Predicting Respondent Precision of Geographic Locations

• In Event: 189 - Section on Community and Urban Sociology Refereed Roundtable Session and Business Meeting
  In Refereed Roundtable 60min: Table 02. Conceptualizing Community

Abstract

Understanding the geography of social networks requires measuring not only social ties, but also the locations of the individuals involved; for many sorts of networks, surveys remain the tool of choice for obtaining such information. In a survey where respondents are asked to report the location of themselves or others, spatial uncertainty will undoubtedly factor in into the measurement (whether due to obfuscation by the researcher or imprecision on the part of the respondent), which in turn has important implications for the analyses of the researcher. Using data from the American Social Fabric Study, a spatially stratified egocentric network sample of respondents in the western United States, we fit OLS models predicting precision of both one's own place of primary residence (ego), as well as those of one's social ties (alters). As each individual is geocoded to a spatial polygon, we use two measurements of precision: (1) root mean square distance (RMSD) of sampled points within the polygon to the average point estimate, and (2) polygon area. We test variables related to ego demographics, alter demographics, as well as those related to the geography of the polygon. We find that geographical context of the region (e.g., population density and whether or not the polygon is in the United States) appears to be most important when predicting both measures of precision. We also find evidence of a combined privacy/vulnerability effect, whereby respondents with more education tend to give less precise self-locations, with this effect compounded when the respondent is non-white.

Authors

• Emily J Smith, University of California, Irvine

• Nicholas N Nagle, University of Tennessee, Knoxville