Latest working paper, Ethnic neighborhood maps
We introduce a new statistical definition of an immigrant ethnic neighborhood based on a choice model and using the location distribution of natives as a benchmark. We then examine the characteristics of ethnic neighborhoods in the United States using decadal census tract data from 1970-2010. We find that ethnic neighborhoods are pervasive, often capturing more than 50% of the ethnic population in a city, and differ significantly in housing and demographic characteristics from other locations in the city where a group lives. Most neighborhoods disappear within one or two decades. However, larger neighborhoods persist longer and have a well-defined spatial structure with negative population gradients. Neighborhoods grow primarily through spatial expansion into adjacent locations and lagged measures of the housing stock from previous decades can predict into which specific locations a neighborhood grows.
Our data on population counts by birth country come from the long form of the 1970, 1980, 1990, and 2000 decennial censuses. We use the averages from the 2006-2010 American Community Surveys for 2010 data. It is important to emphasize that these counts are summary level data at the census tract level, rather than complete population counts, and are thus only estimates of the true tract populations. Further, we have converted the original tract-level data for each year to constant boundary Census tracts, using the 2010 boundaries, in order to show population changes over time for consistent spatial definitions. To do so, we use the census boundary crosswalks provided by the S4 Institute at Brown University, and discussed in Logan, Xu, and Stults, The Professional Geographer, 2014. This spatial conversion process introduces additional error into the population counts for any specific tract. Therefore, while the ethnic neighborhoods we identify are quite likely to represent areas of high concentration, the listed population counts are only representative of the true population sizes in any year.
In the mapping application, we restrict the cities (core based statistical areas, or CBSAs) and years plotted for each group to only those CBSAs that had at least 1000 people from a given group in one or more years, from 1970-2010. Our neighborhood definition assumes independent location choices for every two people from an ethnic population. This statistical definition is more sensitive to violations of this assumption, such as when a single large family moves into a house, when cities have small ethnic populations. We prefer to apply our methodology to only those cities with at least 1000 people from a given group in that year. However, we also show plots for smaller city populations so that users can see changes in those cities over time as the group population grows.