| Abstract: | The stationarity of the beta distribution for 1926–1985 is rejected for the entire period as a single sample. However, results for pairs of five-year estimation periods are more consistent with stationarity. Over all possible pairs of five-year periods, stationarity in the pair-wise tests is rejected more often than expected merely by chance. However, excluding two aberrant periods, the number of rejections of stationarity is consistent with chance. Therefore, the distribution of betas is nearly stationary in both the short run and long run. For practical purposes the distribution of betas may be regarded as stationary. |
