Factor Investing: Value, Momentum, Quality

1. What you'll learn

• What factor investing is and how value, momentum, and quality factors are defined
• How to calculate common factor signals step by step (EV/EBIT, 12-1 momentum, ROE, and more)
• How to combine multiple signals using z-scores, ranks, and composite scores
• How professionals handle data cleaning, winsorization, and neutralization (sector, beta, size)
• How to build, rebalance, and evaluate a factor-based portfolio with real numbers
• The role of transaction costs, turnover, capacity, crowding, and timing
• Common misconceptions and how to avoid them in practice

2. Concept explanation

Factor investing is a rules-based approach that selects and weights securities based on measurable characteristics called factors. Instead of relying on a narrative about a single company, you build exposure to traits that historically have been linked to higher risk-adjusted returns. Three of the most studied factors are value, momentum, and quality.

• Value seeks stocks that are inexpensive relative to fundamentals, such as earnings or cash flow. The idea is simple: pay less than something is plausibly worth. In practice, value often uses ratios like Price/Book, EV/EBIT, or Price/Free Cash Flow.
• Momentum looks for stocks that have been winning recently, on the premise that trends can persist due to behavioral and structural reasons. A common measure is the past 12 months' return excluding the most recent month (12-1), to avoid short-term reversal effects.
• Quality aims for companies with strong and sustainable business performance: high profitability, conservative leverage, and earnings stability. Metrics include Return on Equity (ROE), gross profitability, low accruals, and interest coverage.

Together, these factors can complement each other. Value can lag when cheap companies are cheap for a reason, but momentum often does well in trending markets. Quality can mitigate value traps by filtering for financially sound firms. Combining them helps smooth the ride.

3. Why it matters

Academic research and practitioner results show that factor premiums have been persistent across decades, geographies, and different datasets. While nothing is guaranteed, allocating to factors has offered investors a middle ground between pure indexing and discretionary stock picking. It is systematic, transparent, and testable.

For individual investors, factor investing is accessible via rules you can implement yourself or through ETFs and mutual funds that target specific factors. The key is knowing how signals are constructed, cleaned, and combined—because these details drive real-world outcomes including performance, risk, and costs.

Professional implementation also involves controlling unintended risks. For example, a value screen might inadvertently load on small, distressed companies; a momentum portfolio might concentrate in one hot sector. Advanced methods—like z-score standardization, sector-neutral ranking, and beta/size neutralization—aim to isolate the factor you want while limiting what you do not.

4. Calculation method

Below are commonly used formulas and practical steps. We will standardize signals using z-scores so different metrics can be combined into a single composite.

• Value signals (higher is better once we invert the ratios):

  • Earnings yield via EV/EBIT
  • Free cash flow yield (FCF/Price)
  • Book-to-Price (B/P)

• Momentum signal:

  • 12-1 total return (sum of monthly returns over months 2 to 12, excluding the most recent month)

• Quality signals:

  • ROE
  • Gross profitability (Gross Profit/Total Assets)
  • Accruals (lower better)

Step 1: Compute raw metrics per stock.

• EV/EBIT:

  EV = MarketCap + TotalDebt - Cash EarningsYield = EBIT / EV

  Higher EarningsYield implies better value.

• Free cash flow yield:

  FCFYield = FreeCashFlow / MarketCap

• Book-to-Price:

  BtoP = BookValueOfEquity / MarketCap

• Momentum (12-1):

  Mom12-1 = Product(1 + monthlyReturn_t) from t = -12 to -2 minus 1

• Quality metrics:

  ROE = NetIncome / AverageShareholders'Equity GrossProfitability = (Revenue - COGS) / TotalAssets Accruals = (NetIncome - FreeCashFlow) / TotalAssets

Step 2: Clean the data.

• Winsorize outliers, e.g., cap each metric at the 1st and 99th percentiles.
• Handle missing or negative denominators sensibly (e.g., exclude firms with negative equity for B/P; set EV minimum to avoid division by near-zero).
• Define a minimum liquidity threshold (e.g., median daily dollar volume) to avoid illiquid names.

Step 3: Standardize via z-scores by cross-section (e.g., within your investment universe or within sectors).

