Market shares and rivalry in the U.S. cigarette industry

In a recent article Gallet and List (2001) examined whether relative market shares in the U.S. cigarette market were mean-reverting using traditional univariate unit root tests and a test that allows for a breaking trend. Their results indicated most of the series were nonstationary, suggesting rivalry remained strong throughout most of the 20th century, a result confirmed recently by Adhikari (2004) using a different methodology. The purpose of this note is to apply univariate and panel tests of stationarity to determine whether relative market shares can be assumed to be individually trend-stationary. If the relative shares are characterized by a unit root, the stationarity tests should confirm the results of the unit root tests. Additionally, panel unit root tests have been shown to have greater power than univariate tests. Results based on several different panel tests for stationarity and for unit roots, some of which allow for a structural break, are presented. These results suggest the U.S. cigarette industry was not as competitive as one might expect on the basis of univariate tests and the estimates reported by Adhikari (2004).