No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Performance Measures for Computers
Chips can be easily copied and semiconductor firms are not monopolies. Nevertheless, in the semiconductor industry patents protect Ricardian rents against free riding. Ricardian rents—rents wrought by a firm’s differential ability to produce more output or value per unit of input—remunerate the investments in R&D that semiconductor firms make in the expectation of profit. In addition, patents enlarge the set of business models, strategies, and contracts that firms can use to trade. Many practices that emerged over time—for example technical marketing, second sourcing, licensing, trade in intellectual property—and the observed evolution of horizontal and vertical specialization would not have been feasible without patents. Last, patents and Ricardian rents in the semiconductor industry conciliate protracted investments in R&D with exceptionally fast growth of multifactor productivity and falling prices over almost 70 years.
The present study analyzes computer performance over the last century and a half. Three results stand out. First, there has been a phenomenal increase in computer power over the twentieth century. Depending upon the standard used, computer performance has improved since manual computing by a factor between 1.7 trillion and 76 trillion. Second, there was a major break in the trend around World War II. Third, this study develops estimates of the growth in computer power relying on performance rather than components; the price declines using performance-based measures are markedly larger than those reported in the official statistics.
Performance modeling can be used throughout the life of a computer system, from initial design, through implementation, configuration (and reconfiguration) and even tuning. Performance models are usually solved by numerical techniques, where possible, and by simulation, otherwise. This paper summarizes IBM's contributions to performance modeling and the solution of performance models.
This paper provides a justification for hedonic price indices and details the properties of hedonic price functions. The analysis is done in a market setting in which a finite number of goods, each defined by its characteristics, interact. We note that proper hedonic indices can be constructed from the same data currently used to construct matched model indices. Since the matched model index does not incorporate price changes for goods which exit, and the goods that exited tend to be those goods whose prices fall, the matched model index has a selection problem which biases it upwards. The hedonic index does not have this problem. We illustrate with a new study of price indices for PC's. The hedonic index shows steep price declines in every year. On average, the matched model indices indicate no price fall at all and one commonly used matched model index is negatively correlated with the hedonic. We also construct and compare alternative price indices used either in research or by the federal statistical agencies. Of these the one that seems to work well is a Pasche style hedonic. Its advantage is that since it does not require computation of the current period's hedonic function, it is easier to use when monthly timetables need to be met.
In the last 20 years, the expenditure share of prepackaged software in the national output has grown. The large number of characteristics in computer software make hedonic regression techniques impractical for purposes of controlling for quality changes. In this study, matched model price indices are constructed using monthly scanner data on prices and unit values for various prepackaged computer software titles and categories sold in Canada from January 1996 to June 2000. Quality differences are controlled for by applying the maximum overlap method. Results show that prices declined during the period studied at an average annual rate of 5.9%. JEL classification: C43, L86
Un indice de prix de logiciel à partir de données scanographiques. Au cours des derniers 20 ans, la portion des dépenses pour les logiciels dans la dépense nationale s’est accrue. Le vaste éventail de caractéristiques des logiciels rend l’utilisation des techniques de régression hédonique inutilisable à cause de leur incapacitéà normaliser les changements de qualité. Dans ce texte, on construit des indices de prix mensuels à partir de données scanographiques sur les prix et valeurs unitaires pour différentes familles de logiciels vendus au Canada entre janvier 1996 et juin 2000. Les différences de qualité sont contrôlées en utilisant la méthode du maximum de recoupement. Les résultats indiquent que les prix ont chuté au cours de la période étudiée à un rythme annuel de 5.9%
The computing performance expressed in millions of instructions per second (MIPS) or transactions per second (tps) is discussed as there are no accepted industry standards for measuring the performances of different computers. Various benchmark programs used with different architectures and based on most popular instruction sets are discussed and benchmark results are reported.
The development of the first electronic digital computers in the 1940s marked the beginning of a technological impulse that continues to have widespread effects on the economy and society. The beginnings of the computer reflect the postwar emergence of high technology firms characterized by the large percentage of revenues continually spent on research and development to create innovative products. During the last four decades, complex forces - including technological advances, international competition, and national policies on technology, trade and investment - have shaped the evolution of this new and distinctive type of industry. In this study, the author identifies and analyzes the origins of technologies important to the development of computers. He traces the roots of individual technologies to the specific research groups and programs responsible for major advances and analyzes the impact of these innovations on industrial competition.
The economic model of production and production cost is extended to encompass heterogeneous outputs. The normal goods model shows how input usage varies as the rate of output changes. The extended model concerns output per unit of output—the way costs change as more outputs (the characteristics of goods) are packaged per unit of goods.The extended characteristics-space production model is used to integrate a variety of engineering approaches to measuring technological change. All are shown to provide information about pieces of the production or cost surfaces for heterogeneous goods. The model is also applied to uses of hedonic functions for measuring technological change, and some of the theoretical and interpretative difficulties of these new approaches to measuring technical change are assessed.
