Confidence
Tracking credit spreads is our version of reading 10-Ks in the bathtub
Every day investors are greeted with a stream of headlines proclaiming what is special about this moment in market history. Some narratives from the past 12 months have included Evergrande, supply-side inflation, demand-side inflation, the Federal Reserve raising rates, and most recently Russia’s invasion of Ukraine.
These events, of course, are important and newsworthy, but how can they inform an investment strategy and generate actionable signals for investors? How can we separate the signal from the noise and contextualize each current news event within a framework that has stood the test of time?
When we’re looking at stocks, we make sense of the options by using ranking models based on cross-sectional analysis. We look back at what factors (size, value, etc.) would have ranked stocks effectively in the past and then look at which stocks best fit those criteria today.
But cross-sectional analysis is not very well suited to parsing narratives and understanding how asset classes are priced relative to economic conditions. To understand whether a signal is robust, we look not only for whether the signal produces high or low returns in a given asset class but also for the range of potential outcomes. We want to make sure that the signal produces non-overlapping confidence intervals, which is a more difficult test of the signal’s predictive strength.
The best way to understand this test is visually, with examples. Compare, for example, how changes in credit spreads predict the future returns of 10-year Treasurys (left-hand side of Figure 1) to how changes in the Federal Reserve rate predict the future returns of 10-year Treasurys (right-hand side of Figure 1).
Figure 1: 1M FWD 10Y Treasury Real Returns by Credit Spreads (LHS) and Changes in Federal Reserve Rates (RHS), 1975–2022
Source: Bloomberg, FRED
Looking at the chart on the left, the top end of the distribution of returns when spreads are falling is below the bottom end of the distribution of returns when spreads are rising. This gives us high confidence that changes in credit spreads are a good predictor of future treasury returns. Treasurys do seem to rally when credit spreads deteriorate. Looking at the chart on the right, we see a significant overlap in the distribution of returns—changes in Federal Reserve rates don’t seem to provide a robust signal at all, contrary to popular opinion. We shouldn’t sell our Treasurys just because the Fed is hiking.
We have found that the level and direction of credit spreads are a powerful macroeconomic indicator. Below, we show a table that looks at stocks (divided by factor), bonds (corporate and government, divided by duration and credit rating), and commodities from 1975 to 2022 and whether credit spreads produced non-overlapping confidence intervals in predicting one-month future returns. The green highlights indicate assets for which a given indicator has been reliably predictive at the one-month horizon.
Figure 2: Non-Overlapping Confidence Intervals of 1M FWD Real Returns by Asset and Predictor Variable, 1975–2022
Source: Ken French Data Library, Thomson Reuters Datastream, S&P Capital IQ, Bloomberg, FRED
Credit spreads produced non-overlapping confidence intervals in predicting future returns for two-thirds of the asset classes we tested—a remarkable degree of statistical power.
We wanted to highlight a few of these relationships so you can understand why we spend so much time talking about credit spreads. Below we show the distribution of returns to the Fama-French value factor relative to changes in credit spreads.
Figure 3: 1M FWD Fama-French Value Factor Real Returns by Spread Change, 1975–2022
Source: Ken French Data Library, FRED
The value factor works well when spreads are falling and poorly when spreads are rising. The cheap cyclical stocks captured by the value factor are heavily impacted by changes in credit conditions. Notably, this is the opposite pattern shown by 10-year Treasurys, which is why the two asset classes serve as good complements in a business-cycle based strategy.
We can see another interesting relationship with the Fama-French size factor. The size factor delivers higher one-month forward real returns when credit spreads are above their 10-year median.
Figure 4: 1M FWD Fama-French Size Factor Real Returns by Spread Level, 1975–2022
Source: Ken French Data Library, FRED
The size premium appears to work better when the price of liquidity is more costly, when spreads are above median, and when liquidity is widely available, the returns to investing in small-cap stocks are lower. These confidence interval tests are confirming basic economic intuition.
We find running these types of analyses to be essential to separating real signals from noise. On Google trends, “interest rates” has ~90x more search interest than “credit spreads.” For all the mindshare that interest rate hikes are currently taking up, we don’t see much predictive signal in this noise. So, just as Warren Buffett cuts wheat from chaff by reading annual reports in the bathtub, we give closest attention to credit spreads to gauge the relative attractiveness of asset classes.