Many establishments both hire and lay off within a short time window (e.g. within one quarter). This means that there is ‘churn’ (e.g. Burgess et al. 2000 and Davis et al. 2012 for the US). We show that churn is not restricted to establishments that keep their workforce roughly constant. Even fast-growing establishments separate from many workers, and fast-shrinking establishments hire new workers. These patterns of churn at the establishment level are informative about the driving forces and the frictions determining job and worker reallocation.
In a recent paper (Bachmann et al. 2017), we shed new light on worker churn in Germany using the newly constructed Administrative Wage and Labor Market Flow Dataset (AWFP) which covers every German establishment from 1975-2014 at a quarterly frequency. The AWFP aggregates from individual administrative social security data to the establishment level (Seth and Stüber 2017). It contains job and worker flows as well as wage information for the entire universe of German establishments. Therefore, it is a unique, high-quality data source. We also replicate many findings about job and worker flows in the US (e.g. Davis et al. 2012). Perhaps interestingly, in contrast to the US, Germany did not experience a secular decline in job and worker reallocation. However, the main focus of our study is worker churn.
The churning rate is defined as (twice) the number of hires relative to establishment size for shrinking establishments, i.e. for establishments where more workers leave than new workers get hired. It is (twice) the number of workers that leave relative to establishment size for establishments where fewer workers leave than new workers arrive. There may be different reasons why workers leave an establishment that grows. Workers may have better outside options, or the establishment may want to replace them with more suitable workers. Similarly, there are many reasons why an establishment hires workers even though it is shrinking in size.
Figure 1 shows the aggregate churning rate (as a share of total employment) for Germany. Three features are important. First, churn is quantitatively large. During the last four decades, it was never below 5% in any quarter. Its average is larger than that of job flows. Second, the churn rate shows a strongly pro-cyclical pattern (the grey areas in Figure 1 indicate recessions) – it can be 40% larger in booms compared to recessions. Third, we show that, at the aggregate level, churn and job-to-job transitions move almost one-to-one over the business cycle. Thus, the increase of churn in booms is generated by an increase of job-to-job-transitions. The last finding lends support to theories that stress job-to-job transitions as means of pro-cyclical worker reallocation. More generally, we document that booms are times of high job creation and high churn (not high job destruction). This means that churn-induced separations – i.e. job-to-job transitions – ultimately get replaced by some establishments through hiring from non-employment, and vice versa for recessions. What is more, in terms of timing, job creation and churn both start early in a boom, but churn is more persistent and continues to increase into the maturing boom.
Figure 1 Aggregate churn rate in Germany
Note: The figure displays the (seasonally adjusted) aggregate churning rate, CHR, as a share of total employment for all West German establishments. The gray shaded areas represent periods of at least five consecutive quarters of unemployment growth.
The cross-sectional patterns for churn are also interesting. Churn is smallest for establishments with a constant employment stock. It increases for establishments with a large positive or negative employment growth rate (Figure 2). Intuitively, this means that shrinking establishments hire many new workers and growing establishments lose many incumbent workers. On average, shrinking establishments hire more than establishments with a constant employment stock. Similarly, on average, growing establishments separate from more workers than establishments with a constant employment stock.
What could be the reasons behind this cross-sectional pattern? We use a simple dynamic labour demand model as an accounting framework to explain the data. We first document that a standard model with idiosyncratic revenue productivity shocks and employment adjustment costs fails in important dimensions. Churn in such a model is zero for shrinking firms and the highest churning rate occurs for firms with a constant employment stock. By contrast, we find that stochastic separation rate shocks and time-to-hire frictions do a good job explaining the U-shaped churning rate pattern (which is not to say that we think productivity shocks and employment adjustment costs are not important in general, but they are not sufficient to explain the churning data). The fact that rapidly shrinking establishments also hire a substantial number of workers implies that these establishments separate from more workers than they had planned or had foreseen. In a reduced-form sense, our framework interprets this as stochastic separation rate shocks to the establishment. In addition, what matters is that these separation shocks cannot be undone immediately. If it takes time to hire, establishments will try to rehire the separations they expect in excess of their desired employment changes, and when fewer workers leave than what was expected, the establishment grows.
Figure 2 Cross-sectional patterns for worker churn
Note: The figure shows churning rates as a function of the establishment-specific employment growth rates. We represent the employment growth category by its midpoint as an estimate of the average growth in that category. The blue solid line represents the churning rate profile averaged over all time periods (1975-2014). The red dashed (the black dotted) line represents the churning rate profile in the ten quarters with the lowest (highest) HP-filtered unemployment rate, a boom (recession).
Figure 2 also shows that the churn rate shifts upwards in booms, uniformly across the employment growth distribution. This uniform upward shift is difficult to reconcile with the idea that booms are times where workers systematically reallocate to establishments that are desired by all workers. Thus, our results are at odds with theories where workers share a common ranking of establishments, i.e. all workers agree where it is best to work – for instance because all workers only care about wages. In such frameworks, highly ranked establishments have low separation rates and grow more in booms, when workers can easily find new jobs. This means that churn should go down in fast-growing establishments, because fast-growing establishments in booms would be highly ranked places where no worker leaves.
Note that our results do not imply that churn is always something undesired by the establishment. As an example of churn, consider a change of an establishment’s workforce composition when an establishment automates its production, and therefore fires many of its production workers, while, at the same time, hiring automation specialists. Also, sometimes churn will be the result of a worker’s decision – for example, when nonpecuniary aspects of a workplace are revealed only after a worker starts working, rendering the fit between worker and workplace suboptimal.
Bachmann, R, C Bayer, C Merkl, S Seth, H Stüber and F Wellschmied (2017), “Worker Churn and Employment Growth at the Establishment Level”, CEPR Discussion Paper, no. 12343.
Burgess, S, J Lane and D Stevens (2000), “Job Flows, Worker Flows, and Churning”, Journal of Labor Economics, 18 (3), 473–502.
Davis, S, J Faberman and J Haltiwanger (2012), “Labor Market Flows in the Cross Section and over Time”, Journal of Monetary Economics, 59 (1), 1–18.
Seth, S, and H Stüber (2017), “Administrative Wage and Labor Market Flow Panel (AWFP) 1975–2014”, FAU Discussion Paper, no. 1/2017.