VEA produced the same 10.02% annualized return over the same 58 months in 2 FactorLens runs. The model explained 72.67% of its variation with US factors and 90.81% with developed-market factors.
The ETF stayed fixed. The regional yardstick changed.
This is an easy setting to miss in FactorLens. A regression always explains a portfolio relative to a chosen factor set. Running a developed-markets fund against US factors can turn a model mismatch into a story about alpha or unusual tilts.
Hold the return window fixed
Both runs used VEA at 100%, monthly data from September 2021 through June 2026, and the same 6 factors: market, size, value, profitability, investment, and momentum. Each run contained 58 observations. Only the region changed from US to Developed Markets.
| Reading | US factors | Developed-market factors |
|---|---|---|
| Data window | Sep 2021 to Jun 2026 | Sep 2021 to Jun 2026 |
| Months | 58 | 58 |
| Portfolio annual return | 10.02% | 10.02% |
| R-squared | 0.7267 | 0.9081 |
| Annualized alpha | -0.45% | +0.10% |
| Alpha t-statistic | -0.10 | 0.04 |
The lower US R-squared leaves 27.33% of monthly variation outside the model. The developed-market run leaves 9.19%. A low R-squared can reflect a genuinely unusual return source. Here it mostly says that US factor returns are a poor ruler for a fund holding developed markets outside the US.
The loadings change too
| Factor | US loading | US t-stat | Developed loading | Developed t-stat |
|---|---|---|---|---|
| Market | 0.8690 | 11.05 | 1.0616 | 20.75 |
| Size | -0.0661 | -0.47 | 0.4386 | 3.45 |
| Value | 0.1162 | 0.79 | 0.0943 | 0.71 |
| Profitability | 0.0763 | 0.57 | 0.2612 | 1.79 |
| Investment | 0.2740 | 1.49 | 0.2587 | 1.59 |
| Momentum | -0.1790 | -1.79 | -0.0868 | -1.13 |
The size result is the clearest failure. US factors give VEA a -0.0661 size loading with a -0.47 t-statistic. The developed-market factors give it a 0.4386 loading with a 3.45 t-statistic. The 2 readings even disagree on the sign.
Market beta also moves from 0.8690 to 1.0616. The developed-market result says VEA behaved a little more aggressively than its regional market factor over this window. The US result says it carried less market exposure. Both coefficients are mathematically valid. They answer different questions.
Alpha absorbs a bad benchmark
The US run reports -0.45% annualized alpha. The developed-market run reports +0.10%. Both alpha t-statistics sit near 0, so neither result supports a persistent residual return. The shift still shows how the intercept absorbs whatever the selected factors fail to explain.
This is why a positive alpha beside a weak R-squared deserves suspicion. The residual can contain manager skill, but it can also contain currency, region, sector, or asset-class exposure that the chosen factors never represented.
Choose the question before the model
Use US factors for a US equity portfolio, developed-market factors for developed international equity, and emerging-market factors for an emerging-market allocation. A global mix needs separate sleeve analysis or a factor set built for the combined universe.
I would read the developed-market VEA result as the primary one. It explains more of the same return path and finds a statistically significant size loading that the US model misses. The 2nd run used a ruler cut for the right market.
Past performance does not guarantee future results. Regression results describe the historical sample and are not return forecasts.