Income Sorting : Measurement and Decomposition

Segregation of households on the dimension of income at the jurisdictional level is interesting to economists because, under some conditions, it is an equilibrium condition in the political economy models of jurisdiction choice that follow from Tiebout. This paper addresses the measurement of household income sorting across jurisdictions and the attribution of observed sorting to a pure Tiebout mechanism. A standard decomposition of income variance into within- and between-jurisdiction components is biased downward by roughly 50 percent due to measurement error and differences between transitory and permanent income. Adjusting US Census data accordingly, an average across US Metropolitan Areas (MSAs) of six to eight percent of income variation can be explained by differences across jurisdictions; approximately 21 percent in the much-studied Boston MSA. Variance decomposition overstates the role of locally provided public goods because jurisdictions are differentiated by both government and location. Comparing pairs of adjacent, randomly defined "neighborhoods" in the Boston MSA, I find that boundaries between physically adjacent jurisdictions explain no more than three to four percent of income variation