Through the analysis of data collected in Gwangju Metropolitan City in September 1997, it was demonstrated that regression analysis, analysis of variance, and cross-tabulation analysis yield different test results regarding the relationship between household income and the degree of patronage by retail store type. As an alternative, stochastic dominance analysis, which directly compares two cumulative probability distributions, was proposed. The results of existing research on this technique were reviewed, and among the developed stochastic dominance criteria, the appropriate criterion was applied to each case to empirically test the competing claims of Goldman (1975) and Hirschman (1978), who hold opposing positions on the relationship between income and patronage by store type. The finding that consumer expenditures for adult clothing were primarily made at department stores, and food expenditures primarily at neighborhood stores and markets, was consistent with Hirschman's predictions. However, in the case of adult clothing, consistent with Goldman's argument, the relative patronage rate was higher for low-income groups at discount stores and for high-income groups at department stores.