Simulation experiments demonstrate the strengths and weaknesses of the
Panzar-Rosse, Hall, and structural approaches to estimating or testin
g market power. The structural model works well only if it is properly
specified. The Hall methods works well in industries with constant re
turns to scale (CRS); however, even slight deviations from CRS lead to
serious biases. Although easier to use than the structural approach,
the Panzar-Rosse approach cannot distinguish between collusion and com
petition for some technologies. Some empirical applications of the Hal
l and Panzar-Rosse methods are provided.