Working Paper: CEPR ID: DP967
Authors: Kenneth Burdett; Melvyn Coles; Jan van Ours
Abstract: The matching function describes the flow of job creation as a function of the stocks of unemployed and vacancies. Most empirical work tries to identify such a relationship by regressing the flow of matches (aggregated over the month) on the stocks of unemployment and vacancies measured at the beginning of that month. It is shown that estimates obtained using this procedure will be downward biased if unemployment and vacancies are mean-reverting processes. If the bias is small, the size of the bias is proportional to the length of the period interval. By further aggregating the data, say from monthly to quarterly data, the downward bias should triple. The resulting change in the parameter estimates can then be used to estimate the size of the original bias.
Keywords: matching; temporal aggregation bias
JEL Codes: C13; J64
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| High stocks of unemployment (J64) | Lower measured flow of matches (Y10) |
| Low stocks of vacancies (J60) | Higher measured flow of matches (Y10) |
| Mean-reverting properties of stocks (G17) | Downward bias in estimates (C51) |
| Length of period interval (C41) | Downward bias in estimates (C51) |
| Non-mean-reverting stocks (G19) | No bias in estimates (C51) |