Fig. 1 Unemployment and duration of unemployment: deviations from Hodrick–Prescott trend. Average
duration has been adjusted for the 1994 CPS redesign. Note: NBER dated recessions are shaded in gray
with the official business cycle peaks and troughs. Often lags in the recovery of the
labor market are referred to as jobless recoveries. Estimating the difference between
the phases of the labor market, as measured by the regimes in the unemployment rate,
and the NBER business cycle phases have had mixed results. Boldin (1993) finds that
a MS model of the unemployment rate identifies regimes that align closely with the
NBER phases, with the exception of the recession of the early 1990s, often thought of
as a jobless recovery. Hamilton (2005) reaches a similar conclusion, suggesting that
movements in the labor market are closely aligned with the phases of broad economic
activity. In contrast, Deschamps (2008) estimates a MS model of the unemployment
rate using a Markov Chain Monte Carlo estimator. His results indicate that labor market
contractions that are longer than NBER recessions are the rule rather than the
exception.
To date the phases of the labor market this paper applies a MS model to the unemployment
rate, as well as three series that the literature has yet to examine: the average
duration of unemployment, jobless claims and the exhaustion rate of regular unemployment
insurance (UI). The three additional series have something unique to tell
us about the labor market independent of the unemployment rate. For instance, the
recession of the mid-1970s had a peak unemployment rate of 9.0% compared to 7.8%
for the recession of the early 1990s, while the peak average durations were 16.9 and
19.0 weeks during these periods. As a result, while the unemployment rate indicates
a larger pool of unemployed workers in the mid-1970s, the average duration indicates
that the average costs of being unemployed, in time out of work, was higher in the
early 1990s. Jobless claims, which serve as a proxy for the separation rate, or the
probability of transitioning from employment to unemployment, allow for seperate
anaysis of unemployment inflows. The exhaustion rate proxies the unemployment
survival rate, or the probability of remaining unemployed, which is simply one minus
the job finding rate. The exhuastion rate analysis focuses attention on the outflows
from unemployment.
The regime estimates suggest a complex set of labor market dynamics over the
business cycle, where the separation rate recovers prior to the job finding rate. Similar
to Gordon (2009), who examines the relationship between NBER troughs and peaks
and turning points in jobless claims, there is a close relationship between the regimes
of the jobless claims series and the official NBER business cycle phases.