I routinely get asked to consider crime data in land use models. Part of why I lean against doing that is because it's inherently difficult and nonlinear, even on longer time scales. Yes, it's something that firms and households look at in location decisions, but we don't have good models of how it changes over time in response to environmental factors and transportation policy. In the absence of those, we'd just be plugging in unreliable data from the base year and assuming it holds constant...