On Friday, BEA released the 2017 Annual Update of the National Income and Product Accounts. The U.S. national accounts data for the last three years, including GDP, were revised. At the level of total GDP, the revisions were small and the overall picture of economic activity isn’t a lot different from what we previously understood.
Here are some interesting nuggets of information to be found in the revised numbers:
- GDP growth during 2014 and 2015 was revised a little higher and during 2016, a little lower. Consumer spending grew more than in the previous estimates.
- Wage growth during 2016 was slower than we had previously thought, which was also reflected in downward revisions to gross domestic income (the income-side measure of GDP) and personal income. However, wage growth for the first quarter of 2017 was revised up.
- Because of the downward revisions to personal income and upward revision to consumer spending, the level of personal saving is lower than we previously thought. The revised personal saving rate for the first quarter of 2017 is 3.9%, which compares with the previously published estimate of 5.1%.
For the last 19 years, my job at BEA put me at the center of the annual update, so this July it was really different for me to be an outside observer. I’d like to spend the remainder of this post trying to explain why BEA’s annual updates are so important to maintaining the quality of the national accounts data.
I sometimes think of the various GDP estimates as lying on a continuum. At one end, quarterly GDP is a timely economic indicator that hopefully captures the broad ups and downs, the accelerations and decelerations of economic activity, albeit with imperfections and various gaps in coverage. This is what we see when the advance estimate is released roughly 4 weeks after the end of the quarter. At the other end of the continuum we’d like to use to compile a comprehensive set of accounts for all of the important sectors of the economy. For this goal, timeliness is less important than having and using the most comprehensive and detailed source data. The annual and five-year benchmark accounts, which are based on detailed and comprehensive source, are what you really want to use to fully describe the economy through an accounting system.
Here are some examples. For manufacturing, the annual GDP data are based on the Census Bureau’s annual survey of manufactures, which provides detailed data on the characteristics and activities of some 50,000 manufacturing establishments. In contrast, the quarterly GDP estimates for manufacturing are based on the monthly “M3” survey, which collects a limited set of data from large companies, with the data generally reported at the level of company divisions rather than establishments. The much more detailed and comprehensive annual data are needed to provide a valid benchmark for the quarterly estimates. Similarly, for business income, the source data for quarterly national accounts estimates are generally missing data from pass-through entities like partnerships and S corporations. Comprehensive business income data only becomes available when the IRS Statistics of Income Division publishes tax return data, which typically only becomes available with a lag of a year or more.
In addition to benchmarking, the annual update is important because the seasonal adjustments for recent years are updated. For the last two years, BEA’s GDP estimates have been criticized for residual seasonality—especially for relatively weak estimates for first quarter GDP growth—and BEA has committed to remove the residual seasonality as part of next year’s comprehensive revision. After Friday’s GDP release, Federal Reserve staff members Paul Lengermann, Norman Morin, Andrew Paciorek, Eugenio Pinto, and Claudia Sahm have published a note updating their analysis of residual seasonality. They argue that “residual seasonality is unlikely to be the primary reason for the slowdown in first-quarter growth this year.” I’ll also recommend some analysis by Claudia Sahm of the latest revised estimates of the first quarter of 2015 (which was the quarter that kicked off the discussion of residual seasonality), which is available as a set of Twitter posts beginning here.
Finally, the BEA has traditionally used the annual updates to introduce methodological improvements. Some of these methodology changes help to fill known gaps or weaknesses in the source data that are used for compiling the estimates, while others are undertaken in response to changes in the economy. This year, BEA took steps to remove a bias in personal consumption expenditures that arose because the previous methodology attempted to separate gasoline sales from other retail sales, but didn’t remove the gasoline sales made by warehouse and grocery stores. The 2016 Economic Report of the President (p. 74) noted that this omission may have caused consumer spending in 2015 to have been understated, but BEA has now taken steps to remove that bias. Other methodological improvements include the use of new source data for music and digital downloads, new data on e-commerce and scanner data on electronics, and use of a more representative price index for software.
1 Comment