Gross Domestic Product: Understanding News from Noise

Multiple choice guess-tion: On an annualized basis, the nation's gross domestic product in the fourth quarter of 2001 went up: (a) 0.2 percent, (b) 1.4 percent, (c) 1.7 percent, (d) still to be determined. Answer: Yes.

Technically, all of the answers are correct or were correct at one point. The best answer right now is 1.7 percent—that's the most recent figure released in late March by the Bureau of Economic Analysis (BEA), the federal agency responsible for calculating growth of the nation's gross domestic product, or GDP. But the BEA makes three estimates for every economic quarter, and it just so happened that each estimate came up with a different number in the fourth quarter. Same thing happened in third quarter 2001.

In fact, GDP estimates are similarly restless over time. In a 1999 comprehensive review, the BEA remeasured 34 quarters of the economic expansion up to that point, and not a single quarter came in at the previously published figure. It also revised annual GDP from 1990 to 1998 an average of almost one-half of a percentage point. It came close to revising away the negative annual growth of 1991, shaving a -0.9 percent annual growth rate to just -0.2 percent.

Someone asleep at the national abacus? Not really. Much of the seeming fickleness is part of the inherent difficulty of getting a tape measure around the country's $10 trillion economy—which is more than twice any other single economy and one-third of the planet's GDP. In that context, subsequent revisions to GDP seem pretty minor—akin to hitting a bull's eye from a great distance, and then being disappointed that the arrow wasn't dead center.

GDP revisions come from two main sources. The first is the time-lag nature of data collection in such a large, complex economy, which mainly affects the three estimates (advance, preliminary and final) made for each quarter (which are released one, two and three months, respectively, after the end of a quarter).

The second major source of revisions—which mostly affect historical GDP figures, rather than recent estimates—is a change in the definition of what GDP actually counts or how it is counted. These definitional changes must then be applied to past GDP estimates so historical figures are comparable over time.

Though small in percentage, both types of revisions can produce changes in GDP equal to the output of not-so-small countries. For example, the cumulative increase of 1.5 percentage points from the first to last GDP estimate for the fourth quarter of last year (the result mostly of data lags) meant that the nation's measuring stick initially missed $33 billion in output—the equivalent of Morocco's entire annual production.

Brent Moulton, BEA associate director for national economic accounts, called the fourth quarter revision "fairly dramatic"compared with the average revisions of one-half a percentage point, but said GDP watchers "should be aware" that significant revisions can and do happen.

"If our typical revision were that large, it would be cause for concern," Moulton said, but revisions of such magnitude are "relatively rare," he said, estimating that only 10 percent of quarterly revisions are that large.

A + B = C(omplex)

The task of measuring GDP is both fairly simple and very complex. On the simple side, it is a basic, aggregate measure of the market value of new production in the United States. It measures only the value of final goods and services, and does not "double count" items as they are bought and sold among, say, manufacturers, distributors and retailers on their way to consumers. Nor does it count the resale of used merchandise. It counts only the final sale of new items.

GDP measures final purchases by summing consumption, investment and net exports of households, business and government. It considers only those goods produced in the United States, regardless of whether production is owned by foreign interests. Similarly, it does not count production by U.S. nationals abroad. (The precursor to GDP, gross national product, or GNP, did the opposite, excluding U.S. production by foreign-owned companies but counting production abroad by U.S. nationals. The shift to GDP was made in 1991.)

Even the building blocks of GDP—so-called national income and product accounts, or NIPAs—are fairly straightforward. One measures income, the other product, or output. NIPAs trace basic economic flows through the country's major industrial sectors and give GDP a useful cross-reference—income should equal the value of output.

Sounds simple, right? Well, it ain't. Peek behind GDP—if you dare—and things get complicated quickly. Not only is the measurement task itself Herculean, but quarterly and annual GDP estimates entail innumerable factors, many of which are subject to change or interpretation.

For example, GDP is itself something of a proxy measure and is several interpretive steps removed from what we're actually after, which is national output—the "new stuff"produced in a year: cars, apples, shoes, lawyer consults, restaurant meals and innumerable other final goods and services. But moving from "national output" to GDP requires a little mental gymnastics, particularly if the goal is to be able to compare one year's output with another.

One can't simply count the number of new things produced because cars and apples are fundamentally different economic units. So rather than a numerical count of items produced, we substitute the cumulative sales of all newly produced final items, which also establishes the relative importance of cars and apples to the nation's output. This value—what people actually paid for all final goods—is called current, or "nominal," gross domestic product.

