Suddenly, The Economy Doesn't Measure Up

If you had to add up everything produced in the U.S. economy, from apples to zippers, and create an inflation-adjusted measure of national output, how would you do it? The government has been toting things up the same way since 1933, when economist Simon Kuznets devised the present system of national accounts, later winning a Nobel prize for his efforts.

Get ready for a big change. In December, the Commerce Dept.'s Bureau of Economic Analysis (BEA) will make structural changes in the way it computes and presents its flagship measure of economic performance--real gross domestic product. More than that, even without these modifications, the new numbers will have broad implications for the way we perceive the current and past health of the economy.

This is no run-of-the-mill revision. It will dramatically change the face of a familiar economic statistic. Moreover, the new numbers will say that economic growth in this expansion has been a lot weaker than we thought, in part because there has been less business investment. As a result, the productivity gains we thought we were getting will evaporate, dimming hopes for improvement in real wages, living standards, and U.S. competitiveness. All this will raise important questions for policymakers, especially at the Federal Reserve.

HARD TO FIGURE. To understand what the BEA is doing, consider its task. Real GDP is a weighted sum of 1,100 or so components, including detailed categories of consumer spending, business investment, government outlays, and foreign trade. In an attempt to correct a long-recognized flaw introduced by the current method of adding up the pieces, the BEA will now feature what it calls "chain-weighted" GDP instead of the traditional "fixed-weight" measure. The current measure in constant dollars will be shoved into the background. In its place will be an index, and the BEA will emphasize period-to-period percentage changes. For example, first-quarter real GDP, which now reads $5,470.1 billion in 1987 dollars, will appear as 118.3, meaning the economy has grown 18.3% since the base year of 1987.

It gets worse. Because each GDP component will be an index, you can no longer add its subcomponents to get a total. So traditional analyses of, say, a sector's changing share within the economy or of how much an individual sector contributes to overall growth, will be next to impossible. Want to compare the contribution of capital spending in this expansion to past upturns? Forget it.

To circumvent some of these problems, Commerce plans to provide information on broad sectoral contributions to growth, and it will supply dollar-equivalent measures of the new index. But guess what? The dollar figures won't add up to the total, either, and the further back in time you go, the worse the problem gets.

The issues go far beyond familiarity and convenience. For the current expansion, the economy's growth rate will fall to 2.6% annually from its estimated 3.1%--and this was already the slowest upturn in the postwar era (chart). The 2.6% rate is not much greater than the economy's long-term growth potential. Even without the new chain-weighted numbers, the old fixed-weight version of GDP, which the BEA also will update by moving its base year to 1992 from 1987, will yield similar results for recent trends in output.

The expansion's weaker look is partly the result of how the new GDP will treat key categories of business investment, especially computers. That sector has been the real power behind this expansion, accounting for 36% of the economy's growth. Under the new data, it will take a big hit. Of course, the capital-spending boom is still there. Growth in equipment outlays, up 13.4% annually in this expansion, will downshift to a not-too-shabby 11% pace, but the boost to GDP is less than we thought.

The new numbers so dramatically rewrite history that almost every existing interpretation of economic data will have to be reconsidered, especially because of what they mean for productivity growth. Measuring between business-cycle peaks, the old data showed that productivity climbed 1% a year in the three business cycles since 1973 and 2% so far in the current cycle. By contrast, the new estimates show slightly higher gains in the 1970s and 1980s of 1.3% a year vs. a mere 1.4% in this cycle (chart). This is a negligible uptick. "The new numbers show that the impact of computer technology on productivity has been exaggerated," says Maureen Haver of Haver Analytics, a New York economics consulting firm, and president of the National Association of Business Economists.

LITTLE CHANGE. How is this possible, in light of all the downsizings of recent years? One explanation may be that while many companies cut jobs by substituting new technology, high-tech industries have been on a hiring spree. And overall, companies have hired more workers than they have gotten rid of in this recovery.

Slower output growth also suggests slower corporate sales. Result: Many of the job cuts that experts perceived as lifting output per worker may have simply brought employment in line with this recovery's tepid sales growth. "Most people think that when a company lays off, it produces more with fewer workers," says Larry Mishel, chief economist at the Economic Policy Institute, a Washington think tank. "But many are just producing less, so their output per worker hasn't changed."

The new portrait of the economy also raises troubling questions about living standards and wages. Historically, pay and productivity growth have moved more or less in tandem. So it was no surprise that wages stagnated when efficiency gains slacked off after 1973. Economists have been more puzzled about the 1990s, when apparent productivity improvements have failed to translate into significant pay gains. The revised productivity numbers explain in part the poor performance of wages.

