The US yield curve has (almost) inverted, and this has been making headlines for the last couple of months now. This should come as no surprise, as the yield curve is perhaps the most reliable recession indicator out there. But what does an inverted yield curve tell us about future returns? Our analysis shows that while asset class returns in general are somewhat subdued between the first date on which the yield curve inverts and the start of the recession, the inversion of the yield curve is not followed by extraordinary deviations in returns.
Before moving over to the results of our analysis, we would like to dwell briefly on the definition of the yield curve, and the combination of maturities in particular. In most empirical research, the yield curve is either defined by the differential between the 10-year and 3-month US Treasury yield (10Y-3M), or the 10-year and 2-year US Treasury yield (10Y-2Y). The reasons for preferring one over the other depends on many things, including data availability – the 3-month US Treasury yield has a much longer history; the degree to which you want to capture short-term versus long-term views on GDP growth and inflation, which is likely to be better reflected in the 2-year yield; and/or forecasting accuracy and timeliness, a point we will get back to later. As noticed by the New York Fed in its study ‘The Yield Curve as a Leading Indicator: Some Practical Issues (2006): “Spreads based on any of the rates mentioned are highly correlated with one another and may be used to predict recessions.” Hence, we will look at both the 10Y-3M and 10Y-2Y inverted yield curves, also because this leads to at least one interesting observation.
Inversions and recessions
As mentioned, the yield curve qualifies as one of the best, if not the best, recession forecasters. For the 10Y-2Y yield curve, we have reliable data covering the last five US recessions, all of which were accurately forecasted well in advance, as shown in the right panel in the table above. The lag between the first ‘inversion date’ and the start of the recession, as determined by the National Bureau of Economic Research (NBER), averages 21 months, ranging from 11 months until the 1981 recession to 34 months until the 2001 recession. The results for the 10Y-3M yield curve, as shown in the left panel in the table above, are highly comparable, with an average lead time of 19 months until the next recession. The data further reveals that prior to the last five recessions, the 10Y-2Y yield curve inverted before the 10M-3M yield curve on each occasion. From this angle, the 10Y-2Y yield curve should be the preferred recession indicator, as it ‘detects’ the next recession first.
The available data history for the 10Y-3M yield curve is longer, covering the last seven recessions. We find that the 10Y-3M yield curve correctly predicts these two additional recessions (1970, 1973) as well. However, it also seems to have given a false signal. On 12 January 1966, the 10Y-3M yield curve inverted for six days, but the next recession did not start until January 1970, or four years later. Obviously, the time horizon for which to assign forecasting power is arbitrary, but four years is considerably longer than in other cases. In addition, between early 1967 and December 1968, the 10Y-3M yield curve did not invert once, suggesting that we are looking at a separate period of yield curve inversion. Unfortunately, we can’t compare these inversions with the 10Y-2Y yield curve, due to a lack of data. Therefore, we will focus on the last five recessions, for which we have data on both the 10Y-3M and 10Y-2Y yield curve, for the remainder of this analysis.
Inversions and asset class returns
So, what does yield curve inversion tell us about (future) asset class returns? The table above shows the average and median annual returns on most major asset classes, US stocks, global stocks, commodities, gold, US Treasuries and US corporates, as well as US real GDP growth for both yield curves. The returns are calculated as the index change between the first negative reading of the yield curve leading up to a recession, and the first day of that same recession. In short, it calculates the performance between the inversion date and the start of the recession. The last row of the table shows the average annual return for the full sample period, from August 1978 until 1 January 2008. As can be derived from the table, this period was an exceptionally strong period for both stocks and bonds, with average annual returns above their longer-term history.
We will now summarize our main findings. First, while there are differences between the returns calculated using the 10Y-3M and 10Y-2Y yield curve, the results are highly comparable. Choosing either yield curve does not lead to different conclusions. Second, while variation in returns is substantial, they are far from extreme. For example, the average and median annual return on all asset classes is positive. No asset class shows severe and structural weakness after inversion, with only gold realizing a negative return in three out of the five inversion periods. But, as the table shows, gold returns are pretty erratic in any case. At the same time, none of the asset classes – again apart from gold – realized an extraordinarily high average return either. Having said that, for all asset classes the average annual return between yield curve inversion and recession was lower than for the full sample, except for commodities.
The deviation from the full sample average return is relatively large for US corporate bonds. For both yield curves, the average annual return after inversion was significantly below 3%, against a full sample return of 8.9%. This observation fits the perception that credits tend to struggle late cycle, as short-term interest rates are lifted by the Federal Reserve and leverage tends to rise. Global stock performance also trails between yield curve inversion and recessions: the average annual return is less than half than the full sample return. This can be explained by the defensive nature of US stock markets, and the fact that most other regions are highly dependent on the US economy, given their ‘openness’. It is a well-known maxim that when the US sneezes, the rest of the world catches a cold.
Lastly, with a 7% return, commodities are the only asset class which realized a much better return than the full sample average (2.8%) after yield curve inversion. This fits the characterization of commodities as ‘being late cycle.’ As final demand increases during economic expansion, so too does the demand for commodities. Hence, since raw materials are needed to produce goods now, the forward-looking aspect is likely to be of lesser importance than it is for equities and bonds.
A word on growth
Before moving over to the final part of this analysis, a quick word on growth. As is shown in the final column of the table above, average real GDP growth between yield curve inversion and the start of the recession is very close to, and even slightly above, the average of the full sample. This implies there is no such thing as a gradual cooling of the economy before slipping into recession. This helps explain why forecasting recessions is incredibly hard. Just ask the IMF, which has not been able to predict even half of the recessions just months before they started.
Has the yield curve inverted?
The yield curve, be it either the rate difference between the 10-year Treasury yield and the 3-month or 2-year yield, has a strong track record in predicting recessions. But has it inverted? Out of the last five recessions, the 10Y-2Y yield curve was always the first to signal a recession. This time, however, the 10Y-3M yield curve briefly inverted in late March, while the 10Y-2Y yield curve did not. While one should refrain from arguing that ‘this time is different’ as much as possible, the fact that the 10Y-3M curve inverted first makes this case different by definition. Quantitative easing followed by quantitative tightening (balance sheet reduction) could perhaps explain this divergent sequence, providing a potential argument why this yield curve inversion ‘doesn’t count’. But there were compelling reasons (a savings glut, structural budget surpluses) before to explain why yield curve inversions should not precede a recession. When looking at both yield curves and their forecasting history, it’s simply impossible to say if a recession signal has been given.
Still, as we believe it is possible to establish that we are in the later stages of the economic cycle, it could prove prudent to become somewhat less enthusiastic about the return prospects of corporate bonds (as reflected in our multi-asset portfolios) and be a bit more optimistic about those of commodities. To be continued…