Unstable matching functions in search theory

The matching function is at the heart of many of labor search models. It has been estimated many times, with a striking empirical regularity: the coefficient that represents matching efficiency appears to decline over time. While it may make sense to see less efficient matching in times of structural change, as it may be happening with the last recession, one would not suspect this over longer samples. Why would labor market performance deteriorate when the information revolution should have reduced matching frictions?

Friedrich Poeschel thinks this is at least in part due to omitted variable bias in the estimation of the matching function. The extant literature relies too much on stocks and neglects various flows, in particular vacancy creation and job seekers beyond the unemployed. Of course, it is not that easy to find such series, or reliable ones. Thus, Poeschel constructs such series from US data using a model of labor market flows. Then, he estimates the matching function using the traditional set variables and then augmenting it with flow variables on the labor supply side. Half of the downward trend in the matching efficiency is taken care of by the new variables. The trend completely disappears once vacancy dynamics are also included. That is reassuring, and maybe one can obtain improving matching efficiency once we get our hands on better data for this.