Cross-Market Factor Validation
Test durable factors across new assets and markets before trusting them
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 96%
Begin with a limited set of economically intelligible factors, such as value, momentum, quality, or low risk, rather than searching a large factor zoo. Translate each factor into a comparable signal for a new asset class, country, or geography, then test it without redesigning the thesis to fit the new data. A successful transfer produces two benefits: an additional return source that is related but not perfectly correlated with existing strategies, and an out-of-sample test of the original idea. Repeated survival across stocks, countries, bonds, currencies, and emerging markets strengthens confidence that the factor reflects a persistent mechanism rather than a historical accident. Portfolio inclusion still depends on correlation, implementation costs, and whether the added market genuinely diversifies existing exposure.
Origin
Cliff Asness describes how AQR's predecessor team first treated countries like individual stocks, then extended value and momentum across bonds, currencies, equities, and new geographies.
Core principles
- 01Start with a small set of economically plausible factors
- 02Apply the same philosophy across comparable markets
- 03Treat new markets as out-of-sample tests
- 04Expect useful strategies to correlate without moving identically
- 05Expand both factor measurement and market coverage
How to run it
- 1
Choose a durable candidate
Select a factor with a plausible economic mechanism and meaningful prior evidence. Avoid beginning with hundreds of weakly differentiated signals.
Pro tip Start with broad concepts such as value, momentum, quality, or low risk.
Watch out A backtest alone does not establish that a factor is durable.
- 2
Translate the signal
Define how the same factor can be measured in another market without changing its underlying logic. For example, compare prospective real yields when translating value into sovereign bonds.
Pro tip Preserve the philosophy even when the market-specific variables differ.
Watch out Do not redefine the factor after seeing which specification performs best.
- 3
Run an out-of-sample test
Apply the translated signal to a market or geography not used to build the original model. Determine whether the expected relationship holds.
Pro tip Prefer genuinely new markets over another partition of the original dataset.
Watch out A new dataset is not useful if the model was repeatedly tuned against it.
- 4
Measure diversification
Estimate the new strategy's return, risk, and correlation with existing implementations. Look for positive evidence without assuming independence.
Pro tip Value partial diversification even when strategies remain meaningfully correlated.
Watch out Do not market correlated implementations as unrelated return streams.
- 5
Expand cautiously
Add validated implementations incrementally and continue monitoring them as real out-of-sample evidence accumulates.
Pro tip Treat live performance as another test rather than final proof.
Watch out Failure in a new market may expose either a weak factor or a poor translation.
In the wild
Asked to improve country allocation, the team treated countries as if they were individual stocks and applied academic work on value, momentum, and size. The approach held up well enough to support a paper and later expansion into additional markets.
→ A stock-selection philosophy became a reusable cross-asset research program.
After developing factors in established markets, a systematic manager applies the same definitions to emerging markets. If the effects persist and are not perfectly correlated with developed-market returns, the new implementation supplies both diversification and independent evidence.
→ The manager gains a new strategy and greater confidence in the underlying factors.
Common mistakes
Mining the factor zoo
Testing a large collection of weak signals increases the chance that apparently successful results are statistical accidents.
Pretending correlation is zero
The same factor in another market can diversify a portfolio without being independent of the original strategy.
Changing the thesis after every failure
Repeatedly redesigning a factor to fit each market destroys the value of out-of-sample validation.
Is it for you?
Best for
Systematic investors evaluating whether a factor is durable enough for diversified portfolio use.
Not ideal for
Investors without sufficient cross-market data or a consistent way to measure factors.
From the transcript
“which was to pretend countries were individual stocks and treat them like what we had studied in academia”
“Momentum is ridiculously easy. Everywhere you go, things tend to keep trending in the same direction.”
“when you go to a new market you haven't looked at yet, and it holds up, you go, maybe we're actually onto something here.”
From the episode
Cliff Asness on Celebrating 25 Years With AQR