The Big 7: Style Factors
Before seeking exotic alpha, models must explain these known sources of systemic risk.
Momentum
The tendency for assets that have performed well in the recent past to continue performing well.
Economic Rationale
Behavioral Bias: Investors under-react to new information initially, then herd/over-react, creating trends.
Risk
Momentum Crashes: When trends violently reverse (e.g., 2009 recovery), momentum strategies suffer deep drawdowns.
Value
Buying assets that are cheap relative to their fundamental value.
Economic Rationale
Risk Premium: Cheap stocks often face distress risk. Returns are compensation for holding uncomfortable assets.
Risk
Value Traps: Stocks that are cheap because their business model is fundamentally broken (e.g., Kodak).
Size
The historical tendency for small-cap stocks to outperform large-cap stocks.
Economic Rationale
Information Risk: Small caps are less covered by analysts, requiring a premium for the lack of information/liquidity.
Risk
Weakening Signal: The size premium has diminished in recent decades and is highly volatile.
Low Volatility
Stocks with lower price fluctuations tend to deliver higher risk-adjusted returns.
Economic Rationale
Leverage Constraints: Institutional investors can't use leverage, so they bid up high-beta stocks, making them overpriced.
Risk
Interest Rate Risk: Low vol stocks (Utilities, Staples) often act like bonds and fall when rates rise.
Quality
Companies with stable earnings, low debt, and high profitability.
Economic Rationale
Flight to Safety: Investors pay a premium for certainty during uncertain times.
Risk
Valuation Risk: 'Quality at any price' can lead to underperformance if you overpay for safety (e.g., Nifty Fifty).
Liquidity
Illiquid assets must offer higher returns to compensate for the difficulty of selling them.
Economic Rationale
Friction Compensation: You get paid for the risk of being unable to exit a position quickly during a crisis.
Risk
Liquidity Crises: In a market panic, liquidity dries up completely, causing massive price gaps.
Growth
Companies with high expected future growth rates, often with high valuations today.
Economic Rationale
Duration: Growth stocks are 'long duration' assets. You are buying future cash flows.
Risk
Discount Rates: Highly sensitive to interest rates. If rates rise, the present value of future growth collapses.
Factor Crowding
When too much capital chases these standard factors, returns decay and crash risk increases.
The Titan Clash: Models Compared
Deep dive into the mathematical engines and strategic differences between the industry giants.
Head-to-Head Comparison
| Feature | MSCI Barra | Axioma | Northfield |
|---|---|---|---|
| Primary Technique | Cross-Sectional Regression | Hybrid (Fundamental + PCA) | Time-Series / Hybrid |
| Outlier Handling | Winsorization (Clipping) | Huber Weighting (Robust Reg) | Bayesian Shrinkage |
| Blind Spots | Risks not in factor definition | Uninterpretable Stat Factors | Complexity / Over-smoothing |
| Ideal User | Asset Managers / Long Only | Quant Hedge Funds | Multi-Asset Allocators |
| Customization | Low (Standardized) | High (Risk Model Machine) | Medium |
Academic Research & Citations
The number of discovered factors has exploded, leading to a "replication crisis." Modern research focuses on filtering out false positives.
...and the Cross-Section of Expected Returns
Harvey, Liu, Zhu (2016)
Argued that due to data mining, the threshold for significance should be raised from t-stat 2.0 to 3.0. Most 'discovered' factors are noise.
Digesting Anomalies: An Investment Approach
Hou, Xue, Zhang (2015)
Proposed the q-factor model (Investment & Profitability) which explains returns better than Fama-French's original 3 factors.
Betting Against Beta
Frazzini & Pedersen (2014)
Showed that constrained investors bid up high-beta stocks, causing them to have lower alphas. Low beta stocks are underpriced.
The New Frontier: Machine Learning
Linear regression is dying. The new wave of research uses Non-linear models to find interaction effects.
Decision Trees / Gradient Boosting
Factors are conditional. E.g., "Momentum only works when Volatility is low." Trees capture these if/then relationships.
NLP & Alternative Data
Trend: Using satellite data to count cars in retail parking lots (Revenue forecasting).
Momentum: Using Glassdoor reviews to predict "Employee Sentiment" momentum.
Inside the Black Box
How Quant Funds transform raw data into Alpha using the Optimization Workflow.
The Strategy Spectrum
Equity Market Neutral
The "Pure Alpha" Approach.
130/30 (Relaxed)
Enhanced Indexing.
Statistical Arbitrage
High Frequency Mean Reversion.
1. Alpha Model
Scores every stock in the universe based on factors.
XOM: -1.2 (Short)
TSLA: +0.1 (Flat)
2. Risk Model
Calculates the Covariance Matrix (Σ).
3. Optimizer
Solves the Mean-Variance equation.
Subject to β = 0
4. Portfolio
The final list of trades to execute.
Short $8M MSFT
Hedge Sector Risk
The Constraints Matrix
The optimizer is a constraint engine. If you don't constrain a risk, you are betting on it.
- Market Beta Constraint
βport = 0 - Sector Constraint
Σwtech = 0 - Gross Leverage
Σ|wi| ≤ 200%
Success Metric: Information Ratio (IR)
The Fundamental Law of Active Management
Correlation between your forecast and reality. A "Good" IC is only 0.05!
Number of independent bets. Quants win by making thousands of tiny bets.