Unexpected life events as improbable reasons for moving home: a review of two studies' findings

Abstract

In longitudinal studies of panel data, logistic regression coefficients' standard errors may be sensitive to extremely-skewed data from low response probabilities, and so, statistical findings of those studies may have been compromised. The present study reviews the findings of two published studies as examples of panel-data studies in order to be convinced about residents' moving home during the next one or two years in response to the occurrence of a life event. The present study's conclusion however is that four life events of household or family dissolution, becoming a widow(er), becoming unemployed, and finding employment/changing job, may have been traumatic reasons for moving at most, but not that they were unexpected reasons for moving. This conclusion is based upon a statistical contradiction between predicted likelihoods of moving from calculated probabilities for respondents' subgroups, and inferred probabilities from logistic regression coefficients. Methods for more refined calculation of standard errors of logistic regression coefficients for data with low response probabilities are proposed.