But your math homework isn't quite finished. Such simple arithmetic does not actually tell us much. Most importantly, it does not allow us to compare one year's output with another because the value and quality of goods changes regularly, and new goods are constantly being introduced. To compensate, the BEA uses different tools to transform nominal GDP into something more useful.

For example, the agency uses what's called the "implicit GDP deflator." This measure adjusts the value of nominal GDP for price inflation and provides a better apple-to-apple comparison of GDP over time. This adjusted measure is called "real GDP," and is the figure typically reported by the media. (Unless otherwise noted, all subsequent references to GDP in this article will mean real GDP.)

That's merely the tip of the GDP iceberg. The BEA also uses sophisticated measuring techniques that adjust output for relative price changes over time—like the dramatic fall of computer prices—as well as changes in product quality. It is this complexity—both in the size of the economy and the methods employed to measure it—that gives birth to revisions in quarterly and annual GDP reports.

Anyone seen Indonesia?

As the fourth quarter of last year demonstrates, the advance, preliminary and final estimates of quarterly GDP can differ rather significantly. But most of the change stems from the lack of timely, comprehensive data. The BEA gathers GDP information mostly from government and private surveys, censuses and administrative records, but not all data sources are available on a timely basis. Rather than wait until all information is available—some of which is only available annually—the BEA fills data gaps by substituting judgmental estimates about the likely value of missing data until it receives "harder"information. Later, revised estimates often reveal more up-to-date information.

For example, the BEA said that much of the revision between the advance and final estimate of fourth quarter GDP last year was the result of "newly available" Census Bureau data on imports, exports, retail sales and manufacturers' shipments of complete aircraft, along with better data on federal spending from the Treasury Department. In other words, given the information the BEA had at the time—and no one had better—the three different estimates for fourth quarter 2001 were simply the best possible at the time.

Data lags even affect annual estimates. Every summer the BEA crunches GDP numbers for the previous three years. Last year's review downgraded 2000 GDP growth from 5 percent to 4.1 percent—equal to $90 billion, or Singapore's annual GDP—based mostly on lower investments in inventory and software, and lower personal consumption expenditures.

Because new and better information trickles in constantly from a variety of sources, quarterly and annual GDP estimates can see small changes years later. But most historical revisions are the result of definitional changes to the GDP formula—what "counts"and how it is counted—by the BEA to account for a changing economy and to take advantage of better theoretical models and improved measurement tools.

For example, in 1999 the BEA introduced nine definitional and classification changes, the biggest of which was a change in how software was counted. Previously, software purchases by business and government were considered "intermediate" inputs—in other words, they only contributed to the final good and were not the final good itself. The reclassification shifted software to a fixed investment, meaning it had residual value beyond the year of purchase that would be counted toward GDP (specifically, the BEA established depreciation schedules of three years for prepackaged software, and five years for both custom and own-account, or internally created, software).

This simple change had a significant effect on GDP, particularly in the 1990s when software became a bigger part of the economy. In a 1999 comprehensive review, the software change boosted GDP in 1998 alone by $150 billion (the GDP equivalent of Indonesia). Smaller definitional changes, along with $80 billion in statistical revisions meant that 1998 saw a one-year, upward revision of $250 billion (think combined GDP of Ireland, Egypt and Peru).

In fact, it's hard to say we ever have the "final" answer for almost any year. The 1999 study revised GDP back to 1959 to account for the definitional changes. GDP remained unchanged for only five years, all of them in the 1960s. Most years were revised up; only seven were revised down (all in the 1970s).

1.7 percent = picking up speed

So what, if anything, does a "moving GDP" mean for policy- and other decision-makers whose job it is to have their thumb on the economic pulse of the nation? Is it wise to put so much weight on a figure that is sure to change?

That depends on which estimate you pay the most attention to, and how much weight you put on it. Of the three estimates made each quarter, the advance estimate receives by far the most attention, the BEA's Moulton said, because such people are trying to get the earliest glimpse of new information on the economy, and the advance notice "is by far the most timely" of the estimates. But he acknowledged it is also "the least reliable by nature" because it depends more on data estimates than the subsequent preliminary and final estimates.

Mark Zandi, chief economist at Economy.com, said the private market normally does not pay all that much attention to GDP announcements, but he acknowledged that GDP estimates "do garner more attention" at economic turning points like the third and fourth quarter of last year.

