The yield curve is positive again — what that means in the data, and what it doesn't

As of Thursday, May 21, the 10-year Treasury yielded 4.57%. The 2-year yielded 4.08%. The gap between them — the famous 2s10s spread, which FRED publishes as the pre-computed series T10Y2Y — was 0.49 percentage points, and positive. (A pairing note before anything else, because Ravi rules: I take the daily DGS10 and DGS2 closes on the same trading day, and the difference matches the official T10Y2Y reading for that day to the basis point. The single-day spread on the following session, Friday May 22, came in at 0.43; nothing in this post hinges on the gap of a day.)

Two years earlier, on May 24, 2024, the same number was −0.47.

This post is about what happened in between, what the current 0.49 actually tells you, and the much shorter list of things it does not.

The shortest history of this number I can write

When I pulled T10Y2Y back to the start of 2024, I got 598 daily observations through last Thursday. Of those, 166 were negative — about 28% of the period, or roughly the first nine months. The curve last printed an inversion on September 5, 2024, at −0.02. Since then it has been positive every single trading day, drifting gradually from near zero up to its current 0.49.

The deepest reading in that window was −0.47 on May 24, 2024. The 2-year was paying nearly half a percentage point more than the 10-year for lending the government your money over twice as long. The shape of the yield curve was, in plain language, upside-down.

The fact that this is no longer true is a real change in the data. The fact that the current 0.49 is also not historically normal is the other half of the post. The long-run average for the 2s10s, going back to the early 1980s when this comparison really starts being well-defined, sits closer to 0.95 percentage points — roughly twice where I see it now. So the curve is positive, but it is mildly positive. It has not returned to its usual posture; it has only stopped doing the strange thing.

Why 2s10s specifically

There are a lot of pairs of Treasury yields you could subtract from each other. The 2s10s — 10-year minus 2-year — became the canonical one for two reasons that are worth knowing if you are going to read anything about the curve.

First, the 2-year is the short end of the curve that tracks Fed policy expectations most directly. When markets expect the Fed to cut, the 2-year falls; when they expect hikes, it rises. So the 2-year is, roughly, "what the market thinks the Fed will average over the next two years."

Second, the 10-year is the closest thing to a "neutral" long rate — long enough that immediate Fed moves don't dominate it, short enough that liquidity is deep and the price is real. (The 30-year is similar but thinner and more pension-driven.)

Subtract the first from the second and you get a number that is, very loosely, the market's view of how the Fed's path compares to where rates "should" eventually settle. When the curve inverts, the market is saying it expects the near-term path of policy to be higher than the long-term equilibrium — usually because the Fed is fighting inflation right now and is expected to ease later. When the curve steepens to a normal positive slope, the market is roughly back to "the near-term and the long-term agree."

That's the textbook framing. The real world is messier, which brings us to the part most coverage gets wrong.

What the famous inversion-predicts-recession rule actually says

You will read, often, that an inverted yield curve "predicts" recessions. The empirical claim behind this is real and worth understanding — but it's also more limited than most retellings.

The strongest version of the rule comes out of research by economists at the New York Fed in the 1990s, refined since: every U.S. recession since the late 1960s has been preceded by an inverted 2s10s (or by an inverted 3-month-vs-10-year, which is the closer indicator the New York Fed actually publishes). That is a real pattern over about eight or nine cycles.

What the rule does not say — and this matters this cycle — is the inverse. It does not say every inversion has been followed by a recession on a fixed timeline. The lag between inversion and recession has ranged historically from about 6 to 24 months. And the strict bar is that the inversion has to be sustained, not just a single day's print.

The 2022–2024 inversion was both long (well over a year) and deep (nearly half a point negative at the trough). On the standard rule, that should have been followed by a recession within roughly two years. As of this writing, the U.S. has not had one by any official NBER call. The curve has now un-inverted, so the indicator is no longer "armed" — but the predicted recession the indicator was supposedly forecasting did not arrive on the typical schedule.

This is a real, ongoing argument among economists, and I'm not going to settle it in a paragraph. What I will say is the careful version: this cycle either broke the rule, or it merely stretched the lag past the typical 6-to-24-month window. Both are possible. Neither is what gets quoted on cable.

Bear steepening vs bull steepening (you will see these terms)

When the curve steepens — meaning 2s10s gets larger — it can happen two ways, and the distinction is informative.

Bull steepening: the short end (2-year) falls faster than the long end. Usually means the market is pricing in Fed cuts. This is the "things are getting easier" steepening.
Bear steepening: both yields rise, but the long end rises more. Usually means the long-term outlook for rates or inflation has shifted up. This is the "things are getting tighter at the back" steepening.

When I pulled the last 60 days of DGS2 and DGS10 side by side, the recent move fell cleanly into the second category. The 2-year is up roughly 70 basis points from its late-February level. The 10-year is up roughly 60. Both moved up; the spread actually narrowed slightly, from 0.59 in late February to 0.49 last Thursday — a 10-basis-point flattening on top of the broader rise. So technically the curve has flattened over the last two months even as both yields climbed. The headline "yields are up" misses that the relationship between the two changed direction inside the move, and that is the kind of mismatch I find myself wanting a chart of the spread series, not either yield alone, to make visible.

This is the kind of thing a chart of the spread series alone makes visible, and a chart of either yield alone hides.

What this does not tell you

Three limits I want to be precise about, because the failure mode of curve-watching is reading a single number as a forecast.

It does not tell you when, or whether, a recession is coming. The curve normalizing is a piece of data, not a forecast, and the un-inversion in late 2024 has been followed by neither the predicted recession nor a stable "all clear" expansion — just a wider conversation about what the indicator actually measured this cycle.

It does not tell you what the Fed will do next. The 2-year reflects market expectations of the Fed; it does not set them. Reading the 2-year as Fed guidance gets the causation backwards. If I see the 2-year fall, what I am seeing is market participants pricing in cuts; whether the Fed actually delivers those cuts is a separate question the data does not answer.

And it does not tell you where rates should be. "Above the long-run average" or "below it" describes the past distribution, not a target the present is obligated to revert to. Spreads can sit far from their averages for years — they did, in fact, sit far from average for the entire 2022–2024 inversion.

How to pull this yourself

Three free FRED series, no subscription required:

DGS10 — 10-year Treasury, daily.
DGS2 — 2-year Treasury, daily.
T10Y2Y — the spread, pre-computed daily. Saves you the subtraction.

The third one is the one to chart. Add it to a FRED graph, set the range to "Since 2020" or longer, and you can see the entire inversion-and-normalization arc on a single screen, with the zero line as your reference. The negative period from mid-2022 through September 2024 will be obvious. The recovery since will be obvious. The fact that the recovery has plateaued in the 0.3-to-0.6 range rather than running back toward the historical 1.0 norm will also be obvious — and that, more than the un-inversion itself, is the thing worth watching from here.

Next week's post is about real yields — what the 10-year actually pays you once you strip out expected inflation. FRED publishes that one too, as DFII10 (10-year TIPS yield), and it tells you something the nominal yield does not.

— Ravi