For metrics where higher is better (EarningsYield, FCFYield, BtoP, Mom12-1, ROE, GrossProfitability), use z as-is. For metrics where lower is better (Accruals), invert.

Step 4: Combine related signals into factor composites.

• Value composite:

  Val_i = average( z(EarningsYield)_i, z(FCFYield)_i, z(BtoP)_i )

• Momentum composite:

  Mom_i = z(Mom12-1)_i

• Quality composite:

  Qual_i = average( z(ROE)_i, z(GrossProfitability)_i, z(AccrualsInverted)_i )

Step 5: Build a multi-factor score (equal-weighted or optimized).

A common starting point is equal weights: w_V = w_M = w_Q = 1/3.

Step 6: Portfolio construction.

• Rank stocks by MF_i and select the top decile or quintile.
• Apply sector-neutral selection: pick top names within each sector to match benchmark sector weights.
• Weighting: equal-weight or tilt by MF_i (subject to caps). A simple approach is equal-weight within the selected set.

Step 7: Risk controls and neutralization.

• Beta neutrality: regress returns on the market to estimate beta and adjust weights to target portfolio beta near 1 (long-only) or 0 (long-short). beta_i = Cov(r_i, r_m) / Var(r_m)
• Size neutrality: match the benchmark size distribution (e.g., via market-cap buckets).
• Turnover control: use buffers so a stock leaves only if it falls below a threshold rank.

5. Case study (practical example with real numbers)

Suppose your universe has 6 stocks: A, B, C, D, E, F. You compute the following raw metrics:

• EarningsYield: [A: 7%, B: 4%, C: 10%, D: 6%, E: 2%, F: 8%]
• FCFYield: [A: 3%, B: 1%, C: 5%, D: 4%, E: -1%, F: 2%]
• BtoP: [A: 0.6, B: 0.4, C: 0.9, D: 0.5, E: 0.2, F: 0.7]
• Mom12-1: [A: 14%, B: -3%, C: 28%, D: 10%, E: 5%, F: 18%]
• ROE: [A: 12%, B: 6%, C: 20%, D: 10%, E: 3%, F: 15%]
• GrossProfitability: [A: 24%, B: 18%, C: 30%, D: 22%, E: 12%, F: 28%]
• Accruals: [A: 4%, B: 8%, C: 1%, D: 5%, E: 9%, F: 2%] (lower is better)

Step 1: Convert each list to z-scores across the 6 stocks. For illustration, assume z-scores (rounded):

• z(EarningsYield): [A: 0.08, B: -1.21, C: 1.30, D: -0.23, E: -2.14, F: 0.20]
• z(FCFYield): [A: 0.08, B: -0.88, C: 1.03, D: 0.60, E: -1.54, F: 0.71]
• z(BtoP): [A: 0.00, B: -0.87, C: 1.53, D: -0.44, E: -1.75, F: 0.53]
• z(Mom12-1): [A: 0.24, B: -1.18, C: 1.37, D: -0.08, E: -0.70, F: 0.35]
• z(ROE): [A: 0.08, B: -0.88, C: 1.37, D: -0.24, E: -1.42, F: 1.08]
• z(GrossProfitability): [A: -0.13, B: -1.06, C: 0.80, D: -0.40, E: -2.13, F: 0.93]
• z(Accruals): lower is better; invert after z. Suppose raw z: [A: -0.15, B: 1.20, C: -1.35, D: 0.25, E: 1.50, F: -1.45]. After inversion (multiply by -1): [A: 0.15, B: -1.20, C: 1.35, D: -0.25, E: -1.50, F: 1.45]

Step 2: Compute composites.

• Value Val_i = average of z(EarningsYield), z(FCFYield), z(BtoP)

  • A: (0.08 + 0.08 + 0.00)/3 = 0.05
  • B: (-1.21 - 0.88 - 0.87)/3 = -0.99
  • C: (1.30 + 1.03 + 1.53)/3 = 1.29
  • D: (-0.23 + 0.60 - 0.44)/3 = -0.02
  • E: (-2.14 - 1.54 - 1.75)/3 = -1.81
  • F: (0.20 + 0.71 + 0.53)/3 = 0.48

• Momentum Mom_i = z(Mom12-1)

  • A: 0.24, B: -1.18, C: 1.37, D: -0.08, E: -0.70, F: 0.35

• Quality Qual_i = average of z(ROE), z(GrossProfitability), inverted z(Accruals)

  • A: (0.08 - 0.13 + 0.15)/3 = 0.03
  • B: (-0.88 - 1.06 - 1.20)/3 = -1.05
  • C: (1.37 + 0.80 + 1.35)/3 = 1.17
  • D: (-0.24 - 0.40 - 0.25)/3 = -0.30
  • E: (-1.42 - 2.13 - 1.50)/3 = -1.68
  • F: (1.08 + 0.93 + 1.45)/3 = 1.15

Step 3: Multi-factor score MF_i with equal weights (1/3 each).

