metropolitan area) in each of the dependent variables examined. These findings do not address the question of the impact of elasticity on central-city/suburban disparities.

The finding that more fragmented systems of government lead to lower per capita spending does not necessarily mean they are more efficient, since it is virtually impossible in these studies to separate out expenditure differences that reflect efficiency from those that reflect differences in the quantity or quality of services provided. Indeed, we would expect the demand for services to vary with different sizes of government over which demand is aggregated and with different spillover effects associated with different sizes of government.

The most commonly used operational measures is the number of local governments in a metropolitan area or number of local governments per capita. But it is not clear that this is a valid measure of the concept. Should all local governments be counted in fragmentation research? To the extent the research is concerned with coordination difficulties, then perhaps the most commonly used measures are reasonable: the greater the number of local governments (or local governments per capita), the greater the difficulties in arranging coordinated activity across the area. But, if the concern is the "sorting" consequences of fragmentation, as discussed above, perhaps only those local governments that have an important effect on the sorting process, i.e., those with local land use powers (primarily general-purpose units of local government rather than special districts) should be counted. However, it might well be argued that local school districts, although a special district and without land use powers, play such a significant role in the sorting process that they should be included as well.

Lewis' measure of fragmentation differs dramatically from previously utilized measures. He constructs a fragmentation index that is equal to TE(1-SSP), in which TE is total expenditures per capita in the metropolitan area and SSP is the sum of the squared percentages of total expenditures accounted for by each local government. The greater the number of governments, each with a lower share of total government expenditure, the greater will be the fragmentation index.

Bald (1994:297) reports on tax/service differences in 35 metropolitan areas between central cities and suburbs in the aggregate. On average in these 35 areas, central cities spent $1.51 per capita for every $1.00 per capita spent by the area's suburbs. (The difference was due to the much higher level of spending by cities on noneducational expenditures; suburban governments spent more per capita on education than did cities.) However, taxes as a percentage of family income (tax burden) were an average of 44 percent higher in the central cities, and Bahl notes (1994:297) that the tax burden disparity is increasing over time.

Oakland (1994:7-8) argues, however, that fiscal disparities are not necessarily undesirable and that efforts to reduce or eliminate them could have perverse efficiency consequences.

Pagano observes (this volume) that, on the basis of the Ladd and Yinger measures, the cities with the greatest actual revenue-raising capacity are in Ohio, which permits the most progressive earnings taxes on commuters; Cleveland's revenue-raising capacity was 41 percent higher than the average U.S. city and Dayton's 59 percent higher, both substantially above their standardized revenue-raising capacity.