Purchasing Power Parity Measurement and Bias from Loose Item Specifications in Matched Samples: An Analytical Model and Empirical Study
Mick Silver and Saeed Heravi
This article considers the trade-off between two types of bias in the compilation of purchasing power parity (PPP) indices. The first is from the poor coverage of the items compared, or out-of-sample bias, when relatively tight specifications are used to define the items compared. The second is inappropriate quality comparisons, or in-sample bias, when loosely defined specifications are used to allow for greater coverage. An analytical framework is provided, which establishes the nature and extent of bias from poor coverage. Scanner data for three countries are used in this study to investigate the bias. Such data allow matching to be undertaken to different degrees of “tightness” of item specification and the resulting bias evaluated. Hedonic regression indices are argued to be a useful approach to dealing with out-of-sample bias. Such indices do not have to be based on matched specifications, but can extend to a representative sample of all prices, the differences in quality being “controlled” for by the regression as opposed to the matching and its restrictive effect on the sample. JEL classification: C43, C81, C82, E43, I32, and O47.
Purchasing power parity, index numbers, scanner data, hedonic regression