Printer Friendly

Improved deflation of purchases of computers.

Improved Deflation of Purchases of Computers

In the revised GNP estimates released in December 1985, new deflators for computers (processors and peripheral equipment) were used for several components--producers' durable equipment, exports, imports, and government purchases. The new deflators were constructed by BEA primarily from price indexes for computing equipment developed by IBM Corporation.1 This article discusses the selection by BEA--from among the alternatives presented by IBM--of the index most appropriate to the deflation of GNP, the construction of the new deflators, and their use in the revised estimates. It concludes with an evaluation of the new deflators.

1. See Rosanne Cole et al., "Quality-Adjusted Price Indexes for Computer Processors and Selected Peripheral Equipment,' SURVEY OF CURRENT BUSINESS 66 (January 1986): 41-50.

Selection of the index

The IBM work described in the January SURVEY OF CURRENT BUSINESS resulted in four quality-adjusted price indexes for each of four types of computing equipment--computer processors, disk drives, printers, and displays (terminals). One of the four indexes, the matched-model index, uses the conventional method for controlling for the effects of quality change. The other indexes all involve the use of hedonic methods and are referred to as the regression, the characteristics, and the composite indexes.

The matched-model index is formed from the changes in prices for identical models that are sold in adjacent years. This method is similar to that used by the Bureau of Labor Statistics for constructing the Producer Price Indexes, components of which are used by BEA for the deflation of most other types of equipments in GNP. The major strength of the matched-model method is that, in principle, all characteristics are held constant in measuring the price change for each model so that quality change is not erroneously included in the measured price change.2 Use of the method thus requires the assumption that any price change that coincides with new model introductions is equal to the average price change for matched models. When the assumption is not met, price indexes constructed using matched-model methods may be biased. Such bias will be small where prices are relatively stable or where the number of new models is relatively small. However, these conditions do not hold for computers. New models are introduced frequently, reflecting rapid changes in technology, and their introduction is frequently associated with major price reductions.

2. For a discussion of the treatment of quality change by the for alternative indexes, see Jack E. Triplett, "The Economic Interpretation of Hedonic Methods,' SURVEY 66 (January 1986): 36-40.

Hedonic methods, which are incorporated in the other three indexes, impute the prices for all models for all periods based on the characteristics of the models and their implicit prices. Thus, they overcome the weakness of the matched-model index by estimating the price change occasioned by the introduction of new models. On the other hand, a weakness of hedonic methods is that they hold constant only the most important and measurable characteristics in deriving a quality-adjusted price change. If a characteristic that is a significant factor in the prices of models is omitted, the price indexes from the hedonic method may be biased.

The differences among the alternative hedonic indexes relate to the way and the extent to which the hedonic function is used. The regression index is formed from the coefficients in the IBM regression equation on the dummy variables for year or for technology class by year. It does not make use of information on quantities; each model is given equal weight regardless of the number shipped. The characteristics price index is formed from the coefficients in the IBM regression equation on characteristics of the equipment, such as speed or capacity. The indexes are weighted by shipments of characteristics. The composite index uses both reported prices and, for models not sold in the base year, prices imputed from the regression equation. It is calculated as the average of the ratios of current-year prices to base-year prices for each model, weighted by the number of models shipped in the current year. The characteristics price index and the composite index are conceptually equivalent as long as the correct characteristics are used.3

3. See Triplett, "The Economic Interpretation of Hedonic Methods.'

Of the four indexes, the composite price index was chosen for the deflation of the GNP components because it combines the strengths of the matched-model and the hedonic methods, using the hedonic function to impute the base-year price for models not sold in the base year.

Construction of the deflators

A deflator for computers was constructed by combining the IBM composite indexes for computer processors, disk drives, printers, displays, and a regression index for IBM tape drives. The resulting deflator covered the period 1972-84. It was extended back to 1969 using information from other studies of computer prices. Prior to 1969, the deflator was held constant at the 1969 level; the effect on purchases in constant (1982) dollars of holding the index constant was small.