When the advance report for the fourth quarter 2001 came in at a positive 0.2 percent, Zandi said there was a "sigh of relief ... [but] everyone took it with a grain of salt" because revisions in the third quarter saw the advance estimate of -0.3 percent drop to -1.3 percent by the final report. Sure enough, fourth-quarter estimates changed as well, but for the better, moving to 1.4 percent and finally to 1.7 percent.

And despite such revisions, research has shown repeatedly that quarterly GDP estimates—whether the advance, preliminary or final—are all useful indicators of the economy's general direction. A January 2002 BEA study on reliability found, as have previous studies, that the early estimates of GDP and their components "are reliable and present a useful picture of economic activity."

The report looked at final quarterly estimates from the first quarter of 1983 to the fourth quarter of 2000. They found that this estimate successfully indicated the direction of change 97 percent of time and GDP acceleration or deceleration about three-fourths of the time. It accurately predicted whether real GDP was high relative to trend about three-fourths of the time, and low relative to trend two-thirds of the time. It was also useful in finding cyclical peaks and troughs. As Moulton noted, advance estimates are generally less reliable than final estimates but are still useful barometers.

So, it seems, the real skill in reading GDP is being able to see general trends in the forest, rather than exact numbers in the trees. For instance, despite the fairly large revision from the advance to the final GDP report for the fourth quarter, the estimates "are telling us there was some growth in the last quarter, but less than average," Moulton said.

Couple of inches short of a yardstick

Whatever the final figure, some say that any tally of GDP is a poor national yardstick because it ignores vital parts of society. For example, GDP fails to measure any household output done without pay: cleaning clothes, preparing food and, particularly, raising children. GDP goes up, however, when we pay someone to do these things for us.

GDP also can rise because of a response to man-made events or natural disasters. For example, part of the surprise in fourth quarter 2001 numbers came from strong government spending, which grew at an annual rate of 10 percent, much of it the consequence of Sept. 11. Hurricanes, tornadoes and floods also "contribute" to GDP growth because they ruin the old stuff (buildings, cars) and people replace it with new stuff.

Moulton acknowledged that "sometimes there's some confusion" about the nature of GDP. Some people "misinterpret it as a measure of [social] well-being or welfare. ... It is specifically an output measure."

But the output model itself has its critics. For example, economists point out that GDP gives equal weight to expenditure regardless of its contribution to future prosperity—counting research and development spending as no different than expenditure for soda.

GDP, we love thee

Acknowledging the criticisms, it appears no one else has been able to build a better output yardstick than the BEA. A December 2000 paper to the Federal Reserve Board of Governors found that GDP revisions were highly predictable in the United Kingdom, Italy and Japan, and about half of the variability in revisions was due to data available at the time of the announcement—statistical noise that can be fixed, but isn't. For other G-7 countries, including the United States, "it seems that revisions primarily reflect news not available at the time"—inherent noise that has no volume-control knob.

So the BEA and others will continue to build better economic listening devices. But the current model was good enough that the Department of Commerce (which houses BEA) dubbed the development of the national income and product accounts—the component parts of GDP—the department's "achievement of the century."

Before NIPAs and GDP, presidents as recent as Franklin Roosevelt were forced to design policies to combat major economic upheaval like the Great Depression on such meager data as stock price indices, freight car loadings and other crumbs of industrial output. In the 1930s, economist Simon Kuznets was commissioned by Commerce to develop a set of national economic accounts. He presented the original set of accounts in a report to Congress in 1937, and Kuznets would later receive the Nobel Memorial Prize in economics for his work in this area.

What we recognize today as GDP was formalized in the 1940s, when World War II necessitated a better understanding of the nation's output capacity, and national income accounts developed by Kuznets were adapted and expanded to include national output (or product) measures. The rest is history (though it's being revised).

In testimonials on the achievement, renowned economists have gushed. Nobel Laureate James Tobin called GDP "the centerpiece of an elaborate and indispensable system of social accounting," and added that these innovations "deserve much credit for the improved performance of the economy in the second half of the century."

At a commemoration event, Federal Reserve Chairman Alan Greenspan said GDP "is a crucially important statistic to get a sense of where the overall economy is and where it has been." One of the "most significant ... and least heralded" contributions of the national accounts has forced structure and consistency onto an otherwise chaotic task.

"I'm aware that we've got a statistical discrepancy which creates all sorts of havoc, especially in the short run," Greenspan said. "But what is really quite extraordinary is how small that number is in a [$10] trillion economy."

Top