However, the data create another puzzle. Wages outpaced inflation by 1.2% in the two recoveries in the 1970s and by 1% in the 1983-90 upturn, but only by 0.7% in the current recovery. Yet productivity has improved at about the same pace as in prior decades. The implication is that the link between productivity and pay has weakened. It also suggests that the current profit boom--capital returns are at a 45-year high--is not being shared with workers.

CONFUSION. Less productivity growth is grim news on other fronts, too. Since Social Security is funded by a tax on wages, sluggish pay growth would leave the fund short of money when baby boomers retire. Some experts recently began to think the long-term outlook wasn't that gloomy, since the productivity revival eventually would lift pay. Now, the problem may turn out to be just as bad as everyone had feared.

The most important questions raised by all this, however, are the implications for policy, especially at the Federal Reserve. The Fed was well aware when it cut rates at its July 6-7 meeting that, based on current data, the new GDP would result in first-quarter economic growth of 1.7% instead of 2.7% and that second-quarter growth, to be reported on July 28 and expected to look weak, is likely to be revised downward as well. Looking longer-term, Fed Chairman Alan Greenspan has suggested that better productivity growth will help to stem inflationary pressures. But now, the new numbers seriously question just how much progress the U.S. has made toward improved efficiency.

So why create all this confusion? "Our objective is to provide the best possible measure of real GDP that we can," says J. Steven Landefeld, acting director of BEA. To do that, BEA must eliminate something called substitution bias, which is inherent in the current fixed-weight measure. Right now, the output of every good or service in real GDP is weighted by its price in a base year, currently 1987, and the pieces are added up to get total GDP in constant 1987 dollars. But in the real world, prices change, and consumers and businesses substitute the cheaper product.

Current GDP ignores this swapping. It inaccurately measures output because cheaper products are weighted with the old higher prices. And the distortion gets worse the further you get from the base year. However, the new chain-weighted GDP solves this problem, because the measure accounts for changes in relative prices during each year by chaining together price weights from adjacent years for every year that the GDP index is calculated.

The problem has been severest for goods whose prices change dramatically between base years, such as energy in the 1970s and computers and related equipment in the 1990s. Since 1987, quality-adjusted computer prices, for example, have dropped sharply, but such items are still weighted in GDP by their much higher 1987 prices. As a result, GDP has been grossly overstating the contribution of one of the economy's fastest-growing segments, especially in the most recent years. In this expansion, computers account for about 60% of the overstatement of GDP, says BEA.

PUZZLED LOOKS. This exaggeration can be minimized, but not cured, in the most recent years by shifting the base year forward, as the BEA does every five years. This updating decreases the importance of computers, for example, because of their new lower prices and weighting. But problems remain. GDP in the years prior to the base year tends to be understated. Moreover, the BEA had to face criticism every five years for rewriting history. So it will no longer feature this fixed-weight measure, and the agency will not include it in the morning news releases.

Economists still aren't sure what to make of it all. When BEA officials laid out their plans in front of about a hundred analysts at the Harvard Club of New York on June 22, there were more than a few puzzled looks and a lot of grumbling. In particular, because the pieces of GDP no longer sum to their total, the change is sending forecasting firms scurrying to revamp their big econometric models to make them compatible with the new data. "What we're demonstrating here is that growth isn't a number. It's one thing in 1987 dollars, and it's another with chain weights," says Edward McKelvey of Goldman, Sachs & Co.

Indeed, some economists don't believe the new stats give an accurate picture. Conceptually, everyone agrees that switching to chain-weighted GDP is the theoretically pure thing to do. But GDP will still be far from perfect. "It may take Commerce a few more years to catch on to the productivity miracle of America, but the gains are really happening," says Joseph Carson, chief economist at Dean Witter Reynolds Inc..

Carson and others argue that removing the substitution bias is fine, but that adjusting for price changes without an adjustment in quality changes, especially in telecommunications and services, ends up understating growth. Moreover, Corporate America's massive investment in computers has begun to pay off, say many economists, as companies replace thousands of workers with computerized technologies. This is particularly true of service output, which government surveys grossly undercount.

Some economists worry about how the public, especially investors, will greet the new GDP data, whose parts will no longer sum to the total. Making matters potentially worse, the BEA may fudge the dollar equivalents of the chain-weighted indexes to make everything add up. Erich Heinemann, chief economist for Ladenburg, Thalmann & Co., goes so far as to say: "I think Congress should hold hearings to assure the public that the Commerce Dept. will not create more problems than it will solve."

If Kuznets were alive today, he probably would applaud the new chain-weighted GDP as an improvement. But paradoxically, under the new method of adding up GDP, perceptions about this economy don't add up anymore.


-- The nation's wealth won't increase as rapidly

-- Wages and living standards will grow more slowly

-- The Federal Reserve Board won't be able to cut interest rates as easily


Before it's here, it's on the Bloomberg Terminal.