment occur within-job. This fact is not specific to just the FT o PTE flows. In fact, of all job transitions of employed workers into or out of PTE (excluding PTE – PTE flows with the same employer), 82% of all transitions occur within job.
One limitation of the CPS data is the inability to observe whether or not these movements of workers between employment states correspond with specific job transitions, especially for multiple job holders. While the CPS tracks the number of jobs for multiple job holders and job transitions are clearly identified for single job holders, The only job identity maintained is the primary job of the respondent . One cannot distinguish a job transition for a multiple-job holder who remains at their primary job if the number of jobs held remains constant in both months. For example, there is no way to tell if a multiple-job holder moved from FT to PTE because of a job-to-job transition in their secondary job. Fortunately, labor force participation data in the Survey of Income and Program Participants (SIPP) tracks all jobs held in each month, allowing for all job transitions to be tracked. I find that in the 1996 panel of the SIPP, The fraction of flows into and out of P T E which correspond to job changes are similar to those found in the CPS data. 75% of all flows into or out of P T E do not coincide with any job transition.
Before moving on to the model, it is important to establish that although the cutoff for defining part-time vs. full-time work is somewhat arbitrary at 35 hours per week, the actual hours worked by full-time and part-time employees reflect the notion that the part-time workweek is significantly shorter than full-time work. This is especially important to note for the case of part-time work for economic reasons, as it is possible that within-job movements between F T and P T E could be an artifact of firms reducing workers’ hours only slightly below the 35 hour cutoff. This is not the case, however, as both part-time for economic reasons (23.11 hrs/wk) and noneconomic reasons (21.61 hrs/wk) work roughly half of full-time hours (44.55 hrs/wk).