Since March of last year, the US Federal Reserve has raised interest rates in order to cool inflation, with the current range set at 5.25-5.5%. [1] During these times, bond value can deteriorate as they inherently carry interest rate risk. Long-term bonds are generally more sensitive to interest rate fluctuations than short-term bonds, leading to increased volatility. As such, recent rate hikes have highlighted the importance of dynamically managing risk in volatile markets. Subsequently, active management of fixed income portfolios has gained popularity as a strategy to enhance returns through proactive and flexible decision-making. [2]

As shown in [Figure 1], the maximum drawdown (MDD) for US bonds has dropped below the level of US equities in recent years. These trends pose a problem for investors as bonds have traditionally been considered safeguards against the sharp downsides of equities. To address this issue, Qraft AI-Powered Dynamic Fixed Income applies artificial intelligence (AI) to dynamically manage duration and credit risks.

[Figure 1] MDD Comparison for US Equities vs. US Bonds (28 Mar 2008 – 30 Sep 2023)

* SPDR S&P 500 ETF Trust (SPY)
** iShares Core US Aggregate Bond ETF (AGG)
Source: Qraft Technologies, Refinitiv

Method Behind The Machine

Qraft AI-Powered Dynamic Fixed Income seeks to exceed the return of a US Aggregate Bond Index through a hierarchical portfolio construction approach that combines two proprietary AI signals: US Bond Duration Risk Signal and Credit Risk Signal. AI models generating each signal utilize Qraft’s proprietary API, KIRIN API, to pre-process vast amounts of data and build bias (survivorship, look-ahead)-free environments for model training and back-testing simulation.

[Figure 2] Qraft AI-Powered Dynamic Fixed Income Model Structure

(1) US Bond Duration Risk Signal

The US Bond Duration Risk Signal manages duration risk across different maturities and optimizes allocation among short-, medium-, and long-term US treasuries. It incorporates macro data such as interest rates, inflation rates, and economic growth indicators, considering their impact on each asset class. Additionally, it utilizes various technical data including momentum, volatility, and correlations to gain insights into historical behavior and interrelationships among different asset classes. By leveraging deep neural networks and the KIRIN API, it can extract patterns from diverse market situations and make more informed decisions.

(2) Credit Risk Signal

The Credit Risk Signal assesses credit market risk and determines the duration allocation between treasury and credit. Depending on the risk signal level, it allows up to 10% allocation of high-yield bonds as a satellite strategy. In bullish market conditions, the model opts to allocate investment-grade bonds, while in bearish times, it favors treasury bonds. The model incorporates both macro data for the bond market and technical data for equity market risk, enabling a comprehensive analysis of market dynamics and risks. Machine learning techniques enhance the model's ability to learn from diverse nonlinear patterns, uncover complex relationships, and adapt to changing market conditions.

Performance Simulation Review

In comparison to the iShares Core US Aggregate Bond ETF (Ticker: AGG), which passively tracks the Bloomberg US Aggregate Bond Index, the Dynamic Fixed Income recorded a higher cumulative return of 22.63% (vs. 2.06%) as of 3Q23. The total active return for the observed period was 20.57%. [Figure 3] exhibits greater cumulative return for the Dynamic Fixed Income over the timeline, supported by the higher yearly returns of the portfolio revealed in [Figure 4].

[Figure 3] Investment Growth Since 2017 (3 Jan 2017 – 30 Sep 2023

Source: Qraft Technologies, Refinitiv
Disclosure: The presented performance represents hypothetical back-tested results during the measured time period. Hypothetical past performance does not guarantee future results. Trading cost of Qraft AI-Powered Dynamic Fixed Income is 10bp.

[Figure 4] Yearly Returns Since 2017 (3 Jan 2017 – 30 Sep 2023)

[Figure 5] demonstrates the true nature of dynamic risk management by the portfolio. Drawdowns have been generally minimized by optimizing asset allocation, especially during the recent interest rate hikes by the Federal Reserve. During such severe periods, reducing drawdowns is essential for controlling downside volatility.

[Figure 5] Investment Drawdown since 2017 (3 Jan 2017 – 30 Sep 2023)

As shown in [Figure 6], Qraft AI-Powered Dynamic Fixed Income outperformed AGG in all metrics. The portfolio’s dynamic risk management helped reduce volatility to 5.31% (vs. 5.48%) which in turn inferred a higher Sharpe ratio of 0.58 (vs. 0.08). Also, a smaller MDD of -14.37% (vs. -18.44%) over the period lent to the model’s ability to manage downside risk.

