The Inversion That Took Two Years to Mean Anything

I was right about the yield curve and it cost me money, and both halves of that sentence are true in the same way, which is the part worth explaining.

The indicator did what the record said it would do. I read a statement about years as a statement about months. The cost of that misreading was paid every day for about eighteen of them, and it was collected back, partially, on one date I could not have named in advance.

What the curve was telling me, precisely

An inversion is a statement about expectations. Longer dated Treasury yields sit below shorter dated ones when the market expects the policy rate to be lower in future than it is now, and that expectation ordinarily rests on a view that growth or inflation will slow enough to force easier policy. It is a price, set by people committing real money, summarizing what they think the central bank will be obliged to do years ahead, which is why I still regard it as one of the more honest objects in finance.

The record is genuinely respectable. The spread between the two year and the ten year, and the three month against the ten year, have preceded the postwar recessions with a consistency that few macro indicators come close to.

What the record does not contain, anywhere, is a date. The curve inverted in 1978. The downturn that followed took the best part of two years to arrive. It inverted in 2000 and the recession was dated to March 2001. It inverted during 2006 and the recession was eventually dated by the National Bureau of Economic Research to December 2007, well over a year later, during which time equity markets made new highs. The 1966 episode produced a slowdown and no recession at all. That is the case everyone quoting the record politely omits.

Postwar recessions are rare events, so the sample is small, so the average of those lags is an average of a handful of numbers with an enormous range around it. I knew every word of that. I could have written it from memory. I still behaved as though the curve had told me something about the coming quarter.

What I did, and what it cost

I moved a meaningful share of a portfolio toward the defensive end. Shorter duration and more cash. Less in the parts of the equity market that suffer when growth slows.

Nothing about those instruments was wrong. The clock was wrong. The expansion continued. Equities did what they do during an expansion. My cautious allocation earned its cash yield and waited.

When the slowdown eventually arrived, the defensive position behaved exactly as designed, and the protection it delivered was smaller than the return I had given up while waiting for it, so the indicator was vindicated and the trade was still a loss, and I have come to think that combination is the normal outcome for anybody who converts a slow signal into a fast decision.

How I read it now

The curve is permanent furniture. Its role has changed from signal to context.

I watch the shape now. The crossing is one moment inside it. A curve that inverts, deepens, and then steepens sharply as the front end falls is telling a different story at each stage, and the steepening that follows an inversion has historically sat closer in time to the trouble than the inversion did. Watching that shape evolve across months on the yield curve tool carries more information than checking whether one spread is above or below zero.

I stopped using it alone. It belongs to a family of indicators with their own records and their own false signals, and they do not fire together, or in a reliable order, or with anything like the same warning. Claims data, the Sahm rule and the leading index each have something to say, and the case for reading them as a group is made in recession indicators.

The change that mattered most was structural. I no longer ask what the curve means for the coming quarter. I ask whether the allocation I hold is one I would be content to hold through a recession beginning either next month or in three years. If the answer is yes, an inversion requires nothing of me at all. If the answer is no, then the allocation was the problem and the curve has merely drawn my attention to a decision I should have made before it inverted.

The shapes, the spreads people actually watch and the historical record with its qualifications are set out in the yield curve explained. If the price and yield mechanics underneath are unfamiliar, how bonds work covers the seesaw between the two, which is what makes the whole curve move in the first place.