For several years now, economists have bitten their nails as the government published disappointing first-quarter estimates of the nation’s gross domestic product. But observers have come to suspect that the numbers, like the boy who cried wolf, are sounding a false alarm.
At issue: the obscure and difficult process of adjusting numbers to remove the effects of normal seasonal variation.
This year, the first estimate of GDP growth in the first quarter was 0.5%. By the time the third estimate was published on Tuesday, the number had more than doubled, to 1.1%.
As the broadest measure of how the U.S. economy is growing, GDP influences both government and business decisions. Congress and the White House consider it as they write the federal budget; the Federal Reserve looks to it as it weighs monetary policy; and businesses take it into account as they decide whether to ramp up production, hiring and investment.
When the numbers—estimates of the total value of finished goods and services produced in the U.S—are weak, it signals a slowing economy and encourages observers like these to move cautiously.
“People take the first-quarter GDP growth as a warning sign,” said Glenn Rudebusch, director of economic research at the Federal Reserve Bank of San Francisco.
But identifying meaningful change amid normal shifts caused by school schedules, holiday shopping, seasonal weather and the like is tricky because those predictable fluctuations obscure genuine market change.
To make sense of the numbers, GDP is seasonally adjusted to account for the economy’s regular ups and downs, so analysts can discern whether the underlying numbers are unusually high or low—signs of legitimate strength or weakness in the economy.
In the first quarter of the calendar year, GDP growth typically slows down. That’s expected on the heels of the holidays, and seasonal adjustments should account for the predictable slump. But within the past decade, the apparent slump in the adjusted GDP in the first quarter has been unrealistically uniform, a sign the adjustments haven’t fully accounted for seasonality.
“We expect a certain amount of randomness in any economic data,” said Brent Moulton, who oversees GDP and other national economic statistics for the Bureau of Economic Analysis. Given that expectation, he said, first-quarter growth has been more consistent than he would have anticipated.
When adjusted numbers continue to exhibit the influences of seasonal effects, statisticians refer to it as residual seasonality.
Although the GDP pattern has been noticed only recently, a study by the Mr. Rudebusch and his colleagues found the trend goes back half a century.
The BEA, which calculates GDP, has also dug into residual seasonality, and in June released a component-by-component analysis of 2,000 nominal data series included in GDP in an effort to figure out what’s going on.
The most pervasive problem, the agency found, occurs when seasonally adjusted monthly data are rolled up to quarterly values.
To calculate GDP, the BEA assembles data sets from multiple sources, including government agencies such as the Department of Defense and the Census Bureau as well as entities like the National Association of Realtors and the American Petroleum Institute.
Typically, the data have already been seasonally adjusted by the source agencies, although in some cases the BEA makes the adjustments, and in special cases the data are left unadjusted. If the source figures are provided in monthly increments, or some other division of time, the BEA converts them to quarterly data and then aggregates all the components to produce the top-line GDP figure.
The problem is that the adjusted monthly components may retain a hair of insignificant seasonality. But once the numbers are aggregated to create quarterly figures and then combined with hundreds of other components, the GDP disgorges a noticeable hairball of residual seasonality.
“Small patterns of seasonality at the individual granular level, which don’t appear to be that significant, can add up over time, over quarter and over various components to substantial residual seasonality,” said Mr. Rudebusch, who found that adjusting the GDP a second time seems to erase the effect.
The BEA plans to stick with its basic approach, but, among other remedies, it plans to test the GDP’s monthly components for residual seasonality after they have been aggregated into quarterly data.
“We’re pretty confident that by doing this additional work we can correct individual components where there is a problem,” Mr. Moulton said.
By July 2018, the BEA hopes, the residual seasonality, along with any lingering doubt about the numbers, will have been eliminated.
Write to Jo Craven McGinty at [email protected]