A curve steepening back out of inversion
In every month the curve was still inverted but had been steepening for a quarter, what did unemployment do over the following year?
10 windows met these conditions and completed 12 months. US Unemployment Rate was higher 12 months later in 10 of them and lower in 0. The middle half of outcomes fell between 0.45 and 1.
What followed, across 10 completed windows
- Higher after
- 10
- Lower after
- 0
- Middle half
- +0.45 to +1
- Full range
- +0.1 to +2.5
Change in US Unemployment Rate over the 12 months after each match. Windows are listed below so any one can be checked.
Term structure
US Unemployment Rate measured 1, 3, 6 and 12 months after each matched episode, with the count of completed windows at each horizon — not a forecast path, the same distribution above asked again at three more calendar distances.
| Horizon | N | Higher / lower | Middle half | Full range |
|---|---|---|---|---|
| 1m | 10 | 4 / 1 | 0 to +0.1 | -0.1 to +0.4 |
| 3m | 10 | 7 / 3 | -0.05 to +0.35 | -0.3 to +1.2 |
| 6m | 10 | 7 / 2 | +0.02 to +1.13 | -0.2 to +1.4 |
| 12m | 10 | 10 / 0 | +0.45 to +1 | +0.1 to +2.5 |
N is windows with a complete outcome at that horizon; a window that runs past the end of the record is excluded from it, not filled in. A horizon below the 5-window floor is marked thin the same way the 12 months result above is.
The conditions, in full
Generated from the same object that produced the numbers, so the description cannot drift from the computation.
- condition: T10Y2Y below 0
- condition: T10Y2Y rising over 3 observations
- align: each condition is put onto the measured series' dates, taking the last observation on or before each, so windows and spans count in the measured series' own periods
- measure: UNRATE over the following 12 observations
- matches within 3 observations count as one episode
Every matching window
10 episodes, each dated. Consecutive months meeting the conditions count once.
| Began | Ran for | US Unemployment Rate then | After | Change |
|---|---|---|---|---|
| Feb 1979 | 7 mo | 5.9 | 6.3 | +0.4 |
| Feb 1980 | 4 mo | 6.3 | 7.4 | +1.1 |
| Feb 1981 | 3 mo | 7.4 | 8.9 | +1.5 |
| Oct 1981 | 1 mo | 7.9 | 10.4 | +2.5 |
| May 1982 | 3 mo | 9.4 | 10.1 | +0.7 |
| Sep 1989 | 1 mo | 5.3 | 5.9 | +0.6 |
| Jul 2000 | 6 mo | 4 | 4.6 | +0.6 |
| Mar 2007 | 4 mo | 4.4 | 5.1 | +0.7 |
| May 2023 | 9 mo | 3.6 | 3.9 | +0.3 |
| Jul 2024 | 2 mo | 4.2 | 4.3 | +0.1 |
Recompute this
Fetch T10Y2Y UNRATE from the API or from FRED, apply each condition on the same dates, collapse consecutive matches, and measure the change over the following 12 months. The full rule is in the claims ledger.
This page counts and dates. It does not say what will happen next, and no window above is presented as more likely than another — MacroRadar publishes historical indicators, never forecasts.