At 8:30 this morning, BEA released the December personal income and outlays report, and we learned that personal income increased 0.3 percent and real consumer spending (or personal consumption expenditures (PCE)) increased 0.3 percent in the final month of 2016. These estimates were prominently reported by the news media.
Less than a half hour later, BEA posted another set of tables with much less fanfare, but several hundred economic forecasters eagerly awaited their arrival. BEA describes these tables as “supplemental estimates” that are associated with and support the quarterly national income and product accounts and the monthly personal income and outlays estimates. While these data are primarily of interest to macroeconomists engaged in forecasting, they are also an example of detailed data that support the transparency of the featured GDP and personal income estimates.
Among these tables was a spreadsheet of “Key Source Data and Assumptions for ‘Advance’ Estimate.” This table goes through each of the major components of GDP and presents key source data for each component side-by-side with the NIPA estimate. For example, for the NIPA estimate of nonresidential investment in structures, the estimate mostly comes from two sources—the Census Bureau’s monthly construction spending report and, for oil and gas well drilling, data on footage drilled from the American Petroleum Institute. These data are shown side by side in the table, allowing users to see the relationship between the source data and the NIPA estimates.
This table also allows users to see the assumptions that BEA makes when data are incomplete—for example, when data were not available for the final month of the quarter (as was the case for construction spending). Also, for data series that are seasonally adjusted by BEA, the table presents the seasonally adjusted estimates.
In addition to the key source data, BEA also presents the “Underlying Detail Tables.” These tables are like the quarterly NIPA tables that were released Friday along with the fourth quarter GDP report, but they present more detailed categories and, in some cases, higher frequency estimates (such as monthly estimates for inventory investment).
Why does BEA keep these underlying detail tables separate from the main NIPA tables? According to a note on the website, “The Bureau of Economic Analysis (BEA) does not include these detailed estimates in the published tables because their quality is significantly less than that of the higher level aggregates in which they are included. Compared to these aggregates, the more detailed estimates are more likely to be either based on judgmental trends, on trends in the higher level aggregate, or on less reliable source data.”
Fundamentally, what’s going on here is that when BEA calculates real GDP, it uses price indexes to deflate nominal spending for hundreds of spending categories. For example, the nominal PCE for higher education is deflated using a CPI for college tuition and fees.
Unfortunately, the source data for detailed spending categories are not always available at monthly frequency. BEA may have monthly data on things like car purchases, electricity and gas consumption, and movie theater admissions. For other spending categories, like physicians services, they may have quarterly data from the quarterly services survey.
For some other categories, BEA may have high frequency source data that don’t align well with the PCE deflation-level categories. For example, Census reports nominal sales at mass merchandise stores, but those stores sell items from several dozen PCE categories. To calculate monthly PCE, BEA may have to assume that the distribution of items sold within each category of stores stays the same until additional data become available (either annually or from the 5-year economic census) to update the distribution. Consequently, BEA’s cautionary note emphasizes that users shouldn’t expect all the underlying detail categories to reliably measure the month-to-month changes in spending within those categories. On the other hand, movements in total spending on goods purchased from retail stores does match the Census monthly retail sales figures, so overall consumer spending on goods should be more reliable than the details.
While most data users probably won’t need to use data at this level of detail, the underlying detail tables and key source data are important for many of BEA’s most sophisticated and important data users, such as the Federal Reserve staff and private sector macroeconomic forecasters. Furthermore, by allowing users to largely replicate the GDP estimates, they help keep the estimates open and transparent.