Productivity
The NBER's Program on Productivity, directed by Ernst R. Berndt, NBER and MIT, met in Cambridge on March 17. These papers were discussed:
|
Fixler and McClelland estimate one-year log price changes for 175 different items and find that, for all goods, growth rates in at least one Primary Sampling Unit (PSU) were significantly different from the average. For most items, at least one region (Northeast, South, West, or Midwest) had a significant influence on price. Fixler and McClelland then ask whether the size of the PSU drives the results; they find that, in almost every case, large "A-class" PSUs collectively have price growth rates significantly different from the mean. Furthermore, inside each region at least one PSU had significantly different growth rates for most goods. Finally, the authors consider the case of ready-to-eat breakfast cereal. Using CPI data, they find that the price changes for cereal across areas are significantly different from the mean price change. Then they perform the same experiment using CPI data for a major manufacturer of ready-to-eat cereals, and again they find the same result. Looking at the same manufacturer but using weekly scanner data collected in three New York PSUs, Fixler and McClelland find similar results to those from the CPI data for five one-week indexes but not for the weekly indexes.
Brynjolfsson and Smith analyze the characteristics of the Internet as a channel for two categories of homogeneous products: books and CDs. Comparing pricing behavior at 41 Internet and conventional retail outlets, the authors find that prices on the Internet are 9 to 16 percent lower than prices in conventional outlets, depending on whether taxes, shipping, and shopping costs are included in the price. In addition, they find that Internet retailers' price adjustments over time are up to 100 times smaller than conventional retailers' price adjustments, presumably reflecting lower menu costs in Internet channels. Brynjolfsson and Smith also find that levels of price dispersion depend on the measures employed. When they compare the prices posted by different Internet retailers, they find substantial dispersion: Internet retailer prices differ by an average of 33 percent for books and 25 percent for CDs. However, when they weight these prices by proxies for market share, the dispersion is actually lower in Internet channels than in conventional channels. This reflects the dominance of certain heavily branded retailers. Brynjolfsson and Smith conclude that, while there is less friction in many dimensions of Internet competition, it is branding, awareness, and trust that remain the important sources of heterogeneity among Internet retailers.
According to CPI estimates, the "sticker" or "list" price (tuition and fees) of a college education in the United States has risen significantly faster since the early 1980s than the overall rate of inflation. This has raised considerable concern among policymakers, parents, and students that college attendance is becoming less affordable while it is becoming increasingly important for economic success in the job market. Schwartz and Scafidi note that the government collects data for the CPI on the list price of college without adjusting for scholarships or other discounts. Further, it makes no adjustments for changes in the quality or characteristics of the services provided, such as attributes of the faculty, the course offerings, or the facilities. Thus, the estimated price indexes reflect changes in quality and characteristics of college as well as changes in price. Using data from the College Board's Annual Survey of Colleges, Schwartz and Scafidi develop a hedonic model of the price of one year at a U.S. four-year college and explore their estimated quality-adjusted price indexes.
One of the most important changes to the National Income and Product Accounts (NIPAs) in the recently released comprehensive benchmark revision is the recognition of business and government expenditures for computer software as investment. Previously, only software embedded in equipment by the producer of that equipment was counted as investment. Business expenditures for software were classified as inputs to production, and government expenditures for software were classified as government consumption expenditures. In their paper, Grimm and Parker describe the various types of software; provide an overview of the impacts of the recognition of software as investment in current dollars, prices, real software, and real GDP; discuss the effects of software on the definitions of NIPA components; list changes in NIPA tables attributable to the new treatment of software; and list some planned future improvements. Grimm and Parker describe the methodology they use in an appendix, which expands on the descriptions that appeared in the August and December issues of the Survey of Current Business.
Cutler and McClellan present some preliminary estimates of a price index for the treatment of breast cancer. Using data from the National Cancer Institute's SEER database linked to Medicare claims data for 1984-91, they show that treatment costs for breast cancer have increased over time, but that longevity has increased as well. On net, the increase in costs appears to have been worth it; that is, the price index for breast cancer treatment has fallen moderately over time. But these results are sensitive to trends in incidence: the incidence of breast cancer has increased over time. Thus, breast cancer treatment appears to be less successful on a population basis than on a case basis.
White shows that certain CPI basic class indexes averaged higher rates of inflation between 1990 and 1996 than alternate indexes based on a unit value approach or obtained by aggregating outlet-type price indexes. The market entry of discount outlets, with lower prices and apparently lower rates of price increases over time, along with unrepresentative sampling, seems to have contributed to the deviations that White describes. If 30 percent of the CPI basket is prone to such outlet effects, then a plausible estimate of outlet substitution bias for the Canadian CPI is between 0.1 and 0.15 percent per annum (assuming that quality differentials are negligible). Biases arising from an unrepresentative outlet sample range from 0.12 to 0.24 percent per annum for the All-Items CPI. These biases are approximately additive and result in an overall outlet substitution bias and unrepresentative outlet sample bias for the Canadian CPI in the range of 0.2 to 0.4 percent.