Price-cost margins, prices and concentration in US manufacturing: a panel study

Abstract

This study tries to separate the efficiency and market power effects of higher concentration for a panel of 253 US manufacturing industries covering the years 1963 to 1992. To do so, two fixed effects estimations of price equations are introduced. The first includes, among the independent variables, average variable cost and a mark-up term dependent on the four firm concentration ratio and isolates the partial effect of concentration on price holding cost constant (the market power effect). The second omits average cost and uses the bias induced to make inferences about the efficiency effect. The results suggest both market power and efficiency effects exist, with the latter dominating. Specifically, for given average cost, higher concentration leads to higher prices, but when average cost are relegated to the error term, higher concentration leads to lower prices.

Notes

For a summary of the two perspectives, see Martin (2002).

Martin finds evidence for both market power (significant concentration coefficient) and efficiency effects (significant productivity advantage), while Allen finds only market power effects and Chappelle and Cottle only efficiency effects.

Gisser (1984) also relates concentration changes to changes in industry price and unit cost changes from1963 to 1972 and, like Peltzman, concludes in favour of efficiency effects that dominate price effects.

Depending on the oligopoly equilibrium, I can be represented by different measures. For a Cournot equilibrium the appropriate index is the Herfindahl, for a dominant firms model the appropriate index can be a k-firm concentration ratio. See Dickson (1981) for details.

The conclusions are not sensitive to how the price equation is specified. Full log linear and full linear specifications produce similar results.

The problem is not caused by spurious correlation that can be an issue in fixed effects when the time series is sufficiently long. In the sample, the yearly average for the price-cost margin moves from 0.46 in 1963 to 0.54 in 1992, while the capital-output ratio moves from 0.25 to 0.32. Concentration increases from 0.39 to 0.41.

The authors use the capital-output ratio (KQ) and this variable interacted with industry growth (GR) and national unemployment rate (UR). From their Table 4 the partial effect of KQ on the price-cost margin in their fixed effects regression is (standard errors in parentheses) 0.058 (0.026) + 0.009 (0.27) GR −1.711 (0.359) UR. This is negative for all sample values of GR and UR. The authors do not explain the discrepancy.

Crandall and Winston (2003) found, using least squares, for a panel covering 20 2 digit SIC industries with annual observations from 1984 to 1996, a sample where time series variation should be prominent, that increases in KQ significantly reduce the price-cost margin.

The implicit assumption is industries are not large enough to influence input prices. A Hausman test did not reject simultaneity, although straightforward fixed effects estimation produces very similar results.