Quality bias in the measurement of prices: empirical evidence and macroeconomic implications for Spain

Technological progress, which goes hand in hand with economic development, gives rise to continuous and significant changes in the characteristics (and quality) of the goods and services produced by an economy, as well as to the appearance of new products. The statistical apparatus of each country needs to be able to detect these changes and incorporate them rapidly into the measurements of economic variables. However, notwithstanding the effort devoted, this incorporation is never complete, since the adaptation of statistics to new realities is a costly process requiring time and analysis. One of the problems generated by this insufficient adaptation is the biased measurement of price changes, since the statistical techniques applied are unable to distinguish between that part of the change in the price of a good that stems from a change in its characteristics and that part that represents a pure price change. This, in turn, has significant implications for the measurement of real magnitudes, because if the price indices used to deflate nominal magnitudes are biased, then the real magnitudes obtained will be too. In recent years, this type of measurement problem has particularly affected economic sectors in which technical progress has been most rapid, such as information technologies. The traditional methods used by statistical offices to correct observed price changes for changes in the quality of products are often inadequate. As an alternative to these traditional methods, the literature and the experts of government statistical offices have, in some cases, been using the so-called hedonic method popularised by Zvi Griliches in the early sixties. The hedonic approach starts by assuming that the observed price of a good is a function of its characteristics. Accordingly, if the relationship between the price of the product and its characteristics (the hedonic function) is estimated, then an index can be constructed representing the path of prices for a constant level of quality. Countries that have adopted this methodology for productive sectors characterised by rapid technological progress, such as computers, have made significant corrections to the price indices for these products. Moreover, incorporation of these corrections into the deflators used to estimate macroeconomic magnitudes in real terms (such as investment and output) has entailed significant upward revisions to these variables. Given the obvious importance to the economic authorities of having the most accurate possible measurement of growth and inflation, the Banco de España decided to undertake a research project with the aim of approximating the size of the biases that might be operating in the measurement of real magnitudes in Spain, as a consequence of the use of traditional techniques to compile price indices. This project had two basic elements. First, hedonic price equations were estimated for some of the products identified as having experienced particularly significant changes in quality: computers, cars and housing. Electrical household appliances and the products of the other ICT (information and communication technologies) industries also incorporate significant changes in quality, but the absence of suitable data has made it impossible to make the necessary estimates. Second, a macroeconomic exercise was carried out in an attempt to gauge the importance of the resulting bias in the estimation of the main macroeconomic aggregates. This exercise was, in turn, tackled in two phases. First, to obtain a point of reference, the price corrections that had been estimated in other economies pioneering the use of these techniques were applied to Spanish National Accounts data for the period 1986-1994, and second (the phase presented in this article), the price corrections deriving from the sectoral analysis for Spain were applied to the most recent National Accounts data (1995-1999). The reasons for including computers among the three groups of products studied in this project are obvious, since the quality improvements in recent decades have been enormous. Processing power (associated with developments in the semiconductor industry) has doubled approximately every one and a half years and other computer components (hard disk, memory, monitor, graphics card, programmes, etc.) have also improved dramatically. As a result, today's computers are very different from those available a few years ago, even offering attributes that were not then available. Since a price index must compare the prices of identical products at two points in time, the adjustments made to take into account such blatant quality changes are particularly important in this sector and the hedonic methodology has been shown to be particularly suitable for this purpose. Thus, the hedonic price indices estimated for the computer sector in different countries show average annual price reductions of 20% to 40%, depending on the type of product and the period considered, while in those countries that use more traditional techniques the price falls recorded are very small and in some cases prices even increase. Eurostat recommends the use of hedonic methodology to calculate computer price indices and the statistical offices of at least the United States, Canada, Denmark, France and Sweden are applying them. In the case of cars, quality improvements have also been unquestionable, although less dramatic than in computers. The first studies of hedonic prices in this sector were carried out in 1939, when Court estimated that, over the period 1925-1935, the price of new cars had fallen by 55% in the United States, when official US Statistical Office figures showed an increase of 45%. In Spain, the adjustment of the prices of this sector for quality changes is important in macroeconomic terms, given its weight in the economy: in 1999, it accounted for more than 5% of Spanish GDP, 6.5% of total household consumption and more than 15% of exports. As regards housing, it should be noted that the US Bureau of Economic Analysis adopted hedonic techniques for the first time in 1963, precisely for the National Accounts deflator of new housing, in order to take into account the slow but constant improvements in the quality of construction. In all countries, residential construction has a considerable weight in investment and real output. Indeed, in Spain, gross fixed capital formation in residential construction accounted for 5.2% of GDP in 1999 and 21.1% of total gross fixed capital formation. Moreover, there is a growing consensus that the European harmonised index of consumer prices (HICP) should reflect owner-occupied housing by including an index of prices for new housing duly corrected for quality changes. This question is particularly important for Spain, since it has one of the highest percentages of owner-occupied housing (approximately 85%) in Europe. Finally, it should be noted that although the estimates of hedonic functions presented in this article for the prices of computers, cars and housing are all based on the hedonic method, they differ in important aspects, as they have had to be adapted to the particular characteristics of each product and the information available in each case. The article is structured as follows. Section 1 summarises the study carried out on computer prices, while Sections 3 and 4 summarise those on the prices of cars and housing, respectively. Section 5 then uses the results obtained for the biases due to changes in quality in these sectors to assess their impact on the measurement of Spanish macroeconomic variables over the period 1995-1999. Lastly, Section 6 presents some brief conclusions.