A test and modification of zipf’s hypothesis for predicting interstate migration

The purpose of this paper is to evaluate the accuracy of Zipf’s hypothesis in estimating interstate migration streams in the United States, and then to determine whether the predictive power of Zipf’s hypothesis can be improved by adding additional variables. The three independent variables in Zipf’s hypothesis accounted for 57 percent of the variation in interstate migration streams in 1935–1940, 61 percent in 1949–1950, and 68 percent in 1955–1960. The addition of per capita personal income of the states of origin and of destination increased the explained variance by only four percentage points in 1935–1940 and by less than one percentage point in 1949–1950 and 1955–1960. Then, the Petersen-Greenwood hypothesis that interstate migration streams in any period are a function of previous flows was tested by adding cumulative lifetime migration as an independent variable. Over 80 percent of the variation in interstate migration streams was explained by cumulative lifetime mobility, as reflected by state of birth data. The conclusion of the study is that the Petersen-Greenwood hypothesis provides a better estimate of interstate migration streams than does Zipf’s hypothesis. Nevertheless, the Petersen Greenwood hypothesis requires further evaluation to determine whether cumulative lifetime mobility is simply a proxy for some other underlying variable and whether it provides accurate estimates of migration streams for other geographic areas.