DP3241 A Test Between Unemployment Theories Using Matching Data

Author(s): Melvyn G Coles, Barbara Petrongolo
Publication Date: March 2002
Keyword(s): hypothesis testing, matching, unemployment
JEL(s): E24, J41, J63, J64
Programme Areas: Labour Economics
Link to this Page: www.cepr.org/active/publications/discussion_papers/dp.php?dpno=3241

A new methodology is described which tests between various equilibrium theories of unemployment using matching data. The Paper shows how to correct econometrically for temporal aggregation effects, where the econometrician’s aim is to identify a matching process using data which is recorded monthly, and also shows how to identify different unemployment theories on the data. As implementing this test requires information on the inflow of new vacancies over time, this Paper uses employment agency data for the UK over the period 1985-99. Although the standard random matching approach provides a reasonably good fit, the empirical evidence provides greater support for ‘stock-flow’ matching. Estimates find that over this period, around 87% of newly laid-off workers are on the long-side of their markets and so match with the flow of new vacancies as those vacancies come onto the market. In particular, these workers’ experience average durations of unemployment which exceed 6 months and their matching rates are highly correlated with the inflow of new vacancies. This job queue interpretation of the matching data has important implications for government policy on long term unemployment and optimal UI. It also suggests that previous estimates of the so-called matching function have been misspecified, which potentially explains the large variation in results obtained in that literature.