Econometric models of the U.S. computer market have been developed to study the relationships between system price and hardware performance. Single measures of price/performance such as “Grosch's Law” are shown to be so oversimplified as to be meaningless. Multiple-regression models predicting system cost as a function of several hardware characteristics do, however, reveal a market dichotomy. On one hand there exists a stable, price predictable market for larger, general purpose computer systems. The other market is the developing one for small business computer systems, a market which is relatively unstable with low price predictability.
This paper analyses the international comparability of methods used to convert the value added of the office and computing machinery sector (OCM) into constant prices for nine OECD countries. It concludes that the variations which exist in the price indexes are largely due to differences in the methods countries use to capture changes in the quality of an industry's output over time. The paper evaluates the impact of these different OCM price indexes on the growth rate of labour productivity during the 1980s by conducting a sensitivity test where the US OCM index is substituted for each of the individual country indexes. This experiment causes the OCM labour productivity growth rate to change by over a factor of ten for several countries. This result suggests that international comparisons of labour productivity should not be made for the OCM sector using the official data, and that labour productivity comparisons of sectors OCM belongs to—non-electrical machinery and fabricated metal products and machinery—should be conducted cautiously, if at all.
This workshop on Technology Measurement has given me the opportunity to return to and critically review work that I had started 20 years ago. At that time I set out to generate a technological history of the digital computer industry[4]. The research was undertaken to develop and implement a methodology for following the evolution of a single technology—general purpose digital computers—in detail. The methodology was to trace the improvements in the technology to obtain a measure of “value” for new developments and to determine the “radicalness” of these developments.This methodology was developed to go beyond the existing research, which at that time consisted of authors' descriptions of what they had observed in the evolution of a technology. Such research consisted in part of authors giving their opinions on what they felt were key developments. Technological historians depended on opinions and expert judgements for determining “important” improvements or innovations and assessing their impact on the technology. The Functional/Structural Measurement that I proposed was intended to change all this by following the history of digital computers and using a precise technique which would flag all the innovations and then measure the value of each.
In this paper we construct a number of quality-adjusted price indexes for personal computers in the US marketplace over the 1989–1992 time period. We generalize earlier work by incorporating simultaneously the time, age, and vintage effects of computer models and then develop a corresponding specification test procedure. When data on new and surviving models are used in the estimation of hedonic price equations, a variety of quality-adjusted price indexes decline at about 30% per year, with a particularly large price drop occurring in 1992. We conclude that taking quality changes into account has an enormous impact on the time pattern of price indexes for PC's.
Microcomputers are and will be having a significant impact on large U.S. corporations. Expenditure for microcomputers can become a major component of an organization's information system (IS) budget, and the decision process regarding their acquisition is being taken seriously. A key player in the microcomputer class, is the workstation. This article models the relationship between workstation price and associated attributes based on (1) the hedonic hypothesis that goods are valued for their utility-bearing attributes or characteristics, and (2) the premise that differences in the prices of goods offered in the same market effects the purchasing decision.
Does Grosch's law, which postulated that the costs of computer systems increase at a rate equivalent to the square root of their power, still hold? The age of mini-, micro-, and supercomputers seems to have complicated the situation. When computers are grouped according to their size and power, Grosch's law seems to hold within each group, but not between different groups.
"A dissertation submitted to the Graduate Faculty in Economics ... " Thesis (Ph. D.) -- City University of New York, 1986. Includes bibliographical references (leaves 67-69).
Incluye índice Guía de referencia sobre los desarrollos de tecnología en hardware para computadoras personales, cubriendo componentes y periféricos.
This paper estimates price indexes for laptop personal computers using hedonic methods and data taken from PC Magazine technical reviews. We use benchmark test results to construct a measure of system performance that encapsulates factors that have previously gone unmeasured, such as the interactions between hardware components. The resulting hedonic function is parsimonious yet has good explanatory power. A second approach to performance measurement is developed using a set of technical proxies that are shown to closely approximate the benchmark test scores, and are thus nearly perfectly equivalent in terms of resulting price index estimates. While not as parsimonious as a single performance measure, these proxies have the advantage of not requiring direct performance testing, and could thus be applied to larger data sets. Laptops were found to have declined in quality-adjusted price at an average rate of 40% per year for the period 1990-1998.
