ABSTRACT

Many spatial applications involve large data sets. For example, the US Census provides data on blocks (n = 8, 205, 582), block groups (n = 208, 790), census tracts (n = 65, 443), and other geographies. In the case of the Census data, each of these observations represents a region. If spatial dependence is material and each region affects every other region, this leads to n×n dependence relations. Since some applications involve elaborate models, computational aspects of spatial econometrics have been an active area of research for some time (Ord, 1975; Martin, 1993; Pace and Barry, 1997; Griffith, 2000; Smirnov and Anselin, 2001; LeSage and Pace, 2007).