A New Test of the Administered Pricing Hypothesis with Canadian Data

This paper provides a new statistical look at the administered pricing hypothesis. Specifically, we use a time-series technique associated with the efficient-markets literature to study the dynamics of price behavior in concentrated and unconcentrated industries during different phases of post-World War II business fluctuations in Canada. To accomplish this, we compare, for the period 1948-75, time series of Canadian Industrial Selling Price Indexes (ISPI) from concentrated and less concentrated industries with a random-difference series used as a base.' We then test the hypothesis that the two series (concentrated vs. random difference; less concentrated vs. random difference) are obtained from different stochastic processes. The use of a random process as a proxy for business-cycle activity is supported by the recent empirical findings of J. H. McCulloch (1975), E. E. Anderson (1977), and N. E. Savin (1977), and the theoretical discussion of K. J. Arrow (1978). The empirical work shows that cycles in business activity in the United States are merely Monte Carlo In this paper we study the dynamics of price behavior in concentrated and unconcentrated industries during different phases of post-World War II business fluctuations in Canada. We use a nonparametric test to compare, for the period 1948-75, time series of Canadian Industrial Selling Price Indexes from concentrated and less concentrated industries with a series composed entirely of random changes as a base. The use of a random-difference series as a proxy for business-cycle activity is supported by theoretical and empirical work involving rational expectations and efficient information processing in output markets. The use of time-series techniques makes our results somewhat difficult to interpret, but generally they are not inconsistent with one version of the administered pricing hypothesis. *We thank W. M. Dunning and P. Stoochnoff, Prices Division, Statistics Canada, for help in assembling the time series of prices, P. Webb for programming assistance, M. Fortin and H. Gulesserian for research assistance, and the editors of this Journal for helpful comments. Financial assistance from the Canada Council and the Bureau of Competition Policy, Ottawa, is gratefully acknowledged. 1. In a random-difference series, the first differences of the series, and not the number individually, are random.