In this article, I first estimate hedonic price equations for computer spreadsheet programs, and then use the analysis to empirically test whether network externalities exist in this industry. The study shows that consumers are willing to pay a significant premium for spreadsheets that are compatible with the Lotus platform and for spreadsheets that offer links to external databases, and a similar premium for spreadsheets that offer local area network externalities. Finally, the quality-adjusted (real) price of computer spreadsheets declined by approximately 15% per year from 1986 to 1991.
This paper compares the impact of ICT capital accumulation on output growth in Australia, Canada, Finland, France, Germany, Italy, Japan, the United Kingdom and the United States. Over the past two decades, ICT contributed between 0.2 and 0.5 percentage points per year to economic growth, depending on the country. During the second half of the 1990s, this contribution rose to 0.3 to 0.9 percentage points per year. Despite differences between countries, the United States has not been alone in benefitting from the positive effects of ICT capital investment on economic growth nor was the United States the sole country to experience an acceleration of these effects. ICT diffusion and ICT usage play a key role and depend on the right framework conditions, not necessarily on the existence of a large ICT producing sector. (Copyright: Elsevier)
Changing product quality poses a challenge for the computation of price indexes, in particular in technologically advanced industries. We assess the differences between traditional and quality-corrected indexes by computing hedonic and matched-model price indexes for personal computer database software. Our database covers the price development in Germany from 1986 to 1994. Quality-adjusted software prices decline by 7.4 percent according to our hedonic index. Surprisingly, a matchedmodel index based on linking the prices of directly comparable program versions decreases even faster than the hedonic index (9.3 percent). This unusual result is apparently caused by the simultaneous selling of old and new versions of a given software product. The estimation results also confirm the importance of network effects. Code compatibility, i.e. the capability of executing programs written for the dominant database product, yields a significant price premium. The ability to read and write data in the dominant spreadsheet format (file compatibility) is also associated with higher prices, but the price differential is much smaller than in the case of code compatibility.
Changing product quality, rapid technological progress, and short product life cycles make it difficult to compute indices that correctly reflect the true price movements of IT products. This paper assesses the differences between traditional and quality adjusted indices by providing results for PC hardware. Transaction price data for the period from 1985 to 1994 are used to construct price indices for personal computers in Germany. The use of hedonic methods allows to correct for quality changes. Quality-adjusted hardware prices decline by 34 percent on average per year according to our hedonic index. The estimation confirms the importance of quality adjustment since the decline of average prices at 7 percent p.a. is substantially lower, thus overestimating computer prices and inflation. When the hedonic index for German computer prices is compared to various indices calculated for the US market it seems that the decline of quality adjusted prices in Germany has been significantly higher than in the US. However if prices are converted to US$ the average decline is about 30 percent which is consistent with results for the US. Die Berechnung von Preisindizes zeigt für Deutschland ein ähnliches Bild wie in den USA. Die starken Qualitätsverbesserungen schlagen sich nicht in erhöhten Preisen nieder. Dadurch weisen Preisindizes, die keine Qualitätsverbesserungen berücksichtigen - wie in der amtlichen Statistik üblich - zu hohe Indexzahlen aus. Der Preisrückgang wird unterschätzt, d.h. die Inflation wird zu hoch ausgewiesen. Bei der Analyse der Preisentwicklung für PCs in Deutschland (bezogen auf den Direktmarkt) gingen die nominalen Preise 1985 bis 1994 um ca. 7% jährlich zurück, während der hedonische, qualitätskorrigierte Preisindex einen Rückgang von 34% jährlich auswies. Kontrolliert man für den Wechselkurs zum Dollar (der Dollar ging in dem Zeitraum von über 3 DM bis auf unter 1,50 DM zurück, und die Preise für die Komponenten wie Festplatten, Speicher, Prozessoren, usw., die aus Fernost oder USA importiert werden, werden in Dollar abgerechnet), so bleiben noch ein Rückgang von 1% p.a. des nominalen und ein Rückgang von 30% des hedonischen Indexes. Diese Werte werden durch vergleichbare Forschungsergebnisse aus den USA bestätigt . Der deutliche Unterschied zwischen nominaler und hedonischer Preisentwicklung bei Computern und bei Software läßt vermuten, daß auch bei anderen Gütern, die einen raschen technologischen Fortschritt aufweisen, der Index der amtlichen Statistik die Inflation zu hoch ausweist.
In this paper we focus on alternative procedures for calculating and interpreting quality-adjusted price indexes for microcomputers, based on a variety of estimated hedonic price equations. Our data set comprises an unbalanced panel for 1265 model observations from 1982 to 1988, and includes both list and discount prices. We develop and implement empirically a specification test for selecting preferable hedonic price equations, and consider in detail the alternative interpretations of dununy variable coefficients having time and age, vintage and age, and all of the time, age, and vintage dummy variables as regressors. We then calculate a variety of quality-adjusted price indexes; for the Divisja indexes we employ estimated hedonic price equations to predict prices of unobserved models (pre-entry and post-exit). Although our indexes show a modest amount of variation, we find that on average over the 1982-88 time period in the US, quality-adjusted real prices for microcomputers decline at about 28% per year.
