Profiling Factor ETF Correlations

Slicing and dicing the US equity market into factor buckets is, at its core, an effort to enhance return by engineering more control over risk management. A key part of this framework is recognizing that risk and return for the stock market overall is a byproduct of multiple factors, such as shares trading at low valuations or posting strong price momentum in the recent past. In turn, it’s reasonable to assume that a set of factor ETFs will exhibit relatively low correlations with one another, offering a degree of diversification otherwise unavailable via standard portfolio designs for capturing equity beta. To test that assumption, let’s review the return correlations for a broad set of factor ETFs in recent history.

For this test we’ll review numbers for a dozen factor ETFs plus a proxy for the broad market beta:

Let’s start by looking at daily returns for the past 12 months (through yesterday’s close, Nov. 5). As the table below shows, the main result is that correlations are generally quite high (roughly 0.7 and higher) relative to the market overall, as proxied by SPDR S&P 500 (SPY). That’s a clue for thinking that the diversification benefits from factor ETFs may be lighter than the marketing brochures for these funds imply.

Interested In Learning R For Portfolio Analysis?
Quantitative Investment Portfolio Analytics In R:
An Introduction To R For Modeling Portfolio Risk and Return

By James Picerno

The lowest correlation for the trailing one-year period is 0.71 for momentum (USMV) and micro-cap (IWC). That’s far enough below 1.0 (perfect positive correlation) to offer diversification benefits but it pales next to stock/bond correlations that are closer to 0.0, much less the absence of correlation or negative correlation readings.

But let’s not be too hasty. A closer look offers a more nuanced profile when comparing factor ETFs to one another and at different time periods and return frequencies. If we extend the trailing period to the past five years for daily returns, for instance, the correlations fall further for equity factors, although primarily for pairings with mid-cap value stocks. Indeed, as the next chart below shows, iShares S&P Mid-Cap 400 Value (IJJ) reflects correlations of roughly 0.2 to 0.3 vs. the rest of the field – far below the other pairings. What’s going on with mid-cap value stocks? Unclear, but it’s an area that’s worthy of deeper study. The main question: Are the low correlations with mid-cap value recurring or specific to the past five-year period. Or perhaps there’s a data error in the historical record that’s skewing results. For now, the result should be viewed with caution since it’s a clear outlier.

Finally, consider how correlations for rolling one-year returns compare. By this measure there’s a wider range of readings vs. daily returns. For example, low volatility (USMV) shares low correlations (below 0.4) with small-cap growth (IJT) and micro-cap (IWC). A similarly low-correlation profile applies to momentum (MTUM) and mid-cap value (IJJ).

The numbers above tell us that factor ETFs offer a fair amount of diversification benefits, albeit with limits. Surprising? Not really. You can’t get blood out of a stone or consistently low/negative correlations by slicing up the US equity bucket into smaller pieces. Stocks, after all, are still stocks and will usually exhibit a higher degree of market beta.

That said, the diversification opportunities that do exist with factor ETFs are compelling, if only when compared to the usual suspects. There are no silver bullets here, but a carefully designed and managed equity portfolio comprised of factor ETFs may provide marginally improved risk and return benefits vs. a conventional approach to investing in US equity beta.

4 thoughts on “Profiling Factor ETF Correlations

  1. Pingback: Profiling Factor ETF Correlations -

  2. Dan

    try running the correlations on excess returns – versus the total us market, say VTI or SCHB for example. This normalizes the results and provides more stark examples of the diversification benefits. Will find that the Value and Momentum excess correlations will be negative, highlighting there diversification benefits

  3. Pingback: Quantocracy's Daily Wrap for 11/06/2018 | Quantocracy

  4. James Picerno Post author

    Yes, you are correct: analyzing correlations based on risk premia is an alternative and arguably a superior method. But I like to start with total return for a profile of how the numbers stack in terms of what investors are actually receiving. In practice, I crunch the numbers on both fronts, even though I favored one view in a blog post. In a future update I’ll compare correlations on both fronts.

Comments are closed.