Treasury bonds are risk-free in the sense that there's little realistic prospect of either a) the US government declining to print enough dollars to pay them, or b) the US government printing so many dollars that the purchasing power of these bonds is diluted. Both of these risks are, technically, nonzero; debt ceiling debates are scary and the US has recently gone through bouts of excessively expansionary monetary policy. But the network effects of the US dollar and the relative stability of the US government mean that treasury bonds are as close as you can get to a risk-free benchmark.

At least in terms of default risk. The reason the treasury market is one of the world's liveliest, despite no real differences of opinion about whether or not the government will pay its bills, is that it represents the purest way to bet on interest rates. And since interest rates are explicitly an input into every financial decision and implicitly an input into basically every decision, it’s a big deal.

The tradeoff with benchmarking investments to interest rates is that you can move along a continuum:

• Short-term rates are, by definition, an accurate reflection of the current price of money. But you usually don't care about the current price as much as you care about the price of money over the life of whatever investment you're assessing (because that life is usually longer than the life of a t-bill).

• Longer-term rates really reflect two things: first, an expectation about what rates will look like in the future, and second, a risk premium that investors earn for taking the risk that rates will end up being different later on.

Most of the time, long-term rates are higher than short-term rates. Plot them on a curve, and it slopes upward, which makes sense. You're taking very little rates risk when you lock up money for three months; you're taking a whole lot when you buy a bond that matures in the year 2053.¹

But this doesn't always hold. Sometimes, the yield curve is flat or even inverted: the amount of money you get paid to take 10-year risk is lower than the amount you get paid to take 3-month risk. One of those times is now, and on average a yield-curve inversion predicts a recession within the next year or two.²

There are two good ways to think about why yield curves tend to imply recessions: thinking of them as a measure and thinking of them as a cause.

1. An inverted yield curve measures recession risk because it implies that the median bond investor expects short-term rates to drop over the life of a longer-term bond. Short-term rates tend to go down when there's a recession (because of the Fed cutting rates), as does inflation (because people spend less and save more). Since inflation risk is one reason why long-term bonds earn a premium, the inverted yield curve is an informative signal, at least if bond traders know what they're doing.

2. The causal story works like this: durable investments are a relatively small share of GDP, but a larger share of GDP changes. It's a lot easier for the number of new homes under construction to drop by half from peak to trough than for oil consumption to do so. Long-lived assets like houses are often funded by borrowing money, and many long-term lenders are also short-term borrowers, i.e. banks. To a bank, a normal yield curve is a subsidy on the practice of taking deposits and making mortgage loans or loans to businesses; an inverted yield curve is essentially a tax on these activities.

So it makes sense to think of an inverted yield curve as both a measure of and cause of recession risk. But economies are complex systems, and rates are set through the interaction of the real economy, financial actors, and central banks trying to adjust to the behavior of both. Those central bank adjustments usually entail trying to cool the economy when it's overheating, but also trying to engineer a "soft landing" rather than a recession. A recession will generally kill off an inflationary trend, but will also kill plenty of jobs and businesses, so it's a costly way to get the CPI back in line. And, of course, the Fed is abundantly aware of the historical correlation between yield curve inversions and future recessions.

There's a very real sense in which the Fed's job is to make anyone with a strong opinion about bonds look foolish: people who short the 30-year because the dollar's going to be worthless soon will look bad if the Fed is able to curtail inflation, while people who buy that same 30-year at 4.45% when one-month rates are 5.55% will also look bad if, in a few years, those yields have flipped.

Overall, the hardest thing to believe about the yield curve inversion model is that it would continue to work once it was widely known. If it either caused recessions or predicted them with perfect fidelity, central banks would be very reluctant to let their yield curves invert, unless they expected long-term inflation expectations to reset higher without a recession. The yield curve remains a very good way to spot-check the current macroeconomic situation, but its ability to predict what comes next is inversely proportionate to how many people expect it to be able to do so.

Read More in The Diff

In keeping with the argument that yield curve inversions are a weakening signal over time, The Diff hasn't addressed that specific topic in detail. However, the newsletter has looked at many other rates-related phenomena, including:

• Why the US electorate prefers lower inflation that policymakers ($)

• Why central banks buy bonds directly ($)

• Whether or not the treasury should issue perpetual bonds that pay a fixed amount but never mature ($)

• And why modern financial history begins in 1998, which called attention to, among other things, how recent the "stocks and bonds move inversely" narrative is, and how much it implicitly depended on China.

1. This is a good time to note that some rates are more special than others. A common way to talk about the yield curve is the spread between two-year and ten-year treasury bonds. Why those two? The ten-year is very liquid, while the 20- and 30-year are less so—and the treasury stopped issuing 30-year bonds after a few years in the early 2000s, so the long-term data for those are wonky. Meanwhile, rate changes tend to happen in cycles; hikes follow hikes, and cuts follow cuts. So a short-term treasury will not fully price in expectations for where short-term rates will land. The two year turns out to be a happy medium: maturing soon enough that it's highly responsive to Fed policy, but maturing long enough in the future that its value is a function of what they did at the last meeting and what they've said they'll probably do at the next few.

2. That average has at least one ambiguous data point: the yield curve inverted in late 2019, and there was indeed a recession soon after, though its cause had very little to do with interest rates. And that’s compared to a sample size that, depending on exactly which rates you use and how many crossovers from inverted to not you count, ranges from around half a dozen to a dozen data points in the US since the modern Fed started in 1951.

Help us learn more about you!

We're building a media kit and need some anonymous aggregated subscriber data. If you work in finance and have time for a 3-question survey, that would be very helpful:

Join the discussion!