Peter Blau postulated that the demographic attributes of individuals span a multidimensional space and that their relationships are conditional on their coordinates in said space. For example, best friends are usually close to one another in the age dimension whereas romantic relationships are often different in the sex dimension. Attempts to map Blau space and understand which attributes are most predictive of the formation of social ties have made two assumptions. First, respondents are able to recall their immediate social network. Second, they are able to report their friends’ demographic profiles accurately. In this talk, I will discuss how aggregated social network data and census data can be used to address both assumptions. In particular, we have developed a hierarchical Bayesian model for social network data that have been aggregated geographically. I will discuss privacy concerns and the ecological fallacy.