Errors of Measurement in Output Deflators
NBER Working Paper No. 2000 (Also Reprint No. r1165)
In this paper we investigate the incidence of measurement errors in two independent estimates of long-term price change, within the framework of "multiple indicators" models of price measurement. We develop estimates of the measurement-error variances associated with both the Producer Price Index (PPI) and the Census Unit Value Relative (UVR) . Our estimates provide support for the generally accepted view that the PPI is a far more reliable indicator of long-term price change: the estimated signal-to-noise ratios for the PPI and UVR are 2.72 and 0.53, respectively. Our estimates should be useful for both constructing an optimal indicator of price change, and for identifying econometric models including error-ridden price- or output-growth terms as regressors. Our analysis suggests that "scores" assigned to product deflators provide useful information about their reliability. By extending our model to explicitly incorporate product-quality change, we are able to assess the importance of the problem posed by quality change for price and productivity measurement. Less than half of quality change, which we estimate to occur at an average annual rate of 1.3 percent, appears to be adjusted for in the PPI. Consequently, estimates of productivity growth based on the PPI underestimate "quality-adjusted" productivity growth by an estimated 43 percent.
Published: Journal of Business & Economic Statistics, Vol. 7, No. 1, pp. 1-9,(January 1989).