Schumpeter 2000 THE ROLE OF INNOVATION AND QUALITY CHANGE IN JAPANESE ECONOMIC GROWTH Derek Bosworth, Silvia Massini and Masako Nakayama Manchester School of Management UMIST Schumpeter Conference, June 2000 Abstract. This paper explores the use of time series data to isolate quality change in the Japanese economy using a hedonic procedure. We argue that the traditional approaches to hedonic estimation based upon panel data sets of different brands in a given product area is extremely resource intensive and, thus, unlikely to be adopted by official statistical bodies outside of key areas, such as computers. This paper adopts a “top-down” approach to see whether more traditional measures of technical change, such as patents, can be used to separate pure inflation from quality change. If this is possible, it offers a much simpler route estimating the role of quality change in economic growth and performance. In practice, we extend the analysis not only to include patents, but other forms of intellectual property that might reflect technology and attribute changes, such as designs, utility models and trade marks. We begin by taking a longer-term historical perspective, exploring the development of indigenous inventive capacity in Japan during the early years when R&D data are not are not available. It is possible to show that the rise in utility models pre-dates the main growth in patenting activity, suggesting the development in more low-level indigenous creative work prior to higher level inventive activity. The principal aim of this paper, however, is to demonstrate that it is possible to develop robust models to explain changes in the producer price index in Japan, which can then be used to re-examine Japanese growth performance over the period from about 1960. If the official Japanese statistical body has fully accounted for quality change in the price indices (i.e. produced fully quality-constant price deflators), then the official estimates of growth will be correct. However, we provide strong evidence that this is not the case. Changes in quality, proxied by the IP variables, are important determinants of prices in Japan over the period 1960 to 1995 as a whole. Indeed, we provide evidence that the true rate of growth of the Japanese economy, taking into account the rate of quality change, is significantly higher than that suggested in official statistics. 1 1. Introduction This paper focuses on the role played by innovation and quality change as drivers of Japanese growth. Even using official data, the Japanese economic growth record stands out compared with other industrialised countries. However, we will attempt to demonstrate that the official statistics not only conceal an even more dramatic growth, but also the role of innovation and quality change in driving this performance. In order to do this, we develop a time series hedonic procedure that draws upon both standard and novel measures of innovative activity. Our interest in this work stems from an earlier project for the ONS, which not only indicated the deficiencies of standard methods of accounting for quality change in official statistics, but also the problems of operationalising hedonic procedures as the principal alternative approach (Bosworth, et al. 1993). In essence, the issue concerns the isolation of quality constant price deflators. In particular, if quality is constant, then real output growth is simply nominal output growth minus the rate of change in prices. If quality is changing over time, however, real output growth has both quality change and volume change components. Derivation of this real growth requires the change in nominal output to be divided by the purely inflationary component of price change. This pure inflation index is obtained by purging the price index of any effects caused by an increase (or decrease) in product quality. If the price index is not quality constant, the growth in real output reflects only the changes in volume and is biased downwards. The USA has led the way in the exploration and use of hedonic procedures for obtaining estimates of quality constant price indexes and exploring their implications for economic growth and productivity measurement. Gordon (1992), for example, argues that, “…it is likely that measurement issues bias downwards the growth rate of manufacturing output in every country, due to inadequate adjustments of price indexes for quality change, but more outside the US due to the absence of a computer price deflator.” Gordon has estimated that the bias for consumer durables may be around 1.5 per cent per annum and, for producer durables, about 3 per cent per annum. This suggests that measured increases in real output will be underestimated (as are real inputs, although the latter is more downward biased than the former). The results of work by Lichtenberg and Griliches (1989) suggests that, over a period of five years in the 1970s, the failure of official statistics to quality adjust the US PPI, lead to an underestimate of productivity growth of about one-third of its total. Nevertheless, the standard hedonic procedure is extremely resource intensive, both in the sense of the need to collect and collate technical characteristics/attribute data for each sector and in the need to periodically re-estimate the empirical specification. Then there is the issue of aggregation from the bottom up, in order to obtain economy-wide estimates. In other areas of work involving innovation, the literature has suggested the use of proxy measures such as R&D expenditure, patent counts, etc. (Schmookler, 1966). Thus, we present an alternative hedonic route, using time series data first suggested by Bosworth (1976). In the present work, however, we further extend this idea by not only utilising traditional intellectual property measures, such as patents, but other measures thought to reflect product innovation, such as trade marks, designs and utility models (Bosworth, et al. 2000). Assuming that we can demonstrate to the readers satisfaction that this approach works, then, given the relative ease of collecting these types of intellectual property statistics, it opens up the possibility of the much wider application of hedonic techniques. Section (2) explores the underlying issues in more depth, providing some evidence from the hedonics literature. Section (3) continues with an examination of the Japanese experience, both in terms of the intellectual property measures that we use as a proxy for various dimensions of innovation and the official measures of growth. Section (4) develops the hedonic specification and reports the results of controlling for quality in the main price deflator. Section (5) then discusses the implications for growth in the Japanese economy. Finally, Section (6) provides the main conclusions of the present paper.