Four-Criteria Factor Filter
Reduce hundreds of proposed factors to a small, investable set
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
The framework evaluates a proposed investment factor against four gates before it is admitted into a portfolio or product set. First, the factor needs an economic mechanism: compensation for bearing risk, a persistent structural impediment, or a recurring behavioral bias. Second, it needs a long history that covers multiple market environments rather than a short, favorable sample. Third, its returns must be differentiated from ordinary market-cap exposure so it adds something identifiable. Fourth, it must be implementable at sufficient scale and low enough cost to preserve the theoretical benefit. Applying all four criteria converts an unruly universe of hundreds or thousands of reported anomalies into a small set of credible macro and style factors.
Origin
Andrew Ang described BlackRock's screening criteria on Masters in Business while explaining why only about half a dozen macro factors and half a dozen style factors survive scrutiny.
Core principles
- 01Demand an enduring economic rationale
- 02Prefer evidence spanning multiple decades
- 03Require returns distinct from market-cap exposure
- 04Reject strategies that cannot be implemented efficiently at scale
How to run it
- 1
Establish the economic rationale
Determine whether the return source compensates investors for risk, persists because of a structural impediment, or exploits a durable behavioral bias.
Pro tip Write the causal mechanism before reviewing the backtest so attractive results do not dictate the story.
Watch out A statistical pattern without a defensible mechanism may be data mining.
- 2
Demand a long history
Test the factor across decades and through multiple economic and market regimes.
Pro tip Include recessions, recoveries, inflation shifts, and valuation extremes where data permit.
Watch out A short history can conceal cyclicality, fragility, or overfitting.
- 3
Test differentiation
Measure whether the factor produces returns meaningfully different from market-cap benchmarks and already accepted factors.
Pro tip Examine correlations and marginal portfolio contribution rather than standalone returns alone.
Watch out A renamed market exposure does not provide a new source of diversification.
- 4
Test scalable implementation
Estimate turnover, trading costs, liquidity, capacity, and the fee required to deliver the strategy.
Pro tip Model costs under stressed liquidity as well as average conditions.
Watch out A strong paper return can disappear when implemented at realistic size.
- 5
Admit or reject the candidate
Retain only factors that pass every gate and document why each accepted factor belongs in the portfolio.
Pro tip Reassess accepted factors if their economic mechanism or implementation conditions change.
Watch out Do not weaken a gate merely to justify a preferred product or result.
In the wild
A research team finds that stocks sharing a novel accounting characteristic outperformed for eight years. The effect looks impressive, but its economic rationale is unclear, its sample covers only one market cycle, and estimated turnover is high. The team rejects it rather than adding another superficially distinct factor.
→ The portfolio avoids an expensive, potentially data-mined exposure.
Minimum volatility has an economic explanation grounded partly in institutional restrictions that push some investors toward higher-risk securities. It has extensive evidence, differs from market-cap exposure, and can be packaged at scale. It therefore survives the four gates.
→ The factor earns a documented place in a diversified portfolio.
Common mistakes
Trusting the backtest first
Starting with attractive returns encourages researchers to invent a rationale after the fact instead of testing a genuine mechanism.
Ignoring implementation drag
Turnover, market impact, taxes, and capacity can consume the return shown in an academic simulation.
Counting repackaged beta as a new factor
A candidate that closely duplicates a market-cap benchmark or an existing factor adds complexity without meaningful diversification.
Is it for you?
Best for
It is best for investment teams evaluating academic anomalies, smart-beta products, or systematic strategies.
Not ideal for
It is not designed to assess specialized alpha strategies whose edge is intentionally narrow, secret, or short-lived.
From the transcript
“there's four criteria that whittles down the potential hundreds or thousands to just these narrows, just this narrow set.”
“The first is that economic rationale that we talked about earlier, reward for bearing risks, structural and pediment or behavioral bias.”
“We want differentiated returns, particularly with respect to market cap benchmarks”
From the episode
Andrew Ang Discusses Asset Management (Podcast)
Andrew Ang