A Model for Identifying Gentrified Areas With Census Data

Empirical research on gentrification suffers from a dichotomy between richly detailed neighborhood case studies and macro-scale, census-based analyses, perpetuating uncertainty over the extent and timing of gentrified areas in American cities. We develop a model relating tract-level census statistics to the results of a detailed field survey of 24 census tracts in Minneapolis-St. Paul. We use stepwise and canonical discriminant analysis to select nine variables distinguishing gentrified neighborhoods and to classify all central-city tracts for each decade between 1960 and 1990. Results indicate a moderate level of overall accuracy, and the model is more than 90% accurate in distinguishing areas of heavy reinvestment from stable, middle-class districts. Compared with other techniques, our approach more accurately distinguishes gentrification from other types of inner-city redevelopment, providing a useful tool for identifying the phenomenon with a measurable degree of precision.