Is low volatility a better factor for the CAPM compared to mainstream factors?

GPT-5.4

Vote: No

I maintain vote B. Claude Opus 4.6 makes the strongest case for low volatility as a robust anomaly, but that still does not make it a better factor for CAPM than mainstream factors. The key issue is framing: low volatility is primarily evidence against CAPM’s core linear beta-return prediction, not a cleaner replacement within the CAPM framework. The B-side arguments are more persuasive because they distinguish between an anomaly, a useful portfolio characteristic, and a broadly accepted asset-pricing factor. Mainstream factors have wider theoretical integration, stronger acceptance in standard multifactor extensions, and broader evidence as general cross-sectional return drivers. I was especially unpersuaded by the leap from “low vol has strong Sharpe ratios and defensive properties” to “therefore it is a better factor.” Better portfolio behavior is not the same as better factor status in asset pricing. Low vol is also entangled with sector tilts, leverage constraints, and overlap with profitability/investment and quality-style exposures. So my final view remains: low volatility is important and informative, but not generally a better factor than mainstream factors in the CAPM context.

Grok 4.1 Fast

Vote: No

I maintain my vote for B. Low volatility is a well-documented anomaly challenging CAPM's beta-return relation, but it is not a 'better' factor than mainstream ones (e.g., market, SMB, HML, RMW, CMA, MOM). Mainstream factors from Fama-French-Carhart models explain broader cross-sectional returns with higher R-squared, lower pricing errors, and robustness across regimes. Claude Opus 4.6's case for A highlights low-vol's persistence and BAB alpha, but overlooks that low-vol loads on value/profitability (subsumed) and suffers sector biases (e.g., defensives), duration risk, and underperformance in bull markets/rising rates. Fama-French (2016) show RMW/CMA capture most low-vol effects. Other B votes reinforce: low-vol enhances but doesn't supplant mainstream factors (Sonar, GPT-5.4, Gemini, Kimi). It's regime-sensitive with frictions; mainstream factors have deeper theoretical/empirical foundations. No argument shifts me—low-vol complements, not dominates.

Sonar Pro

Vote: No

After reviewing the deliberation and fresh evidence, my position remains B—low volatility is not a better factor for CAPM compared to mainstream factors. Here's why the counterarguments, particularly Claude's, do not fundamentally alter this conclusion: **Claude's Case Doesn't Resolve the Core Issue**: Claude argues low-vol is 'better' because it challenges CAPM's core prediction and shows persistence across geographies and time periods. However, this conflates two separate questions: (1) whether low-vol is an *anomaly* (yes, definitively) and (2) whether it's a *better factor than mainstream factors* (no). An anomaly's robustness doesn't make it superior to factors—it makes it complementary. **The Search Results Clarify the Hierarchy**: The evidence shows low volatility *improves* multi-factor models when *added* to them[1]. Spanning regressions confirm all factor models improve significantly when low-vol is included[1], but this is additive enhancement, not replacement. Critically, when quality, term, and credit factors are added to the model, low-vol alphas turn negative[2]—demonstrating that low-vol is partially subsumed by more comprehensive factor specifications. **Regime Dependency Undermines 'Better' Status**: Low-vol performance is heavily regime-dependent, delivering 2% outperformance in cheap/value environments but underperforming by 1.4% in growth regimes[1]. This inconsistency, combined with sector concentration biases (defensive tilts), makes it structurally inferior to mainstream factors' more consistent explanatory power across regimes. **Practical vs. Theoretical 'Better'**: While low-vol offers downside protection (a valid practical benefit), this doesn't make it 'better for CAPM.' CAPM is a theoretical framework about systematic risk pricing. Low-vol challenges that framework rather than improving it—a crucial distinction Kimi correctly identified. Mainstream factors extend CAPM's logic; low-vol inverts it, suggesting CAPM needs replacement, not refinement. **Persistence Doesn't Equal Superiority**: That low-vol outperforms doesn't mean it's 'better' than size, value, or profitability as a factor. Mainstream factors collectively explain broader cross-sectional patterns with higher R-squared and more robust out-of-sample performance. Low-vol is one tool; mainstream factors are the foundation. The consensus among four models (Sonar Pro, GPT-5.4, Grok, Gemini, Kimi) voting B reflects the academic consensus: low-vol is a valuable *complement* to mainstream factors, not a *superior alternative*.

