Unemployment cycles

1 Introduction
Business cycles have a wide variety of origins, ranging from financial crises, over oil price shocks, to
productivity spurts and slowdowns. Often, of all economic agents, workers are those affected most
dramatically, mainly through unemployment. For long, researchers – most notably Diamond (1982)
– have asked whether frictional markets can generate cyclical outcomes, even in the absence of any
exogenous shocks or changes in fundamentals. But so far there has been no compelling mechanism
where the labor market by itself can generate cycles and that fits the facts. In this paper, we propose a
simple theory that generates self-fulfilling unemployment fluctuations and that can account for the key
labor market facts: fluctuations in unemployment, vacancies and job-to-job flows. We will argue that
our model provides a simple rationale for the Jobless Recovery, the fact that it takes a long time for
unemployment to drop even after vacancies and productivity have recovered, and for the outward shift
of the Beveridge Curve.
The main driving force behind unemployment cycles is the search behavior of those with a job.
Singling out the employed to explain unemployment may seem counterintuitive. But with a share of
over ninety percent of the labor force, any minor change in the behavior of the employed who vie
for job openings side by side with the unemployed in the same labor market, has profound aggregate
implications for unemployment. Even if they search much less intensively than the unemployed, because
of their shear size, on average still about half of the job searchers are workers who were employed already.
Most importantly, we document that there is also a huge variation in the composition of the searcher
pool, ranging from 42% of employed workers in the recession to 62% in the boom, mostly due to the
change in the search behavior of the employed over the business cycle. And it is precisely this variation
that has drastic implications for the unemployed during the recession and recovery.
We contribute to the literature by spelling out a model that features a strategic complementarity
between on-the-job search and vacancy creation, giving rise to multiple equilibria. In their search
decision, workers trade off the matching probability against the cost of searching. In turn, in their
vacancy posting decision, firms take into consideration both the expected quality (or productivity) as
well as duration of the job. When workers actively search while on-the-job, the value to the firm of a job
is higher: In the presence of sorting, searchers tend to move to higher productivity jobs, and with active
search the fraction of on-the-job searchers to unemployed is higher. This gives rise to a composition
externality when firms sample workers. At the same time, the expected duration of a job is shorter.
It is precisely the combination of the composition externality and the different duration of a job that
is at the root of the multiplicity. With active on-the-job search, the higher expected match quality
dominates the shorter job duration, which creates incentives for vacancy creation. More vacancies in
turn create incentives for workers to actively search on-the-job since it is easier to find one. Active job
search has become self-fulfilling. This high churning outcome corresponds to an economic boom with
active on-the-job search, high employment-to-employment (EE) transitions, little mismatch, abundant
1