The statistics corner: the GDP revisions and understanding sluggish productivity growth

Business Economics, April, 1996 by Martin Fleming

From the perspective of early 1996, the recent revisions to both the National Income and Product Account (NIPA) methodology and the Gross Domestic Product (GDP) data have left the economics profession in a state of suspended animation. With the method of calculating GDP so fundamentally revised and every macroeconomic data point restated, every macro relationship must be reestimated. Surely this is a once-in-a-generation if not once-in-a-life-time change.

It is very unusual in this era of antigovernment rhetoric and emotion to have a government agency, BEA, being the driving force for change in the economy and the economics profession. One could certainly observe that, to date, BEA has been less than completely successful in persuading the profession of the need for change, motivating economists to anticipate change in data analysis and model construction, and in conveying the merits of the change to the journalists who cover economics. Barring some sort of yet untried marketing device, overcoming professional inertia among economists has surely got to be one of the most difficult assignments anyone can undertake.

NEED FOR IMPROVED DATA SOURCES

While creating change in organizations and professions is very difficult but very important, in the era in which we live, it is not the only problem BEA faces. While the new chain-weighted methodology is appropriate and useful for analyzing activity in today's global economy, it does rely on existing data sources. The revised methodology is, on balance, an improvement but it does not eliminate the need for increased expenditure for additional and improved data sources.

For instance, a substantial number of medium and large size corporations are making massive expenditures on corporate-wide, networked information systems for financial, marketing, manufacturing and human resources purposes. Such expenditures involve roughly equal doses of consulting, programming and systems development time, i.e., labor, and hardware and software purchases. The construction of such systems are just as surely capital expenditures as the purchase of a new piece of machinery or equipment was fifty years ago. Such expenditures are only partly captured in the GDP data, and the revisions do nothing to alter this omission.(1) In addition, there remain substantial shortcomings in the coverage and timeliness of the source data for expenditures by consumers for financial services and healthcare as well as expenditures by state and local governments. All of these source data problems have seriously hampered our ability to forecast future economic activity. (See Fleming, Jordan and Lang.)

Rectifying these source data problems is very expensive and, thus, presents much greater difficulty for BEA. By contrast, implementing the revised GDP methodology is much less expensive, although not free by any means. As a result, the revised methodology was implemented in conjunction with the scheduled base period revision from 1987 to 1992 in 1995.(2)

THE NEW METHODOLOGY

The case for the new methodology has been made clearly and has been available for some time for those who have the interest and the time to invest in gaining an understanding.(3) Without reviewing the details of the approach here, in essence, the methodology introduces two new concepts to GDP calculations.

First, the Fisher Ideal index is employed. This index is employed in response to the concern that buyers are substituting among commodities as a result of changing relative prices. Thus, in order to separate changes in spending that raise or lower standards of living from those that merely represent an alternative way of achieving the same standard of living, the old GDP calculation method was not appropriate. The Fisher Ideal index was selected because it has been shown to be among a class of index numbers that permit a very good approximation to an "exact" formula. It is also easy to compute and use. (See Triplett and Diewert.)

The chain-type annual-weighted quantity index is as follows:

(1) [Q.sub.1] = [square root of] [Sigma] [p.sub.b] [q.sub.t]/[Sigma]

[p.sub.b] [q.sub.b] [multiplied by] [Sigma] [p.sub.t] [q.sub.t]/[Sigma]

[p.sub.t] [q.sub.]

where

[Q.sub.1] = Fisher Ideal Quantity Index p = price q = quantity b = lagged base period t = current period

It is important to bear in mind that p represents the price of the good or service when expenditures are being used to measure GDP. However, in calculations by industry or sector the p represent value-added per unit, which is the difference between the unit price and the cost of materials or purchased services per unit.

Second, each period the growth of each individual quantity of goods and services is weighted by that quantity's current and lagged share of nominal GDP. With the share lagged by just one year, in the case of annual data, and a centered one-year lag, in the case of quarterly estimates, the base period is constantly rolling forward. (See Prakken and Guirl) The old method employed a fixed base period that was applied to all periods and was updated periodically.