[Figure 6] Portfolio Performance Metrics vs. Benchmark (3 Jan 2017 – 30 Sep 2023)

Performance Analysis During Major Market Events

The active return attribution of the AI-Powered Dynamic Fixed Income can be broken down into two effects: Allocation and Selection. The Allocation and Selection Effects respectively represent the contribution from the US Bond Duration Risk Signal and the Credit Risk Signal.

The portfolio has gradually increased its active return to 20.57% as of 3Q23. [Figure 7] reveals the rising trend of active return during different market events, highlighting the portfolio’s capability to potentially outperform the benchmark and its evolving nature.

[Figure 7] Total Active Return Since 2017 (3 Jan 2017 – 30 Sep 2023)

The Dynamic Fixed Income outperformed the benchmark in the 2022 Bond Bear Market (31 Dec 2021 – 30 Dec 2022), producing a higher active return of 4%. Both effects contributed positively as illustrated in [Figure 8], with the Selection Effect at 2.53% and the Allocation Effect at 1.47%.

During the global outbreak of COVID-19 (28 Feb 2020 – 31 March 2020), the portfolio produced 1.75% in active return. In this case, the Selection Effect contributed 1.81% in active return, while the Allocation Effect offset by -0.05%.

However, the US-China Trade War in 2018 (28 Sep 2018 – 31 Dec 2018) saw the portfolio endure a negative active return of -1.27%, revealed by the deeper drawdowns than the benchmark shown above in [Figure 5].

[Figure 8] Active Return Attribution During Major Financial Market Events

Source: Qraft Technologies, Bloomberg*
Disclosure: The presented performance represents hypothetical back-tested results during the measured time period. Hypothetical past performance does not guarantee future results. Trading cost of Qraft AI-Powered Dynamic Fixed Income is 10bp.
*See Appendix

Looking Ahead

Qraft’s AI-Powered Dynamic Fixed Income can appeal to investors who seek dynamic risk management along with consistent active return inside a fixed income portfolio. Currently, “higher for longer” interest rates are the prevailing belief in the market. The view is backed by the stickiness of the US inflation, demonstrated by the recent larger-than-expected rise in Consumer Price Index (CPI). [3] Hence, duration risk from potential rate hikes and increased government debt issuance could lead to additional volatility. [4] Credit risk may potentially grow as a result of gloomy markets hinted by S&P 500 reports of consecutive declines in year-over-year profit margin. [5]

In response, the AI-Powered Dynamic Fixed Income actively manages duration and credit risk for US bonds. It optimizes allocation using Qraft’s AI duration and credit risk signals, enabling investors to navigate unprecedentedly volatile market conditions.

Appendix

US-China Trade War in 2018 (28 Sep 2018 – 31 Dec 2018) Bloomberg Attribution Analysis

Global outbreak of COVID-19 (28 Feb 2020 – 31 March 2020) Bloomberg Attribution Analysis

2022 Bond Bear Market (31 Dec 2021 – 30 Dec 2022) Bloomberg Attribution Analysis

[1] Duggan, Wayne. “September Fed Meeting: Interest Rates Are Put on Hold, One More Rate Hike Likely.” Forbes, Forbes Magazine, 20 Sept. 2023, www.forbes.com/advisor/investing/fomc-meeting-federal-reserve/

[2] “Investing: Active Bond ETFs.” Fidelity, Fidelity Viewpoints, 15 Aug. 2023, www.fidelity.com/learning-center/trading-investing/bond-investing-active-etfs

[3] "US CPI Report September 2023: Live News on Inflation, Consumer Price Index." Bloomberg, 12 Oct. 2023, www.bloomberg.com/news/live-blog/2023-10-12/us-cpi-for-september

[4] "Market Outlook & Investment Strategy (October 2023)." Bank of China (Hong Kong) Limited, www.bochk.com/dam/investment/invmktcommentary/monthly/monthly_en.pdf

[5] Butters, John. "S&P 500 Reporting A Lower Year-Over-Year Net Profit Margin For The 6th Straight Quarter." FactSet, 24 Jul. 2023, https://insight.factset.com/sp-500-reporting-a-lower-year-over-year-net-profit-margin-for-the-6th-straight-quarter

Leveraging AI For Dynamic Risk Management in the Volatile Bond Market

Related Articles

Search for More Articles