During my 19 years at BEA, I felt that our most significant accomplishments were the four benchmark revisions or updates of the national accounts that were released in 1999, 2003, 2009, and 2013. On Friday, July 27, BEA will release its next benchmark update of the accounts.
What makes these data releases so important? There’s really two things going on. First, there’s the “benchmarking” itself. While users typically think of GDP data in terms of growth rates, the level of GDP also needs to be established. In the U.S. national accounts, this benchmarking of the level of GDP occurs at 5-year intervals coinciding with the economic census, when the most comprehensive and detailed data are available. Based on the economic census data, the BEA prepares benchmark input-output accounts that include industry-by-industry and commodity-by-commodity, estimates of the supply and use of all the goods and services produced during the year. The measurement of GDP depends on separating the goods and services that are “intermediate”—that is, used up in producing other goods and services—from the goods and services that are “final expenditures”—that is, used for final consumption, capital formation, or exports. This separation can be done most accurately when the comprehensive data from the economic census are utilized. This release includes benchmarks for 2012, the most recent year for which complete economic census data are available.
The second thing that happens during benchmark updates is that BEA makes important methodological improvements or conceptual updates to the accounts. An article in the April 2018 Survey of Current Business describes the major changes coming in this year’s update.
The methodological change that I expect will get the most attention will be improvements to seasonal adjustment. In addition to improved seasonal adjustments, BEA will re-introduce GDP estimates that are not seasonally adjusted (NSA). In 2015, after several years of weak GDP growth in the first quarter, residual seasonality in GDP became a hot issue, and since then BEA has done a lot of work to address the problem. (You can read about the problem in this note from the Federal Reserve staff, or in this article I co-authored for BEA during 2016.)
The revised GDP estimates will show improved seasonal adjustments, especially for the period since 2002, but also, in some cases, for earlier periods. BEA will also start publishing NSA estimates of GDP, gross domestic income (GDI), and their major components. (BEA used to publish NSA estimates of nominal GDP but in 2008 it had to eliminate them due to budget cuts. BEA hasn’t previously published NSA estimates of real GDP or of GDP price indexes.) Finally, BEA is making some changes to their seasonal adjustment processes to hopefully any recurrence of residual seasonality problems. They’ve improved their cooperation and communication on seasonal adjustment issues with the Census Bureau and the other federal agencies that provide seasonally adjusted source data that are used in compiling the national accounts. For example, when a Census indicator series is published at monthly frequency, staff from both agencies are now checking for signs of residual seasonality when the series is aggregated at quarterly frequency for the GDP estimates. Also, during the regular July updates, the revised series will be open to revision for longer spans in order to more fully incorporate the latest seasonal adjustments. BEA plans to routinely update the estimates for the most recent five-year span instead of the previous three-year span.
Another set of methodological improvements will center on price indexes for technology—specifically for software, medical equipment, and communications equipment. The price indexes for custom software and for “own-account” software (that is, software that companies produce for their own use) will include, for the first time, an adjustment for productivity. The price index for medical equipment are based on data from the ECRI Institute and better account for quality change than the previously used indexes. The communications indexes will include a quality-adjusted price index for cellular phones from the Federal Reserve Board, which begins in 2002. These changes are part of of a “digital economy initiative” that will probably lead to further progress in improved quality adjustment in future years.
For financial intermediaries, the updated estimates will change the methodology for measuring the output of savings institutions and credit unions. In 2003, BEA adopted the “reference rate” method for measuring the services of commercial banks. This method assigns part of the services of banks to depositors and part to borrowers, where the split is based on a reference interest rate that is risk-free and doesn’t include any services. The services of savings institutions and credit unions, however, were still measured using the older method—that is, the financial services were assigned entirely to depositors. (I don’t think there was any conceptual reason why these institutions maintained the older methodology—I think staff at the time simply didn’t have the resources to extend the new treatment to these institutions.) Because services provided to borrowers are more likely to be counted as intermediate spending, the overall effect of the new treatment will be to reduce PCE (and GDP) for financial services.
The BEA article talks about a number of other upcoming changes—to state and local pensions, to the classification of software research and development, to measures of own-account investment, to Federal Reserve payments to the federal government, and to state payroll taxes. If you regularly use the GDP or national accounts statistics, I recommend that you read it.
While not specifically addressed in the article, I’ll be interested to see how the revised estimates deal with the cloud computing puzzle that was recently identified by David Byrne, Carol Corrado, and Dan Sichel. They found that the financial reports of IT service companies that are involved in cloud computing have reported a surge in capital expenditure that doesn’t seem to show up in the GDP statistics. They believe that the national income and product accounts are missing own-account investment that occurs when these companies buy components and assemble their own equipment for use in providing cloud computing services. On Friday, we’ll see if the new BEA estimates agree with this analysis.