In this update, we’ll explore an approach for using inversions of the yield curve as a market timing tool; this approach has historically delivered higher investment returns at a lower risk, relative to a similar strategy without such a “yield curve overlay."
Previous Strategy: Leaving Gains on the Table
During an earlier post, I outlined a strategy – based on selling the S&P 500 when it falls below 95% of its 200-day moving average, then buying it back when it rises above 103% of the 200-day moving average – which has delivered superior risk-adjusted returns relative to a “buy and hold” approach. However, this strategy would have left us out of the market on 23 occasions since 1950 – and in 16 of these cases, the S&P 500 went up.
Ideally, we’d like to harvest some of these gains, while still protecting ourselves against the most punishing losses that can decimate a portfolio. Is there a market timing tool that would make this possible?
The Yield Curve to the Rescue
An inverted yield curve would therefore appear to represent a useful timing tool that might help us to avoid the market declines that occur in conjunction with recessions, while participating in more of the appreciation that accompanies economic expansions.
Let’s see how we can translate this into a decision rule that will enable more lucrative investing.
A More Profitable Decision Rule
If we use the yield curve as a market timing tool, we can modify our aforementioned decision rule to the following:
If in cash, buy S&P 500 at close when it has risen above 103% of its 200-day moving average; if invested, liquidate S&P 500 at close when each of the following conditions are met: (a) the S&P 500 has fallen below 95% of its 200-day moving average; (b) the 2 Year/10 Year Treasury yield curve has inverted since the end of the previous recession; and (c) the next recession has not yet been confirmed.
Selling after the yield curve has inverted but only if the next recession has not yet been confirmed is a play on the old investment adage “buy (or in this case, sell) on the rumor, sell (buy) on the news”.
If you are not interested in slogging through the nuts and bolts involved in implementing the modified decision rule, please feel free to skip ahead to the section titled “Implications of Modified Decision Rule”.
NBER Offers Us an Assist
An important criterion for evaluating our modified decision rule is that it needs to be applicable contemporaneously, i.e. during the “fog of war” and without the benefit of hindsight. In this case, condition (b) above requires that we know when the previous recession ended, and condition (c) requires that we know the next recession has not yet begun. But how can we be sure about this, when there is typically considerable uncertainty at turning points in the economy?
Fortunately, the National Bureau of Economic Research (NBER) offers us an assist here. Starting with its July 26, 1979 meeting, the NBER began making contemporaneous recession and recovery calls. We’ll use the NBER announcements as the basis for implementing our decision rule. If the yield curve has not inverted since the NBER announcement of a cyclical trough (i.e. end of the previous recession), we will ignore any sell signals that our moving average strategy generates; but if the yield curve has inverted since the NBER announcement of a trough, we will only ignore any sell signals that our strategy generates if the NBER has announced a peak (i.e. beginning of the next recession) in the meantime.
The amount of time between the end of a recession and the NBER’s subsequent announcement of a trough can be quite lengthy – in one case, a period of 21 months. In contrast, the NBER has tended to be timelier in its recession calls it has with its recovery calls – we see below that the longest amount of time between the beginning of a recession and the NBER’s announcement of a peak is 12 months. Based on the NBER announcements, we would have liquidated our holdings in the following cases where a sell signal was generated by the moving average strategy:
On the flip side, our modified decision rule would have led us to ignore sell signals in five cases:
Operating without NBER Business Cycle Dating
To validate our decision rule, we also need to be able to operationalize it for those business cycles which predated the NBER’s contemporaneous recession and recovery calls. Fortunately, we would not have encountered much ambiguity in implementing our decision rule, given the timing of the sell signals generated during these periods.
We could have confidently liquidated our holdings in the following cases where a sell signal was generated following a yield curve inversion, as this sell signal occurred before the next recession would have been confirmed, and well after the end of the preceding recession: Conversely, we could have confidently ignored the following sell signals, since we would have had confirmation that a recession had begun since the most recent yield curve inversion:
Implications of Modified Decision Rule
This decision rule would have enabled us to stay invested during the following periods in which the use of a moving average strategy without a “yield curve overlay” would have had us go to cash:
The use of a “yield curve overlay” on our moving average strategy would still have kept us out of the market during the following periods3:
Our “yield curve overlay” provides us with a significant improvement in performance, relative to our previous decision rule without such an overlay.
