The Facts About Factor Timing

Episode Summary

Executive Summary: The episode examines factor timing/rotation as a way to switch among factor ETFs using valuation, momentum, macro, and composite signals. The hosts note the academic evidence is mixed: Rob Arnott’s value-based timing found some excess return but weaker risk-adjusted results, while Cliff Asness/AQR argue factor timing is difficult and favor momentum. Their own factor-rotation models use 15 Invesco ETFs, monthly rebalancing, and a trend overlay, with momentum and composite approaches appearing strongest in their backtests.

Main Topics: The case for and against factor timing (Priority: 5/5): The hosts frame factor timing as intuitive—buy what is out of favor and rotate into what works—but emphasize that academic evidence is mixed and implementation is difficult. Academic research: Arnott vs. Asness/AQR (Priority: 5/5): They compare Rob Arnott’s value-timing work, which found value beat equal-weight but not on a risk-adjusted basis, with AQR’s work arguing contrarian timing is difficult and factor momentum is more promising. Factor selection universe and ETF structure (Priority: 4/5): Their implementation begins with a 15-ETF universe spanning small, mid, and large-cap versions of value, growth, momentum, quality, and low volatility, used as the starting point for rotation portfolios. Value timing using spreads and composites (Priority: 5/5): Their value model uses valuation spreads between the best and worst stocks within each factor, combined into a composite valuation score rather than relying on a single metric like price-to-book. Momentum timing across multiple lookback periods (Priority: 5/5): The momentum model ranks ETFs by a composite of several momentum horizons, not just one period, and allocates to the top five ETFs with the strongest relative momentum. Macro-based factor rotation (Priority: 4/5): A more top-down model maps factor attractiveness to macro conditions such as GDP growth, inflation, the yield curve, and credit spreads, assigning factors to cycle quadrants and adding bonus signals. Composite model and implementation details (Priority: 4/5): They also combine value, momentum, and macro into a composite ranking system, rebalance monthly, and run the strategies with a trend-following overlay to manage exposure.

Key Arguments: Factor timing is appealing in theory because investors can buy factors when they are out of favor, but it is hard to execute consistently in real markets. Academic research does not clearly endorse factor timing; results differ depending on whether one uses value-based timing or momentum-based timing. Rob Arnott’s approach suggested value timing can outperform an equal-weighted factor mix in raw returns, but not necessarily on a risk-adjusted basis. Cliff Asness and AQR argue that contrarian factor timing is deceptively difficult, while factor momentum may be a more robust way to rotate among factors. Using spreads and composites is preferable to relying on a single valuation metric such as price-to-book, which can be flawed or distorted over time. Momentum is simpler to implement than value timing because it relies on ranking relative performance rather than estimating valuation spreads. Macro timing is the weakest of the three approaches academically, but it may still add value in a composite system by capturing cycle-dependent relationships. A composite of value, momentum, and macro may be more robust than any single timing signal because it diversifies model risk. Their real-time portfolios are designed to test these ideas rather than claim definitive proof; the goal is evidence-based experimentation and ongoing tracking.

Data Points: Number of factor ETFs in universe: 15 ETFs - The starting selection universe includes small, mid, and large-cap ETFs for each factor. Factors included: 6 factor groupings - Size, value, growth, momentum, quality, and low volatility are used in the rotation framework. ETF selections held in portfolio: Top 5 ETFs - Each model whittles the 15-ETF universe down to five holdings based on its ranking method. Value timing paper year: 2016 (approx.) - Rob Arnott’s ‘Timing Smart Beta Strategies, of course, buy low, sell high’ is referenced as a key research paper. Performance conclusion on value timing: Excess return but worse Sharpe ratio - Arnott’s value-timing approach beat equal-weighted factors on raw return but not on a risk-adjusted basis. Backtest start year: 2006 - The hosts present backtested data for the factor-rotation approaches starting in 2006. Rebalance frequency: Monthly / 28-day cycle - The factor rotation portfolios are reviewed and rebalanced on a monthly schedule. Momentum lookback: 12 minus 1 and related periods - The momentum model uses multiple momentum horizons including a 12-month-minus-1-month style measure. Macro indicators: GDP growth, inflation, yield curve - These are examples of the top-down macro inputs used to assess factor attractiveness. Macro bonus signal: High-yield spread widening - A wide high-yield spread can add bonus points to the macro score and has historically favored value.

Pivotal Quotes: "factor timing is challenging" — Jack: Summarizing the central difficulty of trying to rotate among factors based on relative performance. "contrarian factor timing is deceptively difficult" — Jack (quoting Asness/AQR research): Describing the academic skepticism around timing factors by buying what is out of favor. "momentum was the best. And then the second one has been the idea of using all of them together" — Jack: Summarizing their backtest results across the different factor-rotation models since 2006.

Implications: For investors, factor timing may sound attractive but is hard to sustain and can underperform for long stretches. The episode suggests momentum and composite models may be more robust than pure value timing, but discipline, testing, and caution are essential.