Time-series momentum, commonly known as trend following, has been a cornerstone strategy in quantitative finance, attracting substantial interest for its potential to capture persistent price movements across various asset classes. This strategy exploits the tendency of asset prices to continue moving in the same direction, leveraging serial correlations in returns to generate profits. Commodity Trading Advisors (CTAs) and managed futures funds, managing hundreds of billions of dollars collectively, often employ systematic trend-following strategies across equities, commodities, government bonds, and currencies.

In their seminal paper, “Demystifying Time-Series Momentum Strategies: Volatility Estimators, Trading Rules, and Pairwise Correlations,” Nick Baltas and Robert Kosowski delve deep into the mechanics of time-series momentum strategies. They aim to enhance the understanding and implementation of these strategies by examining the impact of volatility estimation methods, trading rule designs, and the role of asset correlations.

A central focus of Baltas and Kosowski’s research is the refinement of volatility estimators and trading rules to improve the efficiency and profitability of time-series momentum strategies. Traditional implementations often suffer from excessive portfolio turnover, leading to higher transaction costs that erode net returns.

The authors propose using the Yang and Zhang (2000) range-based volatility estimator, which incorporates open, high, low, and close prices to provide a more accurate measure of an asset’s true volatility. By capturing more information about price movements within a trading period, this estimator reduces noise and the need for frequent rebalancing. Empirical results show that utilizing the Yang-Zhang estimator reduces portfolio turnover by approximately 17% without a statistically significant impact on performance.

In terms of trading rules, Baltas and Kosowski compare the standard binary approach, which signals a full long or short position based on the sign of past returns, with a more nuanced “TREND” rule. The TREND rule scales positions according to the statistical strength of the observed price trends, reducing exposure to assets with weaker signals. This approach smooths the transition between positions and further reduces portfolio turnover by about 24%, again without compromising performance.

Beyond volatility estimators and trading rules, the authors highlight the critical role of pairwise correlations between assets in a time-series momentum portfolio. Standard implementations often neglect the correlation structure, which can lead to suboptimal risk management, especially during periods of heightened market stress when asset correlations tend to increase.

Baltas and Kosowski introduce a correlation-adjusted variant of the time-series momentum strategy. This model dynamically adjusts the leverage applied to each asset based on the average pairwise signed correlation within the portfolio. By reducing leverage during periods of high average correlation, the strategy aims to mitigate the risk of significant drawdowns.

The figure presents the three-month raw (Panel A) and signed (Panel B) average pairwise correlation of the available contracts at the end of each month; 24-month moving averages of these estimates are superimposed. The sample period is from January 1984 to February 2013.

Figure 1 illustrates how average pairwise correlations have fluctuated over time, emphasizing periods where correlations spike, such as during financial crises. Recognizing and adjusting for these correlation patterns is essential for managing portfolio risk effectively.

To assess the practical benefits of their proposed refinements, Baltas and Kosowski conduct a thorough evaluation of the strategies, accounting for transaction costs. They construct a transaction cost model that distinguishes between roll-over costs (incurred when futures contracts approach maturity) and rebalancing costs (associated with strategy turnover).

Their findings reveal that the traditional time-series momentum strategy incurs trading costs amounting to approximately 10% of its gross returns over a 30-year period. By implementing the Yang-Zhang volatility estimator and the TREND trading rule, these costs can be reduced by about 35%. The correlation-adjusted strategy further enhances risk-adjusted returns, particularly during periods of high market stress.

The table presents performance statistics for correlation-adjusted time-series momentum strategies that differ in the momentum trading rule (SIGN vs. TREND) and the volatility estimator used (SD vs. YZ). The correlation factor is estimated using the average past three-month pairwise signed correlation. For comparison, the first column contains performance statistics for the benchmark strategy with no correlation adjustment. Panel A covers January 1984 to February 2013, and Panel B covers January 2009 to February 2013.

Table 2 compares the performance metrics of various strategies, demonstrating that incorporating the correlation adjustment improves Sharpe ratios and other risk-adjusted return measures. Notably, the TREND rule combined with the Yang-Zhang estimator and correlation adjustment yields superior performance with reduced turnover.

The insights from Baltas and Kosowski’s research hold significant implications for return stacking strategies, which seek to enhance diversification and risk-adjusted returns by combining multiple uncorrelated or lowly correlated return streams within a single portfolio. By refining volatility estimation and trading rules, and by accounting for asset correlations, investors can construct more efficient portfolios that align with the objectives of return stacking.

Implementing these refined time-series momentum strategies can complement other alternative investment strategies, such as those discussed in Trend Following and Portable Alpha. For instance, a trend-following component that adjusts for volatility and correlations can be stacked with other strategies like merger arbitrage or carry to achieve a more robust and diversified return profile.

Moreover, incorporating correlation adjustments echoes the principles of Diversification 2.0, where understanding and managing inter-asset relationships are paramount. By dynamically adjusting exposures based on the correlation environment, investors can better navigate periods of market stress, enhancing the resilience of return stacked portfolios.

Baltas and Kosowski’s “Demystifying Time-Series Momentum Strategies” offers valuable contributions to the field of quantitative investing. By rigorously analyzing the effects of volatility estimators, trading rules, and asset correlations, they present practical methods to enhance the performance and efficiency of time-series momentum strategies.

Their findings underscore the importance of refining traditional approaches to reduce transaction costs and manage risks effectively. For practitioners and investors engaged in return stacking, these insights provide a framework for constructing more resilient and diversified portfolios, ultimately aiming for superior risk-adjusted returns in complex financial markets.