• A: (0.05 + 0.24 + 0.03)/3 = 0.11
• B: (-0.99 - 1.18 - 1.05)/3 = -1.07
• C: (1.29 + 1.37 + 1.17)/3 = 1.28
• D: (-0.02 - 0.08 - 0.30)/3 = -0.13
• E: (-1.81 - 0.70 - 1.68)/3 = -1.40
• F: (0.48 + 0.35 + 1.15)/3 = 0.66

Ranking by MF_i: C (1.28), F (0.66), A (0.11), D (-0.13), B (-1.07), E (-1.40).

Portfolio choice (top 3): C, F, A. Equal-weight each at ~33.3%. Apply sector checks to keep sector weights similar to your benchmark; if C and F are in the same sector, consider substituting A with the next-best name in a different sector to diversify.

6. Practical applications

• Long-only factor portfolio: Select top 20-30% by MF_i, equal-weight, rebalance quarterly with 20% buffer (a name only exits if it falls below the 40th percentile). This reduces turnover while keeping exposure.
• Long-short factor pairs: Long top decile, short bottom decile within sectors to isolate factor alpha and reduce market beta. Use dollar or beta neutrality.
• ETF selection: If you prefer funds, choose value, momentum, and quality ETFs with clear methodologies, low fees, and sufficient liquidity. Check whether they use sector neutrality, rebalancing frequency, and how they handle transaction costs.
• Rebalancing cadence: Value signals are slower-moving than momentum; monthly momentum refresh with quarterly value/quality updates can balance freshness and costs.
• Position sizing: Cap individual positions (e.g., 5%) and sectors (e.g., ±5% relative to benchmark). Consider volatility scaling: scale weights by the inverse of each stock's recent volatility so no single name dominates risk.
• Risk measurement: Track tracking error, information ratio, and factor exposures via a risk model. Monitor drawdowns and rolling 3-year relative performance to understand cyclicality.

7. Common misconceptions

8. Advanced considerations

• Neutralization and orthogonalization: To isolate pure factor effects, regress each signal on size and beta, then use residuals as cleaned signals. For example, for value, regress EarningsYield on log(MarketCap) and beta; take the residual as size/beta-neutral value. EarningsYield_clean = EarningsYield - (a + b1 * log(MarketCap) + b2 * beta)
• Sector and industry structure: Implement rankings within GICS sectors or industries to prevent persistent sector bets. Then re-aggregate to portfolio level with sector constraints.
• Transaction costs and turnover: Use an optimizer that maximizes MF score subject to cost penalties and constraints. Include impact models based on average daily volume. Maximize: Sum(w_i * MF_i) - lambda * TradingCosts
• Capacity and crowding: Factor strategies can become crowded, compressing spreads and amplifying drawdowns. Monitor liquidity and your strategy's share of average daily dollar volume to avoid excessive market impact.
• Evaluation metrics: Beyond raw return, track information ratio, hit rate, t-stats of factor returns, and factor drawdowns. Use out-of-sample and walk-forward tests.

9. Summary

Glossary

Factor: A measurable characteristic of securities linked to differences in returns, such as value, momentum, or quality.

Z-score: A standardized value showing how many standard deviations a data point is from the mean.

Momentum (12-1): The total return over the past 12 months excluding the most recent month.

Earnings Yield: EBIT divided by enterprise value, the inverse of EV/EBIT.

Winsorization: Limiting extreme data points to reduce the effect of outliers.

Neutralization: Adjusting signals or weights to remove unwanted exposures like sector, size, or market beta.

Information Ratio: Active return divided by tracking error, measuring risk-adjusted performance relative to a benchmark.