June 10, 2015

A lot of work goes into measuring a $17 trillion economy. And, at the Bureau of Economic Analysis it’s a process that never really stops.

In addition to calculating Gross Domestic Product, a key economic indicator of how the U.S. economy is faring, BEA also produces thousands of related data points each month that flow from our GDP reports and give us rich detail about consumer spending, business investment and government activity.

The vast majority of data that feeds into BEA’s calculation of GDP is based on data collected by other sources – a mix of government agencies and some private entities. Just to give you a flavor, BEA’s data sources include 190 surveys or administrative data sources provided by 38 federal agencies as well as data from more than 100 private companies that help fill in some of the data gaps.

These source data represent a mix of frequencies – monthly, quarterly, annual, and in some cases, once every five years. Some of the data are seasonally adjusted by the source agencies and some aren’t. For some components of GDP, monthly or quarterly source data aren’t available, and BEA must extrapolate the estimates based on trends or on indirect indicators. (More information on BEA’s data sources can be found here.)

Against that backdrop, how does BEA go about seasonally adjusting GDP?

The approach that BEA uses is described as an “indirect” approach. All the pieces that make up GDP are first seasonally adjusted and then aggregated to arrive at a seasonally adjusted, topline GDP number.

Here’s a snapshot of how BEA’s indirect process works:

  • Detailed components of GDP are estimated from seasonally adjusted source data. Whenever possible, BEA uses source data have been seasonally adjusted by the source agency, in effect applying the same seasonal adjustments made by the source agencies. A majority of the source data for quarterly estimates of nominal GDP come from Census Bureau surveys that have been seasonally adjusted by Census.
  • In cases where the source data are not available on a seasonally adjusted basis from the source agency, BEA performs its own seasonal adjustment.
  • The components of GDP are then estimated from the seasonally adjusted source data. BEA reviews and checks the resulting estimates for seasonality.
  • The components are then aggregated to calculate GDP and other higher level aggregates. For values measured in current dollars, the aggregation is simple addition. Estimates of real (inflation adjusted) GDP are based on the chained Fisher quantity index formula.

The main rationale for using this indirect approach is that it allows users to decompose GDP and trace the estimates of the components back to the source data – without having to deal directly with not seasonally adjusted data. Having the ability to move easily from seasonally adjusted source data to GDP components to GDP itself is an important element of providing transparency to our users. Many users have told us that they find it valuable to maintain that consistency between the GDP estimates and the seasonally adjusted source data.

A potential drawback of this indirect approach is that it raises the risk of residual seasonality. That’s because seasonally adjusting the individual components might not necessarily remove all seasonality from the aggregates. As a result, BEA is working to improve its estimates of GDP by identifying and mitigating potential sources of residual seasonality.