What Your Portfolio's Factor Exposures Reveal About Your Real Risk

A 60-Second Primer on Factor Investing

In 1993, Eugene Fama and Kenneth French showed that stock returns aren't explained by market exposure alone. Small companies behave differently from large ones. Cheap stocks behave differently from expensive ones. Their original 3-factor model of Market, Size and Value has since expanded to six factors:

The regression also produces an alpha (return not explained by any factor) and an R-squared (the share of monthly return variance explained by the fitted model). If R-squared is 95%, the residual accounts for 5% of the sample variance. That residual can reflect omitted exposures, measurement error and asset-specific effects; it does not measure the share of cumulative investment return.

The Building Blocks: Individual Asset Classes

Before we look at portfolios, let's examine the raw ingredients. We regressed four common ETFs against the Fama-French 6-factor model using monthly returns. Here's what each one really is.

US Large Cap (SPY), 395 months

The S&P 500 is almost perfectly explained by six factors (99.3%R-squared). Essentially zero alpha, exactly what you'd expect from a market-cap-weighted index fund. The strong negative SMB loading (-0.151) confirms it is a large-cap fund. But notice the significant positive profitability loading (+0.059) and the negative momentum (-0.029). SPY has a subtle quality bias and acts as a mild contrarian, traits you won't find on the label.

US Aggregate Bonds (AGG), 267 months

Only 10.1% of the sample variance in bond returns is explained by these equity factors. This model leaves interest-rate and other bond risks outside its scope. The tiny but significant market beta (+0.067) captures the credit-spread dynamic: when equities rally, credit tightens, and bonds get a small tailwind. The negative HML loading (-0.083) makes intuitive sense: when value stocks do well (risk-on), safe bonds lag.

International Developed (EFA), 292 months

International developed markets carry a significant value tilt (+0.147 HML). This is no coincidence, European and Japanese markets are structurally more value-oriented than the US, with heavier weightings in financials, industrials, and energy. Many of these markets also trade at significantly cheaper CAPE valuations than the US. Only 75% of the variance is explained by US factors. The remaining 25% is residual variance; it cannot all be identified as a separate diversifying risk source from this regression alone.

Emerging Markets (EEM), 272 months

Only 56.0% explained by US factors, so nearly half of EM's variance is independent. The negative profitability loading (-0.177), while not quite statistically significant, reflects the lower corporate quality typical of emerging markets. The negative momentum loading suggests EM tends to lag when US momentum is strongest.

Portfolio #1: The Classic 60/40

The most popular portfolio in finance: 60% US equities (SPY) and 40% US aggregate bonds (AGG). We blended the monthly return series, weighting by allocation, and ran the 6-factor regression on the combined portfolio. 267 overlapping months of data.

Zero alpha. Exactly what a passive 60/40 should produce. But look at the market beta: 0.622, not 0.60. That's because bonds carry a small positive equity beta (+0.067), which pushes the portfolio's total market exposure above the nominal 60%.

The significant profitability loading (+0.050) is inherited entirely from SPY. The negative momentum (-0.023) means the portfolio slightly underperforms when momentum strategies are winning. Neither of these shows up on a standard asset allocation pie chart.

Portfolio #2: The Three-Fund Portfolio

A three-sleeve illustration: 50% US large cap, 30% international developed equities and 20% US aggregate bonds, using the SPY, EFA and AGG asset-class series. 267 overlapping months from October 2003. This is a proxy illustration, not a backtest of VTI, VXUS and BND; VXUS includes emerging markets as well as developed markets.

Adding 30% international equity introduces a borderline at the ten-percent levelvalue tilt (+0.043 HML) that the 60/40 doesn't have. This comes directly from the structural value orientation of European and Japanese markets.

The profitability loading drops to near zero (+0.015, not significant). International markets dilute the quality bias that SPY carries. Whether that's good or bad depends on your view: value investors welcome it; quality investors might not.

Portfolio #3: All-World Equity

An equity-only illustration: 60% US large cap and 40% international developed equities, using the SPY and EFA asset-class series. No bonds or emerging markets.292 overlapping months from September 2001. This is not a VTI/VXUS fund backtest.

The strongest value tilt of any portfolio we tested (+0.070 HML, significant at the five-percent level). With 40% in EAFE, the value loading is inherited from international markets and becomes a defining characteristic of the portfolio.

The negative alpha, -1.26% a year and borderline significant (p<0.10), reflects the US exceptionalism drag: international developed markets have underperformed US large cap over this sample period. That drag shows up as negative alpha in a US-factor model because the model attributes international-specific underperformance to the intercept.

Comparing All Three Side-by-Side

Here's the summary view. Only statistically significant loadings are highlighted.

Three patterns emerge:

What This Means for Your Portfolio

If you want to reduce hidden equity exposure

Your 60/40 has a market beta of 0.622, not 0.60. If you want true 60% equity exposure, you'd need to allocate roughly 58% to stocks and 42% to bonds to hit a 0.60 market beta after accounting for the bond allocation's equity sensitivity.

If you want factor diversification

A US-only portfolio (60/40) gives you market + size + quality exposure. Adding international (Three-Fund) replaces quality with value. Adding more international (All-World) deepens the value tilt, and as our survey of 18 CMA providers shows, every single firm expects international equities to outperform the US over the next decade. None of these give you momentum, which requires a dedicated allocation to a momentum ETF or strategy.

If you care about fees and alpha

All three portfolios produced zero or negative alpha. This is exactly right for passive index portfolios, after fees and factor exposures are accounted for, there's no return left over. If you're paying high fees for an active fund that also shows zero alpha in a factor regression, you're paying for factor exposure you could get cheaper.

If you use target-date or balanced funds

These funds are just pre-mixed versions of the portfolios above. Run them through the Factor Exposure Analyzer to see if you're getting any hidden factor tilts you didn't ask for. Some target-date funds tilt toward value or small-cap as the target date approaches, and factor analysis makes this visible.

Try It With Your Own Portfolio

The Factor Exposure Analyzer now supports multi-asset portfolios. Enter your tickers, set your weights, and get the full Fama-French 6-factor breakdown in seconds. It works with any ETF or stock available on US exchanges.

You can also use the built-in presets to get started:

• •60/40 (VTI + BND)
• •Three-Fund (VTI + VXUS + BND)
• •All-World Equity (VTI + VXUS)
• •Growth Tilt (QQQ + VTI + BND)

Methodology

All regressions use monthly return data and the Fama-French 6-factor model (Market, Size, Value, Profitability, Investment, Momentum). Returns are excess of the risk-free rate. Portfolio returns are computed as the weighted sum of constituent monthly returns, rebalanced monthly to target weights. ETF proxies: SPY (US Large Cap), AGG (US Aggregate Bonds), EFA (International Developed), EEM (Emerging Markets). The blended illustrations use those same asset-class series, not the fund tickers offered by the tool presets. Regression uses OLS with heteroskedasticity-consistent standard errors.

Factor data is sourced from Kenneth French's data library. Significance uses two-sided large-sample normal probabilities with robust standard errors. Stars indicate: *** p<0.01, ** p<0.05, * p<0.10.