This page synthesizes three practitioner surveys of factor investing: the Invesco whitepaper (Warren & Quance), the MSCI research paper (Bender, Briand, Melas & Subramanian 2013), and the Cazalet & Roncalli (2014) presentation from Lyxor/Amundi.
What is factor investing?
A factor is any characteristic relating a group of securities that is important in explaining their return and risk. Factor investing targets systematic, quantifiable attributes of securities in diversified portfolios, seeking to harvest risk premia associated with those attributes.
Invesco positions factor investing as a third pillar alongside traditional active management and passive market-cap indexing. Active management seeks alpha through security selection and unique insights. Passive indexing captures the broad market risk premium at low cost. Factor investing sits between the two: it uses rules-based, transparent methods to capture specific return drivers at fees below active management but above pure passive.
MSCI emphasizes that factor investing is distinct from market-cap indexing. Market-cap indexes represent the aggregate holdings of all investors and are the only “macro-consistent” portfolio. Factor portfolios rebalance away from market-cap weights and represent active views. Investors must form their own beliefs about whether the premium will persist.
Criteria for a true investment factor
Not every stock characteristic qualifies as an investment factor. Invesco specifies five criteria:
- Persistent — the premium exists over long time horizons and across multiple market regimes
- Pervasive — works across countries, regions, sectors, and asset classes
- Robust — holds under various definitions and measurement approaches
- Distinct — provides exposure not fully captured by other factors
- Tradeable — can be captured cost-effectively in a live, diversified portfolio
MSCI distinguishes between generic factors (useful for explaining risk) and risk premia factors (factors that earn a persistent, significant premium over long periods). For example, Growth and Liquidity factors explain risk well in the Barra models but have not earned a long-run premium. Factors like Value and Momentum qualify as risk premia factors.
Cazalet & Roncalli make a further distinction: if XMY (e.g., HML) is a risk premium, then -XMY (e.g., LMH) is not a risk premium. A risk factor explains cross-sectional return dispersion; a risk premium is the reward for bearing that factor exposure in the correct direction.
Macro vs. style factors
Macro factors relate to broad economic variables — economic growth, inflation, interest rates, currencies, financial conditions. They broadly impact prices across asset classes. Investors can access some macro factors directly (e.g., TIPS for inflation). The Two Sigma Factor Lens organizes its 18 factors around a macro-first hierarchy.
Style factors are security-level characteristics that can be targeted in investment strategies. The six canonical equity style factors are described below. Most of the factor investing discussion today focuses on style factors.
The six canonical equity factors
Value
Excess returns to securities priced at a discount relative to their fundamental value. Commonly captured by price-to-book, price-to-earnings, cash flow yield, or dividend yield. Fama and French (1992, 1993) established value (HML) as a core factor in the three-factor model. Explanations include compensation for higher distress risk (systematic risk view), loss aversion and extrapolation biases (behavioral view), and institutional flows that push prices away from fundamentals (constraint-based view). See value.
Size
Excess returns of smaller firms relative to larger ones. First documented by Banz (1981). Captured by market capitalization. Explanations include lower liquidity, less analyst coverage, and higher distress probability. The size effect has been weaker in some markets and periods since its discovery, but other factors (especially value and momentum) tend to work well within small-cap stocks, increasing its usefulness. See size.
Momentum
Excess returns to securities with stronger past performance (typically 6-12 months, excluding the most recent month). Discovered by Jegadeesh and Titman (1993). Justified primarily by behavioral biases: herding, anchoring, underreaction to information. Recent research shows earnings momentum largely subsumes price momentum. Momentum is subject to severe crashes (e.g., 2009 reversal). See momentum.
Low volatility
Securities with lower-than-average volatility or beta earn excess risk-adjusted returns. First described by Haugen and Heins (1972). The anomaly contradicts the CAPM prediction that higher risk should bring higher return. Explanations include leverage constraints (Frazzini and Pedersen 2014, the Betting-Against-Beta or BAB factor), lottery preferences, and benchmark-relative institutional mandates. Most robust evidence comes from poor returns of highly volatile stocks, not outperformance of low-vol stocks. See residual-volatility.
Quality
Excess returns to stocks with low debt, stable earnings, high profitability, and strong balance sheets. Asness, Frazzini and Pedersen (2019) formalized this as the QMJ (Quality Minus Junk) factor. Novy-Marx (2013) showed gross profitability alone is a powerful predictor. Quality is related to profitability and investment factors in the FF5 model. See quality.
Dividend yield
Excess returns to stocks with higher-than-average dividend yields. Related to but distinct from value. Tax clientele effects, investor preference for income, and systematic risk explanations have been proposed. See dividend-yield.
Why do factor premia persist?
