The dollar is up, term premiums have come down, credit spreads are wide, and the stock market is down. That means there’s a heightened risk of a recession, right? Maybe, but probably not as bad as the market suggests. For decades, economists and pundits have turned to a range of recession indicators, from yield curve trends, to commodity price movements. While some of these methods have merit, they’re often incomplete, too subjective, or don’t take into account our economy’s multifaceted nature.
Because investors and advisors alike are wondering if a recession will once again rear its ugly head, we set out to create a better indicator. Using a 15-part pipeline of economic and market data, we developed a predictive analytics model to determine the probability the U.S. economy will be in a recession in six months. Here are the highlights, followed by insights for investors and details on how we created the model.
First, here’s a breakdown of the economic data we fed into our predictive model.
Next, let’s see the data in action. As the model points out, there’s been a strong correlation between the probability of recession and its actual occurrence over the past four decades.
Focusing on the next six months, our predictive model says there is a 12% probability of a recession occurring. For perspective, what the model is showing looks no worse than what was predicted from 1983 to 1988, or in the late 1990s. This justifies why we view the recent market corrections as being driven more by fear than reality. In 1984, there was a fear that inflation would rear its ugly head again, but it didn’t. The fear was real, but reality didn’t live up to the fears. We’re probably going through something similar now.
Policy errors—like the Federal Reserve being too eager to hike interest rates—could tip the economy into recession and turn a market correction into a bear, but that seems unlikely. Considering it is a risk, we recommend that investors consider the following:
- Fixed income: It’s still prudent to keep those boring core bond funds in your portfolio. Over the past year, the Barclays Intermediate U.S. Credit Aggregate Index returned 0.29%, which might not seem like a lot, but it’s better than the S&P 500 Index’s total return of -6.85%. You want something that zigs when other parts of your portfolio zag.
- Alternatives: It’s also wise to allocate a decent amount (around 10%) to a diversified alternatives strategy. Relative value, event driven, macro driven, and other alternative strategies can not only help diversify the ideas reflected in your portfolio; they can also be less correlated to your equity exposure.
- Equities: In a market rebound—even a rocky rebound like we’re seeing now—the parts of the market that were most oversold can represent the best opportunities. That means non-U.S. equities, which broadly entered a bear market while the S&P 500 Index merely slid into a correction, can represent good value. In the U.S., mid- and small-cap stocks were the relative underperformers over the past few years; those ponds might be the best for fishing in now.
In addition to explaining our model, we thought it would prove helpful for readers to gain a sense of how the markets have traditionally gauged recession risk in recent years. Some traditional indicators—such as yield curves and stock prices—play an important role in our approach.
The yield curve
Many people measure recession risk by looking for an inversion in the yield curve, which is the graphical depiction of the relationship between the yield on a Treasury security and its maturity. An inverted yield curve is when long-term yields—typically the 10-year—drops below a short-term security’s yields. Paul Samuelson famously said in the 1960s that the market correctly predicted nine of the past five recessions. In other words, there are a lot of false positives. The yield curve’s predictive value hasn’t improved much over the years.
Some people measure whether the yield curve inverts based on the difference between the 10-year Treasury and the federal funds rate, or a short-term interest rate, like the 3-month Treasury bill. I’ve found that those aren’t as reliable as comparing the 10-year yield to the 2-year yield, which is one of the 13 indicators we use in our predictive model. Especially in today’s environment in which the Fed has short-term yields stomped down close to zero, it would be nearly impossible to get the yield curve to invert. That doesn’t mean the yield curve lacks any predictive power; it just needs to be taken with a grain of salt.
Since 1950, there have been 14 recessions. Before the 1953 recession, the yield curve did not invert. However, the yield curve did invert four times (based on month-end data) without there being a recession within at least two years in four instances: 1965, 1967, 1998, and 2005. So, as a recession indicator, it missed one and called four that weren’t. For those recessions that the yield curve got right, the curve inverted—on average—13 months before the recession hit, although it’s ranged from 8 months to 19 months.
You can do better than simply looking at the yield curve by considering changes in the term premium. Lately, the difference between the 10-year and 2-year Treasury has narrowed. Some of this—maybe a lot of it—reflects an emotional flight to safety by investors. It could also reflect concerns about inflation being too low or growth faltering. There could also be a structural reason for this drop in yields, as regulations have forced many banks to own liquid assets, such as Treasury securities. Regardless, the term premium has dropped 0.36 percentage points relative to its 6-month average.
As part of our predictive model, using the level of the 2-year Treasury yield, the difference between the 10-year and the 2-year and the difference relative to the moving 6-month average contributes a 1.8% probability of there being a recession in 6 months.
Stock market returns
But the dollar has been strengthening and commodity prices have fallen. That surely must mean something, right? Yes, but as part of our predictive model, these factors move the probability up to 5.3% from 1.8%. The dollar was rising up to 1985 and then again from 1989. That hurt specific companies and certain industries, but it didn’t cause system-wide problems. Similarly, the drop in oil prices—while extreme—isn’t wholly unprecedented. Broadly, oil price declines cause some short-term pain in the energy sector, but sustained lower prices are conducive to growth.
Don’t mistake the current pain in concentrated areas for some broader economy-wide problems.
The widening of credit spreads and the drop in stock prices could be pointing to a slowdown, but it seems like the market often gets the direction right without nailing the magnitude. Right now, the widening of spreads and drop in stock prices—which are 2 of the 13 indicators in our predictive model—suggests a 20% probability of there being a recession in six months. Since 1950, the economy has been in recession 16% of the time, but credit spreads and the stock market would predict the economy being in recession 25% of the time. So, I think it’s important to temper market predictions with other economic data series that might not be as noisy or fickle.
Leading economic indicators
The search for less-fickle predictors is what led The Conference Board to construct the aforementioned 10-part index of Leading Economic Indicators (LEI). While the LEI has posted declines for two months, most of that came from the stock market, while other economic indicators such as manufacturing hours, Institute for Supply Management new orders, and money supply growth have performed quite well. This is why, in our predictive model, we break down the LEI into subcomponents, because not every piece is of equal value in predicting recessions.
We also made some adjustments. We used the 10-year Treasury minus the 2-year Treasury instead of The Conference Board’s 10-year minus federal funds rate as an indicator. We also added:
- Credit spreads, as measured by the Moody’s BAA yield minus the 10-year Treasury yield
- The six-month percentage change in the CRB spot commodity index and raw material index
- The six-month percentage change in the dollar’s nominal trade-weighted index
This gives a pretty rich model for determining the probability that the economy will dip into a recession in six months and is how we reached our calculation of 12%.
Methods we did not use
Despite what some introductory textbooks say, recessions are not defined as periods when gross domestic product drops two consecutive quarters. The National Bureau of Economic Research’s Business Cycle Dating Committee is the widely accepted arbiter of recession dates. It uses a loose definition that focuses on whether a slowdown is systemic (goes throughout the whole economy), sustained, and significant. To a large extent, professional judgment is exercised in making this determination. There’s also a long lag in when the economy’s turning points are identified. A lot of macroeconomic data is subject to revision even well after the original publication date. That’s why it can take years to call when a recession occurred, and that’s why more real-time measures are sought.