Index numbers and revealed preference rankings

For previously identified weakly separable blockings of goods and assets, we construct aggregates using four superlative index numbers, the Fisher, Sato-Vartia, Törnqvist and Walsh, two non-superlative indexes, the Laspeyres and Paasche and the atheoretical simple summation. We conduct several tests to examine how well each of these aggregates "fit" the data. These tests are how close the aggregates come to solving the revealed preference conditions for weak separability, how often each aggregate gets the direction of change correct and how well the aggregates mimic the preference ranking from revealed preference tests. We find that, as the number of goods and assets being aggregated increases, the problems with simple summation manifest.