This study constructs new hedonic price indexes for electronic computers covering the period 1951-84. Regressions are estimated for four data sets, two used in previous studies by G. Chow and E. Dulberger, and two new data sets used for the first time in this study. Coverage is limited to mainframes until the late 1970s, but includes both " super-mini" computers and personal computers in the 1980s. The end result is a price index that exhibits a 1951 index number, on a base 1984 = 100, of 147,692, implying an annual rate of price change over the 33 years of -19.8 percent. Price changes for personal computer (PC) processors during the 1982-86 period appear to have been similar to those for mainframe computers during the 1977-84 period, in the range of -20 to -25 percent per year. Evidence for PC peripheral equipment is limited to 1984-86 and indicates a faster rate of price decline than for processors, particularly if the increasing availability of "clones" is taken into account. The paper places considerable emphasis on problems of weighting price indexes for computers together with price indexes for other types of "Office, Computing, and Accounting Machinery" (OCA) and other types of producers' durable equipment (PDE). The methodology used to construct the implicit price deflators in the National Income and Product Accounts, with a fixed 1982 base year, leads to a significant downward bias in the implicit OCA and PDE deflators after 1982, and an upward bias prior to 1982. A particularly disturbing aspect of the present national accounts is a spurious rise in the implicit OCA deflator of 157 percent between 1957 and 1971, despite the fact that its computer component exhibits a price decline and its non-computer component increases by only 8 percent. The paper recommends adoption of a chain-linked Laspeyres index number for any price index aggregate that includes computers. A properly weighted PDE deflator, using our computer price index, declines relative to the official implicit PDE deflator by 0.74 percent per year during 1957-72 and 0.87 percent per year during 1972-84.
Sharp declines in semiconductor prices are largely responsible for observed declines in computer prices. Although communications equipment also has a large semiconductor content, communications equipment prices do not fall nearly as fast as computer prices. This paper partly resolves the puzzle-first noted by Flamm(1989)-by demonstrating that prices for chips used in communications equipment do not fall nearly as fast as prices for those chips used in computers, and those differences are large enough to potentially explain all of the output price differences.
This study estimates quality-adjusted price indexes for personal computers. Three separate hedonic models are estimated using data from 1,841 personal computers over the period 1984-91. In addition to the traditional linear model, a nonlinear model is developed and estimated. The nonlinear model is parsimonious in parameters, allows time-varying attribute prices, and can be estimated using a pooled data set. The results indicate that nominal quality-adjusted prices of mail-order firms declined at an average annual rate of 24.62 percent; quality-adjusted prices of major manufacturers declined at a slower rate.
The present study analyzes computer performance over the last century and a half. Three results stand out. First, there has been a phenomenal increase in computer power over the twentieth century. Performance in constant dollars or in terms of labor units has improved since 1900 by a factor in the order of 1 trillion to 5 trillion, which represent compound growth rates of over 30 percent per year for a century. Second, there were relatively small improvements in efficiency (perhaps a factor of ten) in the century before World War II. Around World War II, however, there was a substantial acceleration in productivity, and the growth in computer power from 1940 to 2001 has averaged 55 percent per year. Third, this study develops estimates of the growth in computer power relying on performance rather than on input-based measures typically used by official statistical agencies. The price declines using performance-based measures are markedly higher than those reported in the official statistics.
In this paper we examine Australian data on national and regional employment numbers, focusing in particular on whether there have been common national and regional changes in the volatility of employment. A subsidiary objective is to assess whether the results derived from traditional growth rate models are sustained when alternative filtering methods are used. In particular, we compare the results of the growth rate models with those obtained from Hodrick-Prescott models. Using frequency filtering methods in conjunction with autoregressive modeling, we show that there is considerable diversity in the regional pattern of change and that it would be wrong to suppose that results derived from the aggregate employment series are generally applicable across the regions. The results suggest that the so-called great moderation may have been less extensive than aggregate macro studies suggest.
Anyone concerned with understanding the past, present, and future of high-speed computing must of necessity focus on progress in the components field. He must formulate searching questions about the outlook of that technology and how it affects progress in high-speed computer development.
SYSmark ® 2002: An Overview of SYSmark 2002 Business Applications Performance Corporation
- Bapco
Bapco. 2002. "SYSmark ® 2002: An Overview of SYSmark 2002 Business Applications
Performance Corporation." Available http://www.bapco.com/SYSmark2002Methodology.pdf,
accessed February 19, 2003.