October 10, 2023

- In our quest to find out if an investor can successfully navigate US recessionary periods, we construct systematic allocation strategies - easily implemented with ETFs - that use our estimated recession probabilities from our machine learning model.

- We find that a few straightforward strategies could potentially help investors steer through difficult macroeconomic environments. Adding a trend-following filter or diversifying across both macro and momentum strategies seems advisable. For the best strategy - in a horse race back-test since 1999 - we reach an 8.9% return versus 5.5% for the 60/40 benchmark with more limited drawdowns of 22% versus 31%.

For investors without a data-driven approach, it’s probably safe and wise to do absolutely nothing and just ride it out. First of all, calling recessions is quite difficult. Secondly, we know from the past that even if a recession occurs, the damage might not be that large at all. Finally, when are you ever going to get back in? Relying on the official NBER publication dates to make allocation decisions might be severely damaging to your investment portfolio. Our previous blog post discusses these topics in more detail.

We devise a purely data-driven methodology that beats the NBER approach. We implement a rule-based framework that will guide us through recessionary and potentially dangerous periods.

We present a straightforward systematic tactical allocation model that uses our out-of-sample recession probabilities for further evaluation purposes. As we have written in our previous blog post: it’s one thing to predict recessions, it’s a whole other challenge to honestly estimate how the market will react. History taught us that recessions can even be accompanied by positive equity performance. Imagine that.

To compare as honestly as feasible we construct five systematic allocation models that we will compare to a 60/40 benchmark of 60% MSCI USA Mid & Large Caps & 40% US Treasuries. All trading signals occur at the end of the trading month and returns are shifted correctly to avoid any data leakage.

**Recession-only strategy:**

This strategy only acts on the recession signal and goes defensive as soon as the recession model warns of potential danger. We define “as soon as” as a probability of 10 % or higher. We could have easily chosen another (better) threshold as the thresholds seem robust between 0% and 100% (see main results). The slightest warning is enough for us to trigger a response. We have 2 scenarios:

- Invest in US equities if the probability of recession is low at < 10%.
- Invest in Treasuries when the probability of recession is >= 10%.

**Momentum-only strategy:**

This strategy only tracks a momentum signal and goes defensive as soon as momentum on equities turns negative. Momentum is measured using a composite momentum score computed as the simple average of 1,3,6 and 12 months of momentum (rate of change). So we have 2 scenarios as well:

- Invest in US equities as long as composite momentum is positive.
- Invest in US Treasuries when composite momentum is negative.

**Recession + momentum filter strategy:**

This strategy stays invested in US equities as long as the probability of recession stays below 10%. The strategy switches to momentum-modus when the probability of recession crosses above 10%. The recession signal thus acts as a first warning signal. We have 4 possible scenarios. We are invested in equities when at least recession is < 10% or when momentum is positive. Else we are long treasuries.

- Invest in US equities when the probability of recession is < 10% and composite momentum is positive.
- Invest in US equities when the probability of recession is < 10% and composite momentum is negative.
- Invest in US equities when the probability of recession is >= 10% and composite momentum is positive.
- Invest in US Treasuries when the probability of recession is >= 10% and composite momentum is negative.

**Combined signal strategy:**

This strategy only stays invested in US equities when two criteria are met: The probability of recession has to be low at below 10% and the composite momentum on equities has to be positive. The strategy switches to Treasuries in all other scenarios. So we have 4 scenarios again here:

- Invest in US equities when the probability of recession is < 10% and composite momentum is positive.
- Invest in US Treasuries when the probability of recession is < 10% and composite momentum is negative.
- Invest in US Treasuries when the probability of recession is >= 10% and composite momentum is positive.
- Invest in US Treasuries when the probability of recession is >= 10% and composite momentum is negative.

**50% recession-only strategy + 50% momentum-only strategy**

This strategy invests in both strategies at the same time allocating 50% of capital to each every month.

- The results of our back-test on these rule-based strategies are shown in the figure and tables below.

- The recession-only strategy beats the buy-and-hold benchmark on return (8.99% versus 5.5%) and on a risk-adjusted basis (0.84 Sortino versus 0.58). The strategy seems well-placed to avoid big carnage from large recessions such as 2008. However, when large bear markets occur without real recession signals, the strategy will suffer along with equities. This is what happened in the 2000 - 2003 bear market where the official recession was limited in time and only occurred between 2001-04 and 2001-11. The rest of the bear market was mainly the consequence of the implosion of the bubble of the nineties. Momentum-only strategies usually perform much better in these environments than macroeconomic-focused strategies.

- The backtest uses 10% as the main recession probability threshold. We could have easily chosen another threshold. The last table below shows the risk-performance results of the recession-only strategy for different levels of recession probability thresholds. We believe the results to be more than robust for parameter variation.

- Adding a momentum condition to the recession-only strategy did not lead to any improvement. Importantly, however, it did not worsen the strategy either! So using a trend filter might still not be a bad idea. The out-of-sample sample we tested (1999 - 2023) was too small to show the benefits of trend-following during recessionary periods. None of the 3 recessions during this period had positive equity returns, a condition on which a momentum rule could keep an investor in the market even though the probability of recession is high. Numerous academic studies and empirical research (for example here and here) have consistently demonstrated he benefits of using momentum conditions in these environments.

- Magic mainly happens when we combine signals or strategies. Both the combined signal strategy and the equal weight strategy deliver Sortino ratios of 1. The drawdowns are much more limited as well which results in higher MAR ratios. This once again shows that it can pay to diversify across signals and strategies when applying systematic models in practice.

- In our quest to find out if an investor can successfully navigate US recessionary periods, we constructed systematic allocation strategies that use our estimated recession probabilities from our machine learning model. This topic was explored in our previous blog post.

- We find that a recession-only strategy that uses just the recession signal beats the buy-and-hold benchmark on a risk-adjusted basis. Adding a momentum condition (trend filter) - that can help an investor stay invested during recessions with positive equity regimes - did not change the results of this strategy. However, our sample did not contain these regimes. It’s therefore difficult to point to the advantages of trend-following during challenging economic times.

- We still favor adding a trend filter as most research on longer time frames and other markets shows clear advantages of switching to trend-modus during volatile macroeconomic periods.

- Combing a recession-only strategy with a momentum-only strategy reveals the advantages of diversification. A very risk-averse investor might implement the combined signal strategy which stays invested only when both the recession and the momentum signal call for a risk-on environment. In any other scenario, the strategy stays risk-off.

- We only used treasuries as a risk-off asset in this back-est. These strategies might potentially benefit if other forms of risk-off baskets are explored. For example: adding other risk-off assets such as T-bills, Tips, and strategies that switch between risk-free assets.

We found inspiration for the use of concepts and predictors in the paper below. Importantly, however, this blog post does not replicate the paper!

Spyridon D. Vrontos, John Galakis, Ioannis D. Vrontos,

Modeling and predicting U.S. recessions using machine learning techniques,

International Journal of Forecasting,

Volume 37, Issue 2,

2021,

Pages 647-671,

ISSN 0169-2070,

https://doi.org/10.1016/j.ijforecast.2020.08.005.

(https://www.sciencedirect.com/science/article/pii/S0169207020301205)