There are three ways to measure GDP:

• The expenditure approach adds private and public consumption, investment, and the trade balance. It’s the famous Y=C+I+G+X-M.
• The income approach principally counts labor income and profits.
• The product approach adds up each step of production.

All three measurements should add up to the same number. But, for various reasons detailed in this blog post, there’s always a small difference, called a “statistical discrepancy.” The ALFRED graph above shows this discrepancy as the proportional difference between GDP and national income. For the period just before and during the recent pandemic, that discrepancy went as high as 3%, for the first quarter of 2022.

Compare that graph with our second graph, which covers a more “normal” period. Here, the largest discrepancy is only 1.3%.

ALFRED’s job is to track “vintages” of data: In these graphs, the vintages are the values assigned to quarterly GDP. Those values for each given quarter were revised over time as more (and/or more-precise) information was collected.

During the “normal” period shown in the second graph, these revisions are minor compared with the revisions during the pandemic, shown in the first graph. This comparison highlights how difficult it was to compute the initial estimates of GDP and national income during the pandemic. Later vintages of the quarterly data had more typical discrepancies. This observation tells us that the BEA was able to adapt to the challenges of the pandemic quite rapidly and maintain the high level of accuracy in their data collection process.

How these graphs were created: Go to ALFRED, search for (nominal) GDP, click “Edit Graph,” search for (nominal) “national income,” and apply formula a/b-1. Finally, play with vintage dates and sample periods to obtain the two graphs.

Suggested by Christian Zimmermann.