A second deflator--covering both computers and other office, computing, and accounting machinery--was constructed using the new deflator for computers and Producer Price Indexes for related machinery, weighted together using business purchases of the two types of equipment. This deflator was needed for some of the components of GNP for which separately identifiable data on computers were not available, as described in the next section. The deflators for computers and for office, computing, and accounting machinery are shown in table 1.

The preferred way to use the IBM indexes for deflation purposes would be to identify and deflate separately the purchases of each of the five types of computing equipment in each GNP component. This approach would recognize differences in composition of the equipment purchased by businesses, government, and foreigners. It was not used because data to implement it are not available. Consequently, the BEA deflator for computers was constructed by combining the indexes for the five types of equipment using shipments of domestic manufacturers as weights.4 This approach assumes that each type of purchaser acquired the same mix of the five types of equipment and that this mix is the same as that of shipments of domestic manufacturers.

4. Shipments data are available annually in Current Industrial Reports MA35R, Computers and Office and Accounting Machines, U.S. Department of Commerce, Bureau of the Census.

Use of the new deflators in the revised estimates

The new deflators for computers and for office, computing, and accounting machinery were used in the revised estimates of several components of GNP, as shown in table 2. This table also shows the published categories that include computers, the deflators used in the previously published estimates for each of the components, and the deflators used in the revised estimates. In addition, the table presents the major source data used to prepare the current-dollar estimates to which the deflators were applied. As indicated in the table, separately identifiable estimates of computers, or of a grouping of products closely related to computers, are available for recent periods for producers' durable equipment (PDE), exports, imports, and government purchases. Separately identifiable estimates for computers are not available for these components for earlier periods or for the change in business inventories and the personal consumption expenditures (PCE) components for any period.

For PDE, purchases of computers were deflated, beginning in 1969, with the new computer defaltor. For both exports and imports, data for computers were available aggregated with other types of office machines. Beginning in 1967, this aggregate--business and office machines, computers, etc.-- was deflated using the office, computing, and accounting machinery deflator. Purchases of computers by the Federal Government for defense and nondefense purposes were deflated separately, beginning in 1972, with the new deflator for computers, as were State and local government purchases of computers, beginning in 1977.

For the two other components of GNP that include computers--the change in business inventories and PCE--no use was made of the new deflators because separately identifiable estimates were not available.

Evaluation of the deflators

The use of the new deflators for computers represents a major improvement over the previously used conventions of assuming no price change for computers or of assuming the same price changes as for related products. Nevertheless, there are several problems with the information used to construct the IBM price indexes and with the construction of BEA's deflators from these indexes.

There are three major problems with the information used to construct the IBM price indexes: (1) Coyerage of the sample is limited to certain types of equipment and selected manufacturers; (2) list, rather than transaction, prices are included in the sample; and (3) the information on shipments is incomplete.

Coverage--The samples represent a substantial portion of computing equipment. For medium and large processors and for printers and general purpose displays, the samples are large and include a wide range of both domestically produced and imported models. In 1984, the covered types of equipment accounted for about 44 percent of the value of shipments of computng equipment.5 However, the samples exclude a number of other types of equipment, including small processors, personal computers (PC's), special-purpose computers --such as those in airplanes, small and double density disk drives, punched card equipment, optical scanning devices, and plotters. In 1984, small processors, PC's, and small disk drives together accounted for about 30 percent of the value of shipments of computing equipment.

5. See the July 1985 issue of Current Industrial Reports MA35R.

The sample for processors is further limited to models sold by IBM and three manufacturers of plug-compatible equipment because of the difficulty of obtaining a consistent measure of speed over a wider number of manufacturers. Although only 67 models are included in the IBM sample, it represents a large proportion of total processor shipments, according to data derived from International Data Corporation censuses.

