Appendix 2 presents technical information on how the NSI was computed at the national and subnational levels.

The National Level

The NSI represents inflation in services other than rent/shelter or medical care. The BLS does not publish such an index—there is one for "services", for "services less medical care" and for "services less shelter" but not for "services less shelter less medical care." It was necessary, therefore, to derive an NSI; we appreciate the assistance we received in this regard from BLS economist Patrick Jackman, who computed the national NSI for this project. He did so by combining the "relative importance" and price changes in particular periods for "services less shelter" and "medical care" to derive "services less shelter less medical care."

For instance, using the "relative importance" for December 1977 and the inflation rates between December 1966 and December 1977, one can derive the "relative importance" for December 1966. This calculation was made for "services less rent" and "medical care," which allows a computation of the "relative importance" for their difference, net services. The growth in relative importance of net services provides the measure of net service inflation for the period December 1966 to December 1977. The same process was repeated for the 1977-82, 1982-86, and 1986-90 periods. The inflation rates of each period were chained together to obtain an index value for December 1966 and December 1990—the net services index rises from 100 to 503. This inflation rate is almost identical to that of services as a whole, whose equivalent value in 1991 (with December 1966=100) is 508.

Subnational Indices

Inflation rates can differ substantially across regions. Consequently, it was necessary to construct a net services index for each of the localities in which the study examined a school district. Regional indices were also constructed. The BLS, however, provides indices only for major urban areas and for certain size categories of cities within each region. The regional indices that correspond to the nine cities are: Baltimore for Anne Arundel; North Central C-size for Bettendorf; Denver for Boulder; South D-size for Clairborne; South C-size for East Baton Rouge; Boston for Fall River; Los Angeles for Los Angeles; New York City for Middletown; and Houston for Spring Branch.

There were several other constraints faced when constructing subnational indices. First, there are no indices for medical services and shelter for the period before 1977 for the areas outside of the large urban areas (including Denver). The indices for these areas are constructed using national trends for the pre-1977 period. Second, the only "relative importances" or "weights" available at the local level for 1977 were those from the CPI-U. In contrast, the national net services index used the CPI-W weights for the 1966-77 period (it was the only national index in existence during that time) and the CPI-U weights for the 1977-82 period. The local indices were constructed using the 1977 CPI-W weights for the 1967-77 and 1977-82 periods and 1982 CPI-U weights for the 1982-91 period. Third, the indices were constructed for the full years 1967 and 1991. Fourth, the most disaggregated level for which "relative importances" were available is region. Consequently, each locality's index is constructed using the relative importance of the appropriate region.The weighting method used for the local indices was applied to the national data as a check. It showed that the national net services index grew 5.33 percent more (when more appropriate weights were used and mid-points in the school year—December—were used). To correct for this bias, all of the local indices for 1991 were increased by 5.33 percent. This step increased their (log) annual inflation rate by 0.2 percent.The resulting local and regional net services indices were also compared to the local service indices. In all cases (except Denver and South C-size) the service index rose faster than the net service index, and most were within 2 percent of each other (except the Northeast, New York, Boston, South D-size, and Denver, which differed from 4 percent to 8 percent). These are not large differences over a 24-year period. Given the parallel trends of services and net services at the national and local levels, it might be easier for future research to simply rely on the service index.

[...Options for Deflating Education Expenditures]