Duration of residence and prospective migration: the evaluation of a stochastic model

Abstract

The objective of this paper is to evaluate the empirical accuracy of the Cornell mobility model. Migration is formulated as a stochastic process governed by non-stationary probabilities: during a given interval of time, an individual is presumed to undergo a risk of migrating that decreases as he continues to reside in the same community. The major hypothesis, then, is that a person's propensity to move declines as his duration of residence increases.A secondary hypothesis proposes that age interacts with this relationship. Longitudinal data (5,000 residential histories from the Netherlands system of population registers) were analyzed and translated into prospective probabilities that are age- and duration-specific.Both hypotheses were substantiated. Specifically, the relationship is negative, curvilinear, and varies significantly by age. To facilitate simulation analysis of the model, the relationships found in the data are summarized in a set of logarithmic prediction equations.The findings of this paper underscore the fundamental limitation of stationary probability models in portraying migration and suggest that the non-stationary alternative is a more accurate formulation. More generally, processes of change which bear only a formal resemblance to migration (for example, brand switching or attitudinal change) may be governed by a principle of cumulative stability too. The evidence warrants further inquiry into the applicability of the model to other social processes where inertialike factors operate.