Macro

Macro20 series9 revise their own history11 never restated

Every economic statistic here revises. No market rate does.

The split is total, not approximate. GDP has revised 100% of its points and payrolls 99.9%, by a median of 434,000 jobs — while the ten-year yield, the two-year, SOFR, the effective fed funds rate and every credit spread on this page have revised essentially none. A statistic is an estimate that gets corrected; a rate is an observation that was never in doubt. Any chart drawing only the latest vintage hides that difference, which is what this page exists to show.

Estimates — the first print was not the number

Median revision is the middle |latest − first print| across this series’ own restated points, in that series’ own units — never compared across rows, because a revision of 1.7M in bank reserves and one of 0.1 in the unemployment rate are not the same quantity. The median is the headline rather than the mean because a rebasing moves an index’s whole level and would dominate it; where the 90th percentile disagrees by more than twenty times, the row says rebased rather than pretending it is revision noise. Read this as: how far the number you traded on ended up from the number that was true.

Series	Latest	vs its own history	Points	Revised	Median revision	First print
GDP	32,4862026-04-01	100.0%	322	100.0%	179.71	revised
PAYEMS	159,0752026-08-01	100.0%	1,052	99.9%	434.00	revised
PCEPI	131.662026-07-01	100.0%	811	99.8%	14.95	revised
CPIAUCSL	334.132026-08-01	100.0%	955	64.0%	0.300(rebased)	revised
WRESBAL	3.01M2026-09-16	87.7%	2,229	60.5%	1.57M	revised
WTREGEN	877,0282026-09-16	95.9%	2,124	59.2%	79,621	revised
UNRATE	4.102026-08-01	18.4%	943	56.1%	0.100	revised
WALCL	6.75M2026-09-16	79.3%	1,240	43.5%	2,326	revised
ICSA	196,0002026-09-12	2.2%	3,115	33.6%	4,000	revised

core.fred_observations · fred_tier_a 06:00 ET, fred_ice_oas 12:00 ET

Observations — published once, never restated

These are not “series that happen not to have been revised yet”. A market rate is the price at which something traded, so there is nothing to correct later; the handful of changes visible here are data-entry fixes measured in hundredths, not restatements. Their first print is their only print, which is why a reader can take them at face value and cannot do the same above.

Series	Latest	vs its own history	Points	Revised	Median revision	First print
DGS2	4.762026-09-18	52.0%	12,572	0.0%	0.015	final
RRPONTSYD	0.5822026-09-21	10.0%	3,330	0.0%	5.05	final