Consider Becker's (1973) classic matching model, with unobserved types and stochastic publicly observed output. If types are complements, then matching is assortative in the known Bayesian posteriors (the `reputations'). By the same token, any extension of Becker's non-stochastic matching model theory to a dynamic world would preserve assortative matching in every period.

We consider the two changes together (the dynamic stochastic world that we all live in), and discover a robust failure of Becker's result. In the first period of the two period model, assortative matching is neither efficient nor an equilibrium for high enough discount factors. A failure occurs for either the highest or the lowest types.

Absent a definite last period, the failure of assortative matching is no longer unqualified. We instead produce a labor theoretic story: Assortative matching fails around the highest (lowest) reputation agents for `low-skill (high-skill) concealing' technologies. We then show that as the number of production outcomes grows, almost all technologies are these two forms. Our theory implies the dynamic result that high-skill matches (like the Beatles) eventually break up.

Given the PAM failure, we show how induced information rents create wage profile discontinuities, and an upward wage-bias not justified by higher productivity.