Why does the growth rate of GDP as reported by the BEA differ from a computation on Data Buffet?
Have you ever done this?
- Download the "percent change from preceding period" growth rates, as published by the BEA. (This report includes annual/quarterly and current/chained figures.)
- Apply a Q/Q percent change transformation to Data Buffet series GDP$.US. (chained dollars, SAAR).
- Compare the two, and see that they don't match exactly?
This is not a mistake.
The U.S. Bureau of Economic Analysis (BEA) computes their growth rates using the quantity indices (e.g., PDQGDP.US), not the dollar levels (GDP$.US), and then rounds to one decimal place. The two series have different precision, so any transformation will produce divergent results. The divergence is statistically insignificant -- it is easily swamped by estimation uncertainty and subsequent revisions.
Specifically, as reported and as stored in Data Buffet, the quantity indices have three decimal places of precision, and dollar levels have one place. The difference between their % Q/Q transformations is less than 0.01 percentage point. For example:
For the National Economic Accounts NIPA dataset, the BEA initially publishes estimates with billion-dollar precision, followed by million-dollar precision, typically with the Personal Income and Outlay release one day later. At present, Moody's Analytics does not incorporate the million-dollar estimates into our NIPA dataset. We plan to do so and will post a timeline for when the detailed data will be available and how it will affect our update schedule, estimates, and forecast series.