Skip to main content Accessibility help
  • Print publication year: 2007
  • Online publication date: June 2010

7 - Estimation of Dynamic Games Involving Economic Fundamentals


The previous chapter explained the difference between open-loop and Markov Perfect equilibria. The Markov Perfect equilibrium assumes that firms are rational, insofar as they expect their rivals to respond to changes in the state variable; the equilibrium is subgame perfect. In contrast, the open-loop equilibrium is (in most cases) not subgame perfect. For this reason, the Markov Perfect equilibrium is more consistent with the standard assumption of firms' rationality. However, the open-loop equilibrium is easier to estimate. We start by showing how to estimate open-loop equilibria using two examples of discrete-time, dynamic games with different types of data and objectives. We then briefly discuss an estimation approach when firms use Markov Perfect strategies.


The first of our two discrete-time examples, the sticky price model, shows how to estimate an index of market power in a dynamic setting when the econometrician has only industry-level (rather than firm) data. The market power (“conjectural variation”) parameter nests a family of equilibria, including the leading cases of competition, cartel, and symmetric Cournot.

The second example, which is based on Roberts and Samuelson (1988) – the RS model, includes the open-loop equilibrium as a special case. In this example, the econometrician has firm-level data. Because the econometrician assumes that firms behave noncooperatively, rather than estimate an index of market structure, the econometrician is interested in estimating a particular feature of the noncooperative behavior and the implication of that behavior.