A matching function is a mathematical relationship that describes the formation of new relationships (also called 'matches') from unmatched
agents of the appropriate types. For example, in the context of job formation, matching functions are sometimes assumed to have the following '
Cobb–Douglas' form: :m_t \; = \; M(u_t,v_t) \; = \; \mu u_t^a v_t^b where \,\mu\,, \,a\,, and \,b\, are positive constants. In this equation, \,u_t\, represents the number of unemployed job seekers in the economy at a given time \,t\,, and \,v_t\, is the number of
vacant jobs firms are trying to fill. The number of new relationships (matches) created (per unit of time) is given by \,m_t\,. A matching function is in general analogous to a
production function. However, whereas a production function usually represents the production of goods and services from inputs like labor and capital, a matching function represents the formation of new relationships from the pools of available unmatched individuals. Estimates of the labor market matching function suggest that it has
constant returns to scale, that is, a+b\approx 1. If the fraction of jobs that separate (due to firing, quits, and so forth) from one period to the next is \,\delta\,, then to calculate the change in employment from one period to the next we must add the formation of new matches and subtract off the separation of old matches. A period may be treated as a week, a month, a quarter, or some other convenient period of time, depending on the data under consideration. (For simplicity, we are ignoring the entry of new workers into the labor force, and the death or retirement of old workers, but these issues can be accounted for as well.) Suppose we write the number of workers employed in period \,t\, as \,n_t=L_t-u_t\,, where \,L_t\, is the
labor force in period \,t\,. Then given the matching function described above, the dynamics of employment over time would be given by :n_{t+1} \; = \mu u_t^a v_t^b + (1-\delta)n_t For simplicity, many studies treat \,\delta\, as a fixed constant. But the fraction of workers separating per period of time can be determined endogenously if we assume that the value of being matched varies over time for each worker-firm pair (due, for example, to changes in
productivity). ==Applications==