Three schools of thought:
-
Systematic risk (efficient markets view): Factor premia compensate for bearing undiversifiable risk. Value stocks are riskier because of financial leverage and sensitivity to economic downturns. This view implies premia persist indefinitely.
-
Behavioral biases (systematic errors view): Cognitive weaknesses — loss aversion, herding, overconfidence, anchoring — create persistent mispricings. Premia persist as long as biases remain and arbitrage is costly.
-
Structural constraints: Rational investors operating under constraints (leverage limits, short horizons, benchmark mandates, regulatory rules) produce factor premia as a byproduct. Premia persist as long as constraints remain.
Most factors have multiple theories supporting them. MSCI’s Exhibit 5 maps each factor to both risk-based and error-based explanations.
Factor cyclicality
All factors are cyclical and have experienced extended periods of underperformance. There is no free lunch. Key observations:
- Risk premia are time-varying and low-frequency mean-reverting. Cycle length is 3-10 years (Cazalet & Roncalli).
- Each MSCI factor index experienced at minimum a consecutive 2-3 year underperformance period. Small cap went through a 6-year underperformance in the 1990s (MSCI).
- The explanatory power of factors beyond the market factor declined after 2003 as “beta came back” (Cazalet & Roncalli).
- Different factors perform in different economic regimes: value and size do well in recovery, quality and low-vol in downturns, momentum in sustained trends (Invesco).
- Cyclicality may itself explain why premia are not arbitraged away — short-horizon investors cannot ride through full cycles.
Diversifying across factors historically reduces the length and depth of underperformance periods.
Long/short vs. long-only factor portfolios
This distinction is critical and often misunderstood. Academic factors (SMB, HML, WML) are long/short, dollar-neutral portfolios. Practical factor investing is overwhelmingly long-only.
Cazalet & Roncalli show:
- Long-only factor portfolios (e.g., HML+, the long side of HML) are highly correlated with the market. Correlations between long-only SMB+, HML+, WML+ are 0.81-0.92 (US, 1995-2013), versus near-zero for their long/short counterparts.
- Long-only portfolios inherit substantial market risk. Bad times for the market are bad times for nearly all long-only factor portfolios. Max drawdowns for long-only 5-factor portfolios are 54-60%, similar to the market (53%).
- Long/short portfolios offer dramatically better Sharpe ratios: the global equally-weighted 5-factor long/short portfolio achieved SR 1.10 vs. 0.31 for the market (1995-2013). But long-only 5-factor portfolio SR was only 0.54.
- The Sharpe ratio of a long/short multi-factor portfolio scales approximately as sqrt(m) times the average single-factor Sharpe ratio.
Key “fantasy” debunked: long-only risk factors are not necessarily less risky than long/short factors. The long/short WML factor has extremely high risk due to momentum crashes.
Capacity constraints
Factor strategies face real-world capacity limits:
- Lesmond et al. (2004): momentum profits can be offset by trading costs.
- Korajczyk and Sadka (2004): break-even fund size for long-only momentum is $2-5 billion.
- Frazzini et al. (2012) estimated break-even sizes for global long/short strategies: SMB 190B, WML 13B.
- The issue for long-term investors is the absolute value of transaction costs, not the relative value. A strategy with 5% alpha and 1% trading costs is very different from 3% alpha and 1bp costs.
The factor zoo
Cochrane (2011) observed: “Now we have a zoo of new factors.” Harvey et al. (2014) catalogued hundreds of proposed factors. Novy-Marx (2014) showed that standard regression methods fail to reject that even presidential party affiliation, Manhattan weather, or planetary conjunctions predict returns — highlighting the data-mining problem.
The progression from CAPM (one factor) to FF3 to Carhart 4-factor to FF5 and beyond raises the “alpha puzzle”: every time a new factor is added, alpha is explained, but new alpha always re-emerges.
The consensus is that a small number of factors (roughly 5-7) are genuinely robust; the rest are likely data-mined, redundant, or spurious.
Factor investing as asset allocation
Cazalet & Roncalli frame factor investing as an asset allocation problem. In a long/short setting, optimal factor allocation reduces to mean-variance optimization on independent risk factors. In a long-only setting, the problem is harder because factor portfolios share substantial market risk, and alpha, beta, and idiosyncratic volatility of long-only factors are difficult to estimate.
Strategic allocation with risk factors is not easier than with asset classes — a key fantasy debunked by the Lyxor research.
Related pages
- overview — wiki-wide overview of factor investing
- smart-beta-disruption — how factor investing is reshaping active management
- value-momentum-interaction — negative correlation and combination benefits
- return-based-vs-holdings-based-models — factor model taxonomy
- fama-french-three-factor — original factor model
- fama-french-five-factor — expanded model with profitability and investment