A new method has been developed to measure directly changes in transit accessibility—the combined spatial effect of shifts in land use patterns and transit service—between metropolitan jobs and census tracts with high proportions of the people who most depend on good transit. Through focused analysis of transit routes serving one neighborhood in Buffalo and one neighborhood in Rochester, New York, two main questions are addressed. First, did transit-dependent poor people who lived in inner-city neighborhoods lose capacity to access jobs by transit during the 1990s? Second, if so, how much of the reduction in accessibility was due to changes in transit service rather than to dispersion of land use? Steps include formulating a gravity model using geographic information systems (GISs), calculating an accessibility index at two times during the 1990s at the census tract level, and disaggregating the accessibility change into subcomponents of change in land use and change in transit service by holding relevant variables constant to a base year. Findings do not support the a priori expectations: the transit component of change does not appear to contribute to a loss in accessibility from high-poverty neighborhoods. The model provides insights into the causes of accessibility change, the geographic distribution of accessibility change, and better assessments of whether transit agencies are successfully adapting to changes in land use.