The quarter-on-quarter growth rate for Q2 at 0.9% was below consensus by 1.3 ppts (SAAR), but above GDPNow and IHS-Markit (0.4 ppts, 0.5 ppts respectively). While the mean revision for advance to third release is essentially zero over the 1996-2020 period (see BEA), over the last 5 quarters (2021Q1-22Q1), they have been 1.9 ppts (MAR 12.9 ppts, RMSR 6.8 ppts). That is, the characteristics of revisions from advance to 3rd seem to have changed (although I’ve not done a test for statistical significance), with the advance being revised upward over releases.
Figure 1: Actual advance GDP growth rate (bold black), latest vintage (thin black), Atlanta Fed GDPNow for 7/27 (red line), St. Louis Fed News index for 7/22 (teal line), IHS Markit for 7/27 (sky blue triangle), Goldman Sachs for 7/27 (blue +), Bloomberg consensus as of 7/27 (pink open square), all in %, SAAR. NBER defined peak-to-trough recession dates shaded gray. SAAR. Source: BEA, BEA via ALFRED, Atlanta Fed, St. Louis Fed via FRED, IHS-Markit, Goldman Sachs, Bloomberg, and NBER.
Five quarters is a small sample, so don’t take too much from the mean revision (that is, the uncertainty surrounding the point estimate is large given the small sample). But something to keep in mind.
And as for advance to final, for the 1996-2020 period, while mean revision was zero, the MAR was 1.2 ppts, RMSR 0.9 ppts. That suggests, to me at least, that advance to final revisions will be larger for the 2021-22 period.