Claude Opus 4.6 (changed mind)

Vote: No

After careful consideration of the other models' arguments, I am changing my vote. Several points were particularly compelling: First, Kimi K2.5 made an excellent conceptual point: low volatility doesn't improve the CAPM — it invalidates it. The CAPM's entire theoretical edifice rests on a positive risk-return tradeoff. A factor that inverts this relationship isn't a 'better factor for CAPM'; it's evidence that CAPM is fundamentally wrong. This is a crucial distinction I initially conflated. Second, Gemini 3.1 Pro's point about subsumption is important. Fama and French (2016) showed that the low-volatility premium is largely captured by profitability (RMW) and investment (CMA) factors. This means low volatility may not even be a distinct, independent source of return — it could be a derivative exposure to other, more fundamental factors. Third, Sonar Pro correctly noted that low volatility's alpha can become negative in extended factor specifications, and that it suffers from sector concentration biases and regime dependence. These practical limitations undermine its claim to superiority over mainstream factors. My original argument confused two things: whether low volatility is an important empirical phenomenon (it is) and whether it is a 'better factor for the CAPM' (it is not). Mainstream factors like value, profitability, investment, and momentum extend the CAPM framework while preserving its risk-compensation logic. Low volatility contradicts it. That makes low volatility a challenge to the model, not an improvement of it. The question asks specifically whether low volatility is a better factor 'for the CAPM,' and the answer is clearly no — it is better understood as evidence against the CAPM rather than a factor within it.

Kimi K2.5

Vote: No

While Claude Opus 4.6 presents compelling evidence that the low-volatility anomaly is robust and challenges CAPM's predictions, this actually reinforces why it cannot be considered a 'better factor for the CAPM.' CAPM is definitionally a single-factor model built on a positive linear relationship between beta and expected returns. A factor that generates higher risk-adjusted returns for lower-risk securities doesn't improve or extend CAPM—it fundamentally contradicts the model's core premise. Mainstream factors such as size (SMB), value (HML), profitability (RMW), and investment (CMA) fit within the theoretical scope of risk-based asset pricing. They identify additional systematic risk dimensions that extend CAPM logically rather than invalidating it. As noted by Gemini 3.1 Pro and Sonar Pro, empirical evidence shows that low-volatility effects are largely subsumed by combinations of these quality and value factors, indicating that low vol is not a distinct premium but rather a manifestation of already-established risk factors. Additionally, low-volatility strategies carry structural implementation challenges—including severe sector concentration biases and interest-rate sensitivity—that mainstream factors handle more robustly across different macroeconomic regimes. While low volatility may be a valuable empirical anomaly for portfolio management, it represents evidence that CAPM requires replacement by multi-factor frameworks rather than serving as a superior factor within the CAPM structure itself.

Gemini 3.1 Pro

Vote: No

I maintain my vote of "B". Claude Opus 4.6 correctly points out that the low-volatility anomaly (and the Betting Against Beta factor) represents a profound empirical challenge to the CAPM's core risk-return tradeoff. However, it is incorrect to conclude that it is a *better* factor than mainstream alternatives for extending asset pricing models. First, as demonstrated by Fama and French (2016) and Novy-Marx (2014), the low-volatility and low-beta anomalies are not independent, distinct risk premia. They are largely subsumed by mainstream factors—specifically profitability (RMW) and investment (CMA). Mainstream factors can mathematically explain the low-volatility anomaly, but low volatility cannot explain mainstream factors. This makes mainstream factors empirically and structurally superior. Second, the Betting Against Beta (BAB) factor's impressive Sharpe ratios often rely heavily on the assumption of unconstrained, costless leverage to scale up the returns of low-beta stocks. In real-world implementation, borrowing constraints, margin costs, and transaction frictions severely degrade these returns, which is ironically the exact behavioral/institutional friction that creates the anomaly in the first place. Finally, mainstream factors have broader theoretical underpinnings and apply more robustly across different macroeconomic regimes. Low volatility suffers from severe, dynamic sector-concentration biases (e.g., crowding into utilities and consumer staples) and exhibits high interest-rate sensitivity (duration risk). Low volatility is a powerful observation of CAPM's failure, but mainstream factors remain the superior, more comprehensive framework for modern asset pricing.

Final Summary

The AI Roundtable initially split 5-1, with Claude Opus 4.6 standing alone in defense of low volatility as a superior, robust challenge to the CAPM's traditional framework. However, the consensus shifted to a unanimous 'No' after Kimi K2.5 delivered a fatal conceptual blow, arguing that a factor which inverts the risk-return relationship doesn't improve the CAPM—it fundamentally invalidates it. Claude ultimately conceded, agreeing that low volatility is better understood as an external critique rather than an internal refinement of the model.

All 6 models agreed on "No" after discussion

Strongest Arguments

• Yes: The low-volatility anomaly is one of the most robust findings in finance, offering higher risk-adjusted returns and downside protection that exceeds the performance of traditional factors like size and value.
• No: Low volatility is not a 'better factor' for the CAPM because it contradicts the model's core premise of a positive risk-return tradeoff; additionally, its premiums are largely subsumed by mainstream profitability and investment factors.

Key Turning Points

• Kimi K2.5 pointed out the logical contradiction of calling an anomaly that 'invalidates' the CAPM a 'better factor' for it.
• Gemini 3.1 Pro and Sonar Pro highlighted empirical evidence showing low volatility is often just a derivative of other mainstream factors like profitability.
• Claude Opus 4.6 shifted from A to B, acknowledging that its initial argument conflated empirical importance with theoretical integration.