Economy

US says probe found problem in seasonal adjustment of GDP data

U.S. government statisticians have found evidence that efforts to adjust the country's measure of economic growth for seasonal fluctuations have not been fully successful.

The Bureau of Economic Analysis, the government agency that constructs gross domestic product data, said this week that a component-by-component investigation found evidence in quarterly GDP data over different time spans.

"We did find some evidence of residual seasonality both over the most recent 10-year period and over a 30-year period," Brent Moulton, associate director for National Economic Accounts at the BEA, told reporters.

The government adjusts economic data to remove fluctuations such as seasonal weather patterns and holidays that normally occur at roughly the same time and magnitude every year, to make the series easier to interpret and analyze.

But seasonal effects have lingered in some cases even after the data was seasonally adjusted.

Economists believe residual seasonality has been most prevalent in first-quarter GDP data, with growth underperforming in five of the last six years since the recovery started in mid-2009.

Lending credence to that theory, the government sharply revised up the 2015 first-quarter GDP growth estimate to a 2.0 percent annual rate from the previously reported 0.6 percent pace. But it also revised down the second-quarter GDP estimate for the same year to a 2.6 percent rate from 3.9 percent.

Moulton said the treatment of monthly data in the derivation of quarterly national income and growth estimates was a pervasive problem.

"There are two things that can happen here. One is a series may be tested for seasonality at monthly frequency and may not show any signs of seasonality," he said.

"What we found in our review is that in some cases those series did at quarterly frequency show significant seasonality, suggesting there is some seasonal factor that needs to be adjusted to at least a quarterly frequency."

Moulton said even if a series is adjusted at monthly frequency, which is the case with most of the source data for GDP, there may still be residual seasonality in some cases once the figures are combined to produce quarterly data.

The solution, he said, was to test all monthly GDP source data series at both monthly and quarterly frequencies.

Moulton said the BEA, part of the Department of Commerce, would work with other agencies to ensure that residual seasonality was removed from historical data by 2018. The BEA would also start publishing nonseasonally adjusted GDP and gross domestic income estimates in 2018.

"These estimates will allow users to isolate data revisions more distinctly from revisions to seasonal factors," he said.