How we rate the impact of a release
Kitalpha’s impact tier measures how much the U.S. Treasury yield curve actually moved on an indicator’s past release days, compared with a quiet day on which nothing we track was published. This is a backward-looking measure of realised volatility on past release days. It is not a forecast and it is not investment advice.
The measure
The absolute change in the U.S. Treasury par yield curve from the previous trading day's close to the release day's close.
Why the release day, and not the day after
The Treasury par curve is struck at 15:30 Eastern. The jobs report and CPI land at 08:30 Eastern and the FOMC statement at 14:00 Eastern, so each is already inside that same day's close. Measuring the following day's move would measure the day after the reaction.
The score: the larger of two excesses
The larger of two excesses: the excess move in the 10-year yield (the curve's LEVEL) and the excess move in the 2s10s spread (the curve's SHAPE). A rate decision acts on the front end, so it moves the shape far more than the level — scoring on the level alone rated the FOMC 'low', which was plainly wrong.
The quiet-day baseline
Each excess is measured against a QUIET day: a trading day on which none of the tracked indicators released.
Days with two releases are thrown away
A day on which two or more tracked indicators release is EXCLUDED from every affected indicator's sample. The day's move belongs to both releases and to neither, and it cannot be attributed from a closing price.
The thresholds
The tier is a bucketing of the score. The cutoffs are absolute basis-point numbers, not percentiles: a percentile would mean an indicator’s tier could change colour because we started tracking a different release, which would be absurd. They were chosen as round numbers after inspecting the distribution, then frozen — and we disclose that ordering rather than imply the lines came first.
| Tier | Excess over a quiet day |
|---|---|
| High | ≥ 2.00 bp |
| Medium | ≥ 0.75 bp |
| Low | < 0.75 bp |
The sample floor
Below 20 clean (single-release) days we publish no tier at all. A pill drawn from six observations looks exactly like a pill drawn from six hundred, and that is precisely the problem — so we render nothing rather than something that reads as a finding. On the calendar, an unrated release shows no pill; it does not show a default “low”.
Which stretch of history the tier is cut on
A tier averaged over twenty-five years is well sampled and regime-blind: CPI in 2004 was not the event CPI became in 2023, and the FOMC of 2016–21 was a committee at the zero bound telling everyone in advance that it would do nothing. The published tier is therefore cut on the trailing five years — the regime a reader is actually standing in. Where five years cannot reach the minimum clean sample, the tier falls back to full history and the page says so.
| Release | Trailing 5 years (published) | 2016 to date | Full history |
|---|---|---|---|
| Consumer Price Index | Medium 1.91 bp · n=57 | Medium 1.17 bp · n=107 | Low 0.50 bp · n=219 |
| Employment Cost Index | below floor n=3 | below floor n=8 | Low -0.35 bp · n=26 |
| Employment Situation | High 3.14 bp · n=55 | High 2.15 bp · n=99 | High 2.26 bp · n=248 |
| FOMC Meeting and Rate Decision | below floor n=18 | Low 0.43 bp · n=41 | Medium 0.93 bp · n=101 |
| Gross Domestic Product | Low 0.46 bp · n=46 | Low 0.00 bp · n=100 | Low 0.27 bp · n=223 |
| New Residential Construction | Low 0.35 bp · n=90 | Low 0.18 bp · n=194 | Low 0.19 bp · n=373 |
| Job Openings and Labor Turnover Survey | Low 0.53 bp · n=49 | Low 0.17 bp · n=102 | Low 0.01 bp · n=153 |
| Personal Income and Outlays | Low -0.22 bp · n=33 | Low 0.14 bp · n=80 | Low 0.20 bp · n=209 |
| Producer Price Index | Low 0.20 bp · n=47 | Low 0.16 bp · n=97 | Low -0.12 bp · n=182 |
| Advance Monthly Sales for Retail and Food Services | Low 0.56 bp · n=46 | Low 0.56 bp · n=85 | Medium 0.88 bp · n=188 |
| U.S. International Trade in Goods and Services | Low 0.24 bp · n=51 | Low 0.24 bp · n=88 | Low 0.43 bp · n=213 |
The three columns are published together because the disagreement between them is itself the finding. Consumer prices rate low across the full record and medium over the last five years — a measurable record of a release the rates market learned to care about, and roughly when. Where the trailing window cannot reach the sample floor, the published tier falls back to the full history and the release’s own page says so.
Quiet-day baseline: 4.55 bp over the trailing five years (671 quiet days) · 4.08 bp over the full history (3,497 quiet days). Source: U.S. Department of the Treasury.
Every indicator, with its numbers
The excess, the raw mean and the baseline are all published, so the bucketing can be checked against the numbers it came from — and disagreed with. These are the numbers behind the published tier — the trailing-five-year window, or the full history where a release fell back to it.
| Indicator | Tier | Score (bp) | From | Excess 10Y | Excess 2s10s | Mean |10Y| | Median |10Y| | n clean | n excluded (multi) |
|---|---|---|---|---|---|---|---|---|---|
| Employment Situation | High | 3.13 | 10-year level | 3.13 | 1.31 | 7.69 | 8.00 | 55 | 5 |
| Consumer Price Index | Medium | 1.90 | 10-year level | 1.90 | 1.16 | 6.46 | 5.00 | 57 | 4 |
| FOMC Meeting and Rate Decision | Medium | 0.93 | 2s10s curve | 0.32 | 0.93 | 4.41 | 4.00 | 101 | 96 |
| Advance Monthly Sales for Retail and Food Services | Low | 0.55 | 10-year level | 0.55 | -0.22 | 5.11 | 4.00 | 46 | 17 |
| Job Openings and Labor Turnover Survey | Low | 0.52 | 10-year level | 0.52 | 0.33 | 5.08 | 4.00 | 49 | 10 |
| Gross Domestic Product | Low | 0.45 | 2s10s curve | -0.06 | 0.45 | 4.50 | 3.00 | 46 | 19 |
| New Residential Construction | Low | 0.34 | 2s10s curve | -0.04 | 0.34 | 4.52 | 4.00 | 90 | 23 |
| U.S. International Trade in Goods and Services | Low | 0.22 | 10-year level | 0.22 | 0.18 | 4.78 | 5.00 | 51 | 9 |
| Producer Price Index | Low | 0.19 | 2s10s curve | 0.12 | 0.19 | 4.68 | 4.00 | 47 | 14 |
| Personal Income and Outlays | Low | -0.23 | 10-year level | -0.23 | -0.89 | 4.33 | 3.00 | 33 | 26 |
| Employment Cost Index | Low | -0.35 | 10-year level | -0.35 | -0.63 | 3.73 | 3.50 | 26 | 88 |
Source: U.S. Department of the Treasury daily par yield curve, and the publishing agencies’ own release-date histories. · Computed · Methodology version 2026-07-release-day-excess-trailing5y-v2
These tiers are a backward-looking measure of realised volatility on past release days. They are not a forecast of the next release's effect, and nothing here is investment advice.
Why there is no consensus column
The “expected” figure on a commercial economic calendar is a survey of economists, compiled and sold by a data vendor. It is not a work of the U.S. government and there is no public-domain source for it. We do not license it, so we do not publish it — and we say so on every event page rather than leaving an empty column that reads as missing data. For the same reason you will find no prediction-market odds anywhere on this calendar: venue market data carries the venue’s own terms, and redistributing it here has not been cleared.