Traduction à venir
Most economists know W. Erwin Diewert has published lots and know his main economic theory contributions, but fewer realize how much he has, and is, affecting the evolution of the official statistics of nations. Thus, on the occation of his selection as a Fellow of the CEA, I've chosen to provide a synoposis of his refereed journal contributions on economic measurement from 2000 through mid-2011. I selected 12 papers: 34% of his refereed journal articles for a period when he also published much else.
Price measures are central to all economic statistics, and Diewert has contributed in this area career-long, including recently. The usual approach to index number theory is based ratios. Diewert and Mizobuchi (2009) derive a difference counterpart, with wide applicability.
Diewert, Huwiler and Kohli (2009) note that the consumer price index (CPI) is usually computed as a Laspeyres price index, which causes substitution bias that could be largely eliminated with a Fisher index. Statistical agencies lack the current period expenditure data needed for that solution. However, that information does become available; just not quickly enough to meet CPI timeliness needs. They propose a retroactive Fisher CPI. They find that, using their retroactive index, the Swiss price level increased by about 5.2% over 1993-2000, rather than 6.1% as per the official CPI: not a trivial discrepancy.
Ivancic, Diewert and Fox (2011) have supermarket scanner data for over 8,000 products. Usually, in a time series context, the use of chained superlative indexes is recommended to compute a CPI when using scanner data. However, with price bouncing caused by temporary sales, chaining can cause chain drift. They show that multilateral index number methods provide drift free estimates of price change.
The user cost and rental equivalency approaches are two main ways of accounting for owner occupied housing (OOH) in the CPI, but both approaches have drawbacks. Diewert, Nakamura, and Nakamura (2009) explore a new oppportunity cost approach. They also address empirical methodology problems that statistical agencies face.
The International Comparison Programme (ICP) is a worldwide statistical initiative to enable output comparisons that are unaffected by inter-nation price level differences via the use of Purchasing Power Parities (PPPs). For a single product, a PPP is just the ratio of the price of the product in one country to the price of the same quantity and quality of the product in another country. New methodology was used in the 2005 ICP round. Each of six regions was each allowed to develop its own product list. The regions were then linked, using another separate product list and 18 countries across the six regions. Diewert (2010) developed the methodology used to accomplish the linking of the regions.
Diewert and Nakamura (2003) examine the productivity measures commonly used, and also introduce best practice performance measures of special relevance for inter-establishment comparisons. Empirical resuls for electric power generating plants are presented and discussed.
An inter-producer productivity comparisons method that Diewert was involved in developing in the 1980s is widely used but has important weaknesses. It was derived under unrealistic optimizing assumptions. Moreover, with increasing returns to scale, values for the (local) degree of returns to scale must be known or estimated. Diewert and Fox (2010) show how these problems can be overcome. Using a one equation econometric model consistent with their monopolistic competition framework, they demonstrate how total factor productivity growth (TFPG) can be decomposed into technical change and returns to scale components, even with increasing returns to scale. They then provide an empirical application for the US manufacturing sector for 1949-2000. The choice between value added versus gross output is also addressed.
Empirical support for the importance of increasing returns to scale is mixed. Diewert, Nakajima, A. Nakamura, E. Nakamura and M. Nakamura (2011) (DN4) note also that questions have been raised about published estimates due to the imposition of cross-industry restrictions, the use of value added output data, and the use of aggregate level data. DN4 use establishment-level panel data and a gross output variable, and do not pool over industries. They show results for 24 different 2-year panels for each of three important industries. They also define what they call a "Harberger" scale economies indicator which can be estimated in the same way as the usual returns to scale parameter, and interpreted similarly for certain policy purposes without imposition of assumptions about market conditions. Their empirical results on returns to scale in Japan are intuitive given available narrative evidence. Australias telecommunications provider, Telstra, enjoyed both productivity and profit growth over 1984-1994. Lawrence, Diewert and Fox (2006) develop a method that allows changes in a firms profits to be broken down into separate effects due to productivity change, price change and growth in firm size. This in turn enables calculation of the distribution of the benefits. The traditional model for the aggregation of capital services over vintages neglects induced obsolescence, and thus neglects the effects of increasing real wages. Diewert (2009) derives a vintage model for the measurement and aggregation of capital services using duality theory.
Huang and Diewert (2011) explain that R&D investments act as a mechanism for shifting outward societys production possibility frontier. Thus they treat R&D capital as a technology shifter instead of as an explicit input factor.
At the 2008 World Congress on National Accounts and Economic Performance Measures for Nations (http://www.indexmeasures.com/), Masahiro Kuroda delivered the following citation in naming Erwin Diewert and Dale Jorgenson as the 2008 Laureates: "Our two colleagues, W. Erwin Diewert and Dale W. Jorgenson, provided the seminal beginnings for much that has become so accepted in the fields of growth accounting and index numbers that we all now use these insights and methods no longer stopping to remember who created them." And Ivan Fellegi, Chief Statistician of Canada for 1985-2008, states that: "I and many of my colleagues at Statistics Canada consider Erwin to be one of the most profound thinkers currently active in economic statistics." Erwin Diewert has been a defining force in the evolution of economic measurement methods.
[Alice O. Nakamura, University of Alberta School of Business]