About revisions and seasonality in the era of a pandemic

It is important to remember, especially with some current forecasts suggesting negative growth in the second quarter, that GDP figures are revised. The first release is called a “pre-release” because a significant proportion of the input is evaluated. Below are the various variations of GDP, starting from the low, to better highlight recent changes.

Figure 1: GDP, various periods, in billions of US dollars, 2012, SAAR. Source: BEA via AFRED.

The bugs between the pre-release and the third release are shown in Fig. 2 (in difference logs; hence 0.0002 indicates a 0.02% change).

Figure 2: Changes between pre-release and third release in logarithmic terms. Source: BEA, author’s calculations.

There are always revisions. Previous discussion of the GDP revision by the G7 here. Clearly, in contrast to earlier studies, these revisions tend to be positive, although I have not tested whether the difference is statistically significant.

Note that in the second quarter of 2020, there was a huge positive shift from pre-release to third release—more than half a percentage point in logarithmic terms. This is understandable given the turmoil associated with this period. However, the fact that there was a huge drop in GDP in the second quarter of 2020 raises the question of how to deal with seasonality after that quarter. This problem is known to statisticians – see Lucca & Wright (2021) as well as BLS (2022).

This brings us back to what different approaches to seasonal adjustment show us. Figure 3 shows the official BEA series, the official non-seasonally adjusted series (all recent vintages) and my special seasonally adjusted series using the period 2002-2022Q1, Census X-13 with a log transformation and an additional outlier for 2020Q2.

Figure 3: Seasonally adjusted GDP according to BEA data (bold black), unseasonally adjusted (yellow brown), and seasonally adjusted using an X-13 log transformation with additional emissions in the second quarter of 2020 (red). Peak-to-trough dates as determined by the NBER are in grey. Source: BEA, NBER and author’s calculations.

Please note that my seasonal adjustment process will not repeat BEA. BEA applies seasonal adjustment to the entire sample of individual components before summing. I applied a seasonal adjustment filter to the aggregate series for 2002-2022, the period for which I have NSA data. Therefore, my point here is not to argue that GDP growth is more correctly measured as positive in the first quarter; rather, it is to highlight the pitfalls of overinterpreting the aq/q pre-release. (In fact, on July 28, estimates for the first quarter of 2022 will be revised as part of the annual benchmark review.)