Appendix
Geographical and temporal scope:
This analysis explores trends and patterns in gentrification from 1980 to 2000 using data from 3 United States Decennial Censuses at the census tract level in New York, Boston, Philadelphia, Atlanta, Chicago, Dallas-Fort Worth, Lost Angeles and Seattle. All tracts within counties that intersect the urbanized area of these 8 cities were included in this analysis, though not all tracts are mapped. Tracts that reported no data for a particular variable at a point in time, as well as those tracts whose data suggested error were excluded from this analysis (see Appendix 1 for more information about tract deletions and the data management process). These 8 cities were chosen based on a combination of their total area and their geographical distribution.
Interpolation process:
Data for the three census years were interpolated using Schroeder’s cascading density weighting technique. See Schroeder, Jonathan P. 2009. “Visualizing Patterns in Long Term Urban Population Trends.” Ph.D. Thesis. University of Minnesota.
Tract deletions:
After interpolating the data from the 1980 and 1990 census to make it compatible with 2000 census tract boundaries, the values for some variables in some cases seemed outside a reasonable range and distorted the analysis. For example, there were instances where the percent of the population self-identifying as black was greater than the actual population of the tract. In order to make the results of the analysis more useful, tracts with a percent greater than 200% were excluded from consideration because there was a good chance that these values were inaccurate (as a result of data processesing). The values of cases whose percentage was greater than 100 but less than 200 were reset to 100 because they were relatively close to that value and their inclusion in the study would be helpful. The following is a more detailed summary of tract deletions: Sorted by percent black in 1980: (6 tracts deleted, 16 tracts reset to 1.00), by percent black in 2000 (all tracts in reasonable range), by percent black in 1990 (5 tracts deleted, 11 tracts reset to 1.00). Sorted by percent with a bachelors degree or more in 198o (56 tracts deleted, 13 tracts reset to 13), by percent with a bachelors degree or more in 2000 (all tracts in reasonable range), by percent with a bachelors degree or more (1 tract deleted, 2 tracts rest to 1.00).
The K-means function in SPSS removed 261 tracts from the cluster analysis, most likely because they were unpopulated or reported no data.
Definition of variables:
Standardized Change in % Black: Zscore of (total black population / total population)
Standardized Change in % College Educated: Zscore of (total people with bachelors or higher degree / population 25 years and older)
Standardized Change in Standardized Per Capita Income: Zscore of ((Zscore of Per Capita Income in 2000) – (Zscore of Per Capita Income in 1980)).
Map projections:
All maps were projected using the state plane coordination system most appropriate for the location with the central meridian and parallels adjusted to suit the mapped extent.