Topline Performance Measures
Here are the characteristics of our strategy with a “yield curve overlay”, compared with (a) this strategy without the overlay, and (b) a “buy and hold” approach. To keep things simple, we will assume that no returns are earned during periods when the strategy is out of the market. All figures exclude dividends and cover the period from October 1950-May 2018: The performance advantage of the Moving Average Strategy with a Yield Curve Overlay is even more striking on a risk-adjusted basis. As shown above, this strategy produces a Standard Deviation of 12.5%. To match this Standard Deviation, a Buy and Hold approach would need to be hedged with a 24% cash component, reducing the annualized nominal return of this approach to 5.7%, which is nearly 300 basis points less than that produced by our Moving Average Strategy with a Yield Curve Overlay.
Hedging Our Bets
An inversion of the yield curve has been an uncanny predictor of forthcoming recessions since the 1950’s – boasting a nine-for-nine mark with no false positives. However, it is always possible that it may falter in the future. For example, perhaps we will have a recession that is not anticipated by a yield curve inversion, which might leave us invested during a significant market downturn if we followed the approach outlined above. If this is a concern, we might hedge our bets by modifying our decision rule to the following:
If in cash, buy S&P 500 at close when it has risen above 103% of its 200-day moving average; if invested, liquidate S&P 500 at close when each of the following conditions are met: (a) the S&P 500 has fallen below 95% of its 200 day moving average; (b) the 2 Year/10 Year Treasury yield curve has inverted since the end of the previous recession; and (c) the next recession has not yet been confirmed. When (a) but not both (b) and (c) have been met, liquidate half of S&P holdings until the next “buy signal” but remain invested with the other half.
Compared with the “unhedged” approach, this modified strategy yields an annualized return (8.3% from October 1950-May 2018) that is slightly lower; the standard deviation of annualized returns (12.5%) is the same; but the maximum drawdown (-24.8% vs. -33.5%) is lower; while the average number of trades per year is higher (0.7 vs. 0.4).
One Final Wrinkle
Since yield curve inversions have anticipated recessions, and stock market performance has faltered in conjunction with recessions, you may be wondering whether we would be better advised to liquidate upon the first yield curve inversion following the end of the previous recession, and not wait for the sell signal to occur. Here is what the numbers show:
On average, the S&P has declined between an inversion of the yield curve and the next sell signal – which seems to suggest that it would be better to liquidate as soon as the yield curve inverts. However, during recent business cycles – which have been longer in duration – the performance of the S&P between the yield curve inversion and the sell signal trigger has generally been more favorable. And it is possible that the yield curve may invert in the future without a recession taking place, which would leave us vulnerable to missing out on significant gains if we did not remain invested in the aftermath of the inversion – so I feel that the more prudent course of action is to wait for a sell signal to be generated before liquidating.
Planning for the Journey that Lies Ahead
Knowing the range of possibilities that exist in the road ahead should make it easier for us to navigate with confidence during any turbulent times we may encounter – so in my next update, I will share case studies that provide graphical representations as to how the stock market and interest rates have performed in the wake of previous yield curve inversions, in hopes that the lessons of history will leave us better prepared for the future.
1 Treasury rate data comes from the FRED website. I imputed daily rates going back to January 1954 (importantly, January 1954 was the effective date of a policy that allowed U.S. Treasury interest rates to rise and fall with the market) where FRED did not provide them by using a combination of weighted interpolation (a technique that applies a pattern from a relevant daily dataset to decompose the monthly averages provided by FRED into daily totals – more details available upon request) and correlation (regression using relevant daily rates as independent variables in cases where monthly averages for the dependent variable were not available): 2 Although the March 24, 1980 sell signal took place several months after the commencement of the January 1980-July 1980 recession, this recession was not confirmed by the NBER until June 3, 1980 – so the criteria for liquidating our holdings would have been met. Similarly, we would have liquidated in response to the sell signal generated on September 23, 1957. Although this sell signal took place a couple of months after the commencement of the August 1957-April 1958 recession, we almost certainly would not have had confirmation of this fact in time to ignore the sell signal.
3 Our data on Treasury rates does not go far enough back to enable us to identify whether/when the yield curve inverted in advance of the July 1953-May 1954 recession; to be conservative, we will assume that an inversion preceded the June 9, 1953 breach of the S&P’s 200-day moving average, and consequently we would have been in cash until January 6, 1954, a period during which the S&P appreciated.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.