Accurate Density-Functional Fluctuation Theory (DFFT) approach to forecasting ethnic composition of neighborhoods.

Accurate predictions of human residential dynamics are invaluable to developing housing, transportation, and social policy. Although large-scale forecasts can be made by estimating birth, death, and migration rates; predictions at the neighborhood level (on the scale of ~1000 people) remain a challenge due to (1) the inherent complexity of the underlying interactions and (2) the difficulty of inferring the interactions from available data. Here, we demonstrate the power of Density-Functional Fluctuation Theory (DFFT) to address challenges (1) and (2) to produce novel forecasts of neighborhood-level composition changes. DFFT works by forming an energy-like landscape composed of regional interactions and density-dependent social interactions. Vitally, since these quantities capture cumulative interaction effects, they require minimal assumptions of human behavior. And, surprisingly, they can be determined directly from fluctuations in widely-available demographic data. We demonstrate the efficacy of our approach by forecasting the dynamics of neighborhood ethinic composition from the year 2000 to 2010 using US census data.