Desert first

We explore a systematic bond strategy named Heine’s bond trading model. We find that it beats the benchmark on downside risk, especially during the 2022 crisis, but achieves relatively the same on other metrics. We also try a simpler trend-following system, and although it underperforms the two previous strategies, it does a good job of reducing max drawdown while being much less complex to implement.

Exploring Heine’s Bond Trading Model

Systematic equity investing typically steals the limelight with countless studies dedicated to its analysis and implementation. However, systematic bond investing often remains in the shadows, receiving less attention. Today, we aim to shed light on Heine’s bond trading model.

The strategy, first published in 1995 and developed by a money manager named Richar Heine, tries to beat the buy-and-hold of a single bond ETF by combining both price and macroeconomic signals. These are the rule of the strategy:

1. Is the Dow Jones 20 Bond Index above its 24-wk MA? Score +1 point, else 0.
2. Is the Dow Jones Utility Average above its 10-wk MA? Score +1 point, else 0.
3. Is the Long Bond Treasury Yield above its 6-wk MA? Score +1 point, else 0.
4. Is the 13-wk T-Bill Yield below its 6-wk MA? Score +1 point, else 0.
5. Is the CRB index below its 20-wk MA? Score +1 point, else 0.
6. Buy the bond ETF (i.e. AGG) if the score is 3 or higher, else stay out. 
7. Rebalance weekly.
   ‍

Note that the Dow Jones 20 Bond Index is no longer in existence; however, it is possible to utilize any bond index as a suitable substitute. For our analysis, we employed the Bloomberg US Aggregate Bond Index. Although we were unable to locate a comprehensive explanation regarding the specific rules and parameters, we have developed our interpretation of their logic.

The concept of cross-asset time series momentum suggests that previous bond market returns serve as positive indicators for future bond market returns. Conversely, past equity market returns tend to be negative predictors of future bond returns (for example, refer to this research piece). Consequently, Rule 1 and 2 align logically with this principle.

The disparity between long-term and short-term bond yields can forecast future excess returns on bonds across various time horizons (i.e. this study). Therefore, we assume that the combined influence of Rule 3 and Rule 4 captures a portion of the valuable insights derived from the term structure of interest rates.

A declining trend in the CRB index indicates a more deflationary pattern in prices, potentially bolstering bond returns to some extent. Thus, Rule 6 encompasses the assessment of general price trends.

Putting the Strategy to The Test Post-Publication

We examine whether the model was able to beat buy-and-hold after its publication date over the last 20 years. Moreover, because the model seems to use many (seemingly) arbitrary lookback periods, we suspect some over-fitting might be at play. We therefore also compare the Heine model to a much simpler model that invests when the Dow Jones 20 Bond Index is above its 24-wk MA, and stays out otherwise. We use the AGG ETF (iShares Core US Aggregate Bond ETF) during our tests.

The next figure shows the equity curves of the three systems:

The performance measures of the three strategies are outlined in the following table:

On a risk-adjusted basis, the Heine model seems to be on par with simple buy-and-hold. However, looking at downside risk, the systematic model clearly beats the benchmark by achieving a max drawdown of half the size. This is especially true during the 2022 crisis. On the other hand, it’s not clear whether the complexity of the Heine model is warranted. Although its much simpler sibling performs worse, the difference is quite small. 

Currently, we harbor reservations regarding the robustness of the trading rules employed in the Heine model. Consequently, we do not consider it suitable as a fundamental component of a fixed-income strategy. However, we do perceive potential in utilizing it as a supplementary sub-strategy. Furthermore, we believe that by incorporating accurate cross-asset time series momentum and refining the consideration of the term structure, the strategy could be simplified and potentially enhanced.

‍

‍