Factor investing

Definition

Factor investing holds systematic portfolios tilted toward characteristics with documented historical return premia — value (cheap vs expensive), size (small vs large), momentum, quality (profitability/investment), and low volatility being the canon. It replaces stock-picking with characteristic-picking: the claim is not that any stock wins, but that the tilted basket earns a premium over time. Every equity strategy in this KB decomposes partly into factor exposures, which makes this entry double as the platform’s attribution reference.

How it works / structure

• Factor definitions: value (book-to-market and successors), size (market cap), momentum (12-1 return — strategy-momentum), profitability/investment (the Fama-French five-factor extensions), low-beta/low-vol.
• Construction: rank universe on the characteristic; long-short deciles (academic form) or long-only tilts (implementable form — much weaker exposure); rebalance on schedule.
• Parameters (engine-executable): factor(s), signal definition per factor, universe, long-only vs long-short, holding count/weights, rebalance frequency, multi-factor blending rule (integrated scores vs sleeves).
• Attribution role: replayed strategies are regressed against factor benchmarks; residual (alpha) vs loading (beta) is the platform’s skill/exposure split (port-exposure-netting).

When it applies

Long-horizon systematic allocations expecting multi-year premium realization; attribution and benchmark construction for other strategies; tilts within a core allocation (port-allocation-frameworks). Explicitly NOT a short-horizon signal — factor premia arrive irregularly with decade-scale droughts.

Risk profile & failure modes

• Factor droughts are the price: value underperformed growth for roughly a decade into 2020 — a documented, painful stretch; premium harvest requires surviving droughts longer than most conviction lasts.
• The factor zoo: Harvey-Liu-Zhu (2016) catalogued hundreds of published “factors” and argued most fail appropriate multiple-testing thresholds — the canon above is the defensible core, not the zoo.
• Post-publication decay: McLean-Pontiff (2016) measured substantial attenuation after publication; live premia are smaller than backtests.
• Implementation gap: long-only tilts deliver a fraction of academic long-short spreads; fees, turnover, and crowding take shares.
• Definition sensitivity: “value” by different metrics produces materially different portfolios — specification risk is factor risk.

Evidence & limits

Fama-French (1993) established the framework and documented size and value premia in US data; subsequent literature extended, challenged, and partially eroded the estimates. This is the most scrutinized empirical territory in finance: the honest summary is that a small factor canon has robust long-sample historical support, decayed live magnitudes, and no delivery schedule. Marketing that presents factor premia as reliable annual income is folklore.

Falsifiable-thesis examples

Illustrations only, not signals:

• “A long-only value tilt on universe U will outperform the cap-weighted universe over the next 3 years” — falsified at horizon.
• “Strategy S’s replay return this year will show statistically positive residual after regression on the five-factor benchmark” — falsified by the attribution run.

Cross-references

• Factor with its own entry: strategy-momentum; benchmark baseline: strategy-buy-and-hold
• Method discipline: lens-quantitative (multiple testing, decay — the factor zoo is its case study)
• Fundamental inputs: fa-ratio-analysis, fa-financial-statements
• Portfolio use: port-allocation-frameworks, port-exposure-netting