To date however, empirical findings on this topic have been inconsistent. One of the drawbacks of much of the literature is that most of it is cross-sectional - that is, it examines youth outcomes as a function of the neighborhood the youth lives in at the time the outcomes are measured. It is thus impossible to determine the direction of a causal relationship between neighborhoods and youth outcomes. The problem is especially pronounced because residential policy is one strategy that parents can and do use to further their children's chances, so it is likely to be related outcomes in systematic ways. Furthermore, mobility is selective of persons with particular characteristics, creating neighborhoods with distinct sets of characteristics - including children with particular sets of outcomes. For all these reasons, it is important to use longitudinal data to assess the impact of neighborhoods on children's development.

The Baltimore data are ideal for this purpose, because they contain information over a long time frame for a sample of at-risk youth. In the interview data there is a great deal of information about the youth and their parents over a long time period. My efforts have been concentrated on building a set of data on the youth's neighborhoods over time to append to these data. These efforts involved "geocoding" the data, that is obtaining addresses for all of the subjects, and ascertaining the corresponding census tract of residence. Then, for each census tract, we have drawn data from the 1970, 1980, and 1990 censuses to characterize these tracts. This work has proceeded in collaboration with Katrina McDonald at John Hopkins.

Currently, we are developing measures to summarize neighborhood qualities at each point in time. Once these are developed, we will use them in an analysis of youth outcomes as a function of personal and family characteristics, neighborhood characteristics, and residential mobility, over a 20 year time period.

-M.E. Hughes