The impact of this limitation was evaluated by comparing the IBM regression results with comparable results from a BEA sample of 187 models produced by 17 manufacturers.6 The BEA sample was constructed from sources similar to those used by IBM, but differs from the IBM sample in two ways in addition to its inclusion of more models and more manufacturers. First, the BEA sample does not contain technology variables because of the difficulty of developing such measures consistently for all models. Second, the measure of speed in the BEA sample may not be comparable among all manufacturers.7 For 1972-83 (the BEA sample did not include 1984), BEA applied the IBM regression specifications, to the extent possible, to its own sample for processors. As shown in table 3, despite the possible noncomparability of the measure of speed across manufacturers and the lack of technology measures in the BEA sample, the coefficients for speed and capacity are similar to those obtained by IBM.8 This comparison suggests that the models in the IBM sample for processors are representative of models not included in the sample.

6. Improved Deflation of Computers in the Gross National Product of the United States. Working Paper Series WP-4 (Washington, DC: U.S. Department of Commerce, December 1985).

7. For a dismission of measures of processor speed, see Cole et al., "Quality-Adjusted Price Indexes,' pp. 41-42.

8. Similar results were also obtained from single-year and paired-year equations, indicating that the regression results are not unduly sensitive to the length of time period spoecified for the equitations.

Of the other coverage problems of the IBM sample, two are worth noting. First, the IBM sample for disk drives is limited to large and intermediate drives; these drives accounted for less than 20 percent of total disk drives in 1984. Second, omission of PC's from the IBM sample probably biases the price indexes in recent years. Preliminary research by IBM indicates that the overall impact of the omission during this period may be slight because the prices of both PC's and large and medium processors appear to have declined by about the same amount. However, this research also indicates that, in 1983 and 1984, the prices of PC's declined more rapidly than prices of larger processors.

List prices.--Discounting is common in the computer industry, and the use of list, rather than transaction, prices in the sample may bias the measures of price change if discounts change over time. To the extent that prices of models with "nonbest' technologies were discounted more heavily than those with the "best' technology, discounting will be reflected in the IBM regression coefficients on technology class by year variables. Therefore, the IBM coefficients on characteristics for processors and disk drives may not be biased. However, the composite indexes may be biased because they are calculated, in part, directly from the list prices in the samples.

Shipments.--The shipments data used to construct the IBM indexes are incomplete. Shipments by model are the appropriate weight for price changes for the composite indexes. They are available only for processors and may contain errors because they were derived from data on the stocks of installed systems. The indexes for disk drives and for printers are weighted by shipments by class of equipment. The indexes for displays are unweighted.

The most important of the problems relating to the construction of BEA's deflators is that the five separate type-of-equipment price indexes were not used directly in each of the GNP components. Separate current-dollar estimates were not available, and it is likely that each purchaser acquired a different mix of computer equipment. Other problems are the lack of information for 1985 and the lack of quarterly information. For 1985, a preliminary annual deflator was estimated from midyear 1984 and 1985 prices published in Computerworld magazine; a revised deflator will be estimated by extending the IBM price indexes. For quarterly deflation, the annual deflators were interpolated. For 1986, quarterly deflators will be estimated using information on price changes and on the introduction of new equipment from trade publications.

Table: 1.--Implicit Price Deflators for Computers and for Business Purchases of Office, Computing, and Accounting Machinery

Table: 2.--Deflation of Annual Estimates of Computers in the National Income and Product Accounts

Table: 3.--Comparison of BEA and IBM Regression Results for Processors
COPYRIGHT 1986 U.S. Government Printing Office
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1986 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:new methods for deflating computer purchase costs; includes corrections to the estimates of purchases of computers
Author:Cartwright, David W.
Publication:Survey of Current Business
Date:Mar 1, 1986
Previous Article:The business situation.
Next Article:The cyclically adjusted federal budget and federal debt: revised and updated estimates.

Related Articles
Revised estimates of the national income and product accounts of the United States, 1929-85: an introduction.
The U.S. national income and product accounts: revised estimates.
GNP: an overview of source data and estimating methods.
Gross product by industry: comments on recent criticisms.
Deflators for purchases of computers in GNP: revised and extended estimates, 1983-88.
The U.S. national income and product accounts: revised estimates; annual 1986-88 and quarterly 1986:I - 1989:I.
Gross product in industry, 1977-88: a progress report on improving the estimates.
Mail-order technology.
A computer maintenance and replacement fund.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters