Analyzing AMOM ETF: A Deeper Dive into AI-Enhanced US Large Cap Momentum

Let's delve into last week's analysis where the performance of two ETFs was compared. One that seeks to follow the SP500 (SPY) and another that is designed primarily with AI (AMOM). The QRAFT AI-Enhanced US Large Cap Momentum ETF is managed by Qraft Technologies . QRAFT Technologies is an investment management company that uses artificial intelligence and advanced technology to create investment strategies based on quantitative and qualitative factors. This ETF focuses on large-cap stocks in the United States and uses artificial intelligence algorithms to identify investment opportunities based on market momentum.

Last week I participated in an event organized by VettaFi where Francis Geeseok OH explained the performance of AMOM. They were aligned with my previous one, so to deepen the analysis, we will make some changes. Let's expand the time horizon from 12 to 24 months and use the Adjusted values and not the Closing ones. Now we see what happens with the Volatility of each one, the Return and the possible Future of Volatility.

AMOM ETF Analysis:

1. Mean Returns: The estimated constant term (mu) in the mean model is very close to zero. This suggests that, on average, the AMOM ETF does not exhibit a significant positive or negative return trend. The negative autoregressive coefficient (ar1) suggests that there is a tendency for the ETF to reverse course in the short term. In other words, periods of positive returns are likely to be followed by periods of negative returns and vice versa. The positive moving average coefficient (ma1) indicates that past return shocks tend to persist, meaning that positive (negative) shocks are followed by positive (negative) returns.
2. Volatility: The GJR-GARCH(1,1) model is used to capture volatility dynamics. The estimated parameters suggest that volatility exhibits persistence (beta1 is approximately 0.912), meaning that past volatility influences current volatility. The leverage effect (gamma1) is positive (approximately 0.132), indicating that negative shocks to returns have a slightly stronger impact on increasing volatility compared to positive shocks of the same magnitude. This is consistent with the typical behavior of financial markets.
3. Model Fit: The model seems to fit the AMOM ETF data reasonably well, as indicated by the log-likelihood and information criteria (BIC and Hannan-Quinn). Lower values of these criteria suggest a better fit of the model to the data.
4. Serial Correlation: The weighted Ljung-Box test results suggest that there is no significant serial correlation in the standardized residuals. This implies that the model adequately captures the autocorrelation structure in the ETF's returns.

Overall, the analysis indicates that the GJR-GARCH(1,1) model with an ARFIMA(1,0,1) mean model is a reasonable choice for modeling the AMOM ETF returns. It suggests that the ETF's returns exhibit short-term mean-reverting behavior and volatility clustering. However, as with any financial analysis, it's essential to consider the broader economic and market context, as well as the specific investment objectives, risk tolerance, and time horizon of your investment in the AMOM ETF. Additionally, this model assumes that historical patterns will continue into the future, which may not always be the case in financial markets.

Comparison with SPY ETF Analysis:

1. Mean Returns: The estimated constant term (mu) for the SPY ETF is very close to zero (0.000174), similar to the AMOM ETF. This suggests that, on average, there is no significant positive or negative return trend. The autoregressive coefficient (ar1) for SPY is approximately -0.892, while it was -0.905 for AMOM. Both ETFs exhibit negative autocorrelation in returns, meaning that periods of positive returns are likely to be followed by periods of negative returns and vice versa. The moving average coefficient (ma1) for SPY is approximately 0.922, similar to AMOM. This indicates that past return shocks tend to persist for both ETFs.
2. Volatility: The GJR-GARCH parameters for SPY are generally similar to those for AMOM. The persistence of volatility (beta1) for SPY is approximately 0.930, slightly higher than for AMOM (0.912). This suggests that past volatility has a relatively stronger impact on current volatility for SPY. The leverage effect (gamma1) for SPY is approximately 0.119, similar to AMOM (0.132). Both ETFs exhibit a slight asymmetry in volatility response to positive and negative shocks.
3. Model Fit and Serial Correlation: The log-likelihood and information criteria for the SPY ETF indicate a reasonably good fit, similar to AMOM. The weighted Ljung-Box test results for both ETFs show no significant serial correlation in the standardized residuals. This suggests that both models adequately capture the autocorrelation structure in the ETF returns.

Comparisons:

• The mean return characteristics of both ETFs are quite similar, with small constant terms and negative autocorrelation in returns.
• Both ETFs exhibit volatility clustering and a similar leverage effect.
• The model fits for both ETFs are comparable, as indicated by log-likelihood and information criteria.
• The absence of significant serial correlation in the standardized residuals is a common feature for both ETFs.

My (2 cents) Takeaway: Overall, the GJR-GARCH(1,1) model with an ARFIMA(1,0,1) mean model seems to provide reasonable fits for both the AMOM ETF and the SPY ETF. Both ETFs display similar patterns in mean returns and volatility dynamics. However, the specific characteristics and risk-return profiles of the two ETFs should still be considered when making investment decisions, as they may differ in terms of sector exposure, investment strategy, and underlying assets.

In the long term (24 months) the profitability is similar between the 2 ETFs, but the volatility of AMOM is greater. In the short term (12 months) AMOM's profitability is significantly higher. Some of the possible explanations may be the advance in AI and the better access to alternative data that exists on the market today.

Note: this analysis was carried out with public information and only as an example of didactic financial analysis. It is not a formal or in-depth assessment.