A US recession is imminent

This post first appeared on the Investing with the Trends blog.

Recessions are rare events. Since 1990, the US economy spent about 10% of the time in a recession and 90% of the time in a recovery or an expansion. Most of the time, investors don’t have to worry about it. About once every 10 years, however, we better watch out.

Recessions trigger bad bear markets, leading to big losses in the stock portion of investors’ portfolios. It’s worth trying to anticipate them.

The world is awash in economic data. Data series on payrolls, layoffs, consumer confidence, retail spending, manufacturing outlook, imports and exports, and hundreds of other items get reported continuously. Most of these series are noisy and weakly predictive. We look for indicators that are simple and robust, that have an economic rationale behind them, and that have a long history of prediction of recessions with few or no false positives and negatives. There are two data series that meet that requirement: the slope of the yield curve and the change in unemployment rate.

Yield curve slope

The yield curve indicator was published in 1986 by Professor Campbell Harvey in his PhD thesis. (I recommend this 1989 article in the Financial Analysts Journal as a better introduction.) It has a long out-of-sample history so we can say with confidence that it was not overfitted and it has held well since its publication.

The slope of the yield curve is defined as the yield on 10-year Treasury securities minus the yield on 3-month Treasury securities. The yield curve is an indicator of the sentiment in the economy. Normally, long term interest rates are higher than short term interest rates: investors demand higher interest from their bond holdings than from short-term CDs and Treasury bills. When people expect bad news in the future, however, there is a flight to safe, high-quality assets—like US Treasuries. On occasion (about 10% of the time) that depresses long term interest rates below short-term rates and these occasions are pretty much always followed by a recession.

The only false signal in the past 60 years came in 1966 during a time when US bank interest rates were regulated and subject to interest rate ceilings. In 1971, the first money market mutual fund (the Reserve Fund) launched and it was followed by many others, creating a deregulated marketplace for short term interest rates that remains to this day. The 1966 false signal can therefore be chalked up to a financial environment that no longer exists.

From a financial economists’ perspective, the performance of the yield slope indicator has been somewhat disappointing. When it was published, there were expectations that the level and persistence of the indicator could be used to forecast both the exact timing and severity of the next recessions. Those expectations didn’t pan out.

The magnitude of yield slope inversion has no relationship to how bad the recession will be. The inversion preceding the global financial crisis of 2008 was smaller than the inversion preceding the much milder 2001 recession, for example. The 2021 covid recession was preceded by about the same inversion as 2008. By itself, the yield curve cannot predict the policy response to the recession or how long it will take to emerge from it.

Timing wise, the inversion is not great either. There can be a long lag, 18 months or more, between the time that the yield curve first inverts and the time the recession starts. During the global financial crisis, the yield curve inverted in July 2006 (more than 2 years before the start of recession) and returned to positive slope in May 2007 (a full year ahead of recession).

Invariably, pundits declare that the indicator has stopped working and the investing public ignores the inversion after a while. Amusingly, the inventor of the yield curve himself has fallen prey to this impatience. In May 2008, Professor Harvey gave a presentation in which he cast doubt on his own indicator and forecasts a period of slower growth rather than a recession.

From a market practitioner’s point of view, however, the yield curve indicator is amazing. Not only has it flagged every recession since 1970, it has also correctly classified many market panics that didn’t translate into economic problems.

As an example, in 1987 the US stock market lost 23% of its value on Black Monday, followed by crashes around the world. There were fears of a repeat of the 1929 stock market crash aftermath, maybe even a depression. The yield curve slope didn’t flinch and showed that the economy was likely to be fine. It was also correctly unconcerned about the peso crisis in 1995, the Asian financial crisis in 1997, the Russian default in 1998, and the European sovereign debt crisis in 2013. What’s amazing to me is that the yield curve inverted half a year before the covid pandemic, sending a warning signal that something was afoot. The most likely explanation is that the economy was slowing and headed into a recession anyway.

The biggest weakness of the yield curve indicator is its variable timing. The yield curve starts inverting well before a recession begins. Sometimes, the recession doesn’t start until the yield curve resumes a positive slope, like in 2001. Sometimes, it takes a significant amount of time even after the inversion ends, like in 1991 and 2008. Other times, the recession starts almost immediately, like in 1969 and 1982.

Could we find a different indicator that has a better timing performance? Yes, we can!

Change in unemployment rate

Recessions always come with rising unemployment. Actually, high unemployment is almost the definition of a recession. It is difficult to imagine a recession with low unemployment.

The problem with unemployment rate is that it is a lagging indicator. By the time it reaches high levels, it’s pretty obvious to everyone what’s going on. The ideal indicator would trigger at the time when unemployment bottomed and started accelerating.

Various ways have been proposed to filter the noisy unemployment time series to extract a signal. The economist Claudia Sahm published what is now known as the Sahm rule in early 2019 when she worked at the Federal Reserve. The Sahm recession indicator “signals the start of a recession when the three-month moving average of the national unemployment rate (U3) rises by 0.50 percentage points or more relative to the minimum of the three-month averages from the previous 12 months.”

The Sahm indicator triggered once since its publication when it correctly marked the 2020 recession so its out-of-sample track record is limited. It has been criticized for being overfit and I have some sympathy for that criticism. The 0.5 threshold has certainly been finessed to avoid false signals in 1967, 1976, and 2003 when the indicator spiked up to levels just under the threshold. On the other hand, there is plenty of economic support for the idea that some kind of detection of a turn in unemployment rate would be an effective predictor of recessions. And the Sahm rule is a pretty reasonable changepoint detector.

The rules of the Sahm indicator are not chosen at random. The raw unemployment rate data series is quite noisy, jumping up and down from month to month, and is subject to revisions. It needs to be smoothed. Three month moving average is a simple and effective smoother. It’s about the shortest lookback window that works (two-month smoothing is a bit pointless, and one-month smooothing is just the raw observation). Longer lookback windows would produce less noisy signals but they would add lag to what is already a lagging indicator. Comparing to 12 month smoothed minimum also makes sense, as it will catch moves off the bottom that are jaggedy or interrupted. As trend followers, we spend a lot of time studying smoothing and filtering algorithms. Sahm’s rule is, in fact, a trend following algorithm applied to an economic data series.

Over the past 60 years, the Sahm rule and the yield slope indicator agreed on every single recession, both avoiding false signals. The Sahm rule has been better at timing the onset of recessions, but it is newer and hasn’t had the chance to withstand the test of time. I think that a good compromise is to look at both when evaluating the current economic environment.

Current outlook

For the first time since 2020, both the yield curve indicator and the Sahm rule agree that there is a recession coming. The yield curve inverted in November 2022 and stayed negative since then. The Sahm rule kept its cool and only triggered in July 2024. August numbers recently came out and confirmed its upward trajectory.

Most economic indicators are noisy and subject to giving false signals. That’s great for professional economists who need to have something to talk about during the years of economic expansion, otherwise the CNBC interview invitations would stop coming. Investors, who just need a warning signal the 10% of time that recession is imminent, should prefer simple and robust indicators with few or zero false positives. But when those indicators trigger, they should be taken seriously.

We believe that the yield curve indicator and the Sahm rule are correct and that a recession in the US is imminent. We don’t know how prolonged or how deep it will get.

We recommend that investors review their portfolios and ask themselves, “Am I prepared for a recession that could lead to a 50% drawdown in the US stock market?” Because, historically, 50% drawdowns are not at all unusual during economic downturns. A properly positioned portfolio should be diversified and provide downside protection as well as upside potential. If you’re unsure whether your portfolio is ready, give us a call. We have some ideas.