The first-order Markov chain model is frequently used to project the future size distribution of firms in a particular industry. This model, however, is based on the rather restrictive assumption that the transition probabilities remain constant over time. This assumption was found to be inappropriate in the case of plants manufacturing frozen milk products in Pennsylvania during the period 1944–1963 and, if adopted, could lead to erroneous results as some actual predictions indicated. This article suggests a method, based on multiple regression techniques, of replacing the constant transition probabilities with probabilities which are a function of various factors including structural characteristics in the industry. The results of this procedure compared quite favorably with those of the assumption of constant probabilities.
[1]
J. Wolfowitz,et al.
Introduction to the Theory of Statistics.
,
1951
.
[2]
L. Telser,et al.
The Demand for Branded Goods as Estimated from Consumer Panel Data
,
1962
.
[3]
E. R. Swanson,et al.
MARKOV CHAINS: BASIC CONCEPTS AND SUGGESTED USES IN AGRICULTURAL ECONOMICS
,
1962
.
[4]
L. Telser,et al.
Advertising and Cigarettes
,
1962
.
[5]
T. W. Anderson,et al.
Statistical Inference about Markov Chains
,
1957
.
[6]
John G. Kemeny,et al.
Finite Markov Chains.
,
1960
.
[7]
Emanuel Parzen,et al.
Stochastic Processes
,
1962
.