National Income Accounts
Measuring prices

International and historical comparisons of economic performance require measures of unemployment, inflation and the growth of real output per capita (or per employee per hour). The measurement of the latter two requires an accurate estimate of the aggregate price level, whose measurement is therefore one of the most important tasks of national accounting. In Discussion Paper No. 663, Research Fellow Robert Gordon reviews the extensive US literature on price measurement and points to numerous areas in which techniques of price measurement may affect economic performance by leading policy-makers astray.

In the short run, over-estimates of inflation can lead to the adoption of unnecessary restrictive policies: in 1980, an erroneous treatment of the cost of home ownership panicked the Carter Administration into implementing ill-advised controls on consumer credit; while the inclusion of mortgage interest rates in the best-known inflation rate in the UK can lead to a vicious circle of inflation and restrictive monetary policy. In the long run, measurement errors can lead to incorrect comparisons of output growth and productivity from one decade to the next and internationally. For instance, the US price index for computers entails rapid reductions in the price level in a major manufacturing sector and a corresponding increase in the quantity index. The equivalent European indices do not accurately reflect the rapid improvement in the performance of computers and other high-tech goods, so European measures of output and productivity growth in manufacturing are understated relative to the US. In the services sector, in contrast, European price measures often do better than their US counterparts; for example, they use physical measures of activity in banking and financial services (such as cheques cleared, shares traded and policies issued) rather than labour input (which implies a zero productivity gain by definition).

Gordon notes that the `hedonic regression' technique for the measurement of quality change may be appropriate only for products for which quality differences across models can be captured by just a few characteristics. New research that develops historical, quality-corrected price indices from mail- order catalogues and consumer-testing publications may provide more reliable results than either traditional or hedonic indices. A new problem is the `outlet substitution bias' that arises from the US statistics' treatment of identical goods sold by traditional merchants and discount stores as distinct products (because the latter provide a lower level of service). The effects of the recent massive shift from traditional, self- service outlets to newer and more efficient discount chains, outlet malls and warehouse clubs are not reflected in the CPI; similar problems may arise in Europe and Japan as modern distributional channels proliferate.

Measuring the Aggregate Price Level: Implications for Economic Performance and Policy
Robert J Gordon

Discussion Paper No. 663, June 1992 (IM)