So you’ve read all about the benefits of index investing. Your portfolio is dirt cheap and broadly diversified. Heck, you’ve even stopped checking your portfolio value on a daily basis. Just when you thought you had it all figured out, ETF providers drop a bombshell by claiming there may be a smarter way to invest.
Traditional index investing is clearly not dumb, but it can sometimes feel that way. There’s a lot of academic research out there that has shown certain risk factors have historically generated higher returns than the market. By tilting towards any of these factors in your portfolio, you have the potential for market beating returns. Even better, by combining a number of these factors into your portfolio, you may enjoy a smoother ride in your pursuit of outperformance.
Based on this idea, iShares has created three multifactor equity ETFs (Canada, US and International), which follow the MSCI Diversified Multi-Factor Indexes. Instead of tilting towards a single factor, these indices combine four well-researched factors: value, momentum, size and quality. Let’s start by taking a closer look at the Canadian version. I’ll look at the US and international multifactor indexes in my next blog.
iShares Edge MSCI Multifactor Canada Index ETF (XFC)
With a management fee of 0.45%, you’ll pay 9 times more to hold the iShares Edge MSCI Multifactor Canada Index ETF (XFC), compared to its plain-vanilla counterpart, the iShares Core S&P/TSX Capped Composite Index ETF (XIC). You’ll also reduce your stock holdings by a third, from 240 companies to 81 companies.
Tracking error to the index is always an issue with any active strategy (and yes, factor investing is arguably an active strategy). One behavioural issue I’ve found with factor investing is the tendency to constantly compare your returns to the index. If your investment strategy strays too far below the index returns, you may be tempted to abandon it. By combining factors together into a single fund, it may help reduce the tracking error of the fund relative to the index (and help investors avoid making any knee-jerk reactions).
If we calculate the monthly tracking error of the single and multiple-factor MSCI indices relative to the MSCI Canada IMI Index (“the market”), we find that the MSCI Canada IMI Select Diversified Multiple-Factor Index tracking error is similar or lower than nearly all of the individual factor indices (with the exception of the quality index).
Monthly Tracking Error to Index: January 1999 to December 2015
Sources: MSCI, Dimensional Returns 2.0
Performance Summary: January 1999 to December 2015
Sources: MSCI, Dimensional Returns 2.0
No factor discussion would be complete without a regression analysis. I know that most of my blog readers hate these things. But the fact is, this statistical technique is the best way to measure whether a factor-based index is behaving the way you would expect.
In the tables below, I’ve compared the regression results for a broad-market index with those of the multiple-factor index. For each factor (size, value, momentum, quality) the amounts for the multiple-factor index should be higher than the broad-market index—this indicates the factor tilt. In this analysis, “alpha” represents the portion of the returns explained by something other than those factor tilts. So if the multi-factor strategy performs as expected, you should expect an alpha of zero.
Although the Canadian multiple-factor index had significant tilts towards momentum (+0.07), quality (+0.07) and size (+0.21), it had a negative tilt toward value (-0.03). That’s hard to explain. Also, since size was its biggest factor tilt, I would have expected to see outperformance of smaller companies over the measurement period. But if we review the index performance in the chart above, we find that smaller companies returned about 7.06%, while the market returned 7.24% (whoops).
Things get even more confusing. The unexplained alpha over the period was a whopping 2.41% per year. These results make me question whether there is something else driving the back-tested multiple-factor outperformance – if this is the case, investors may not be receiving a true multiple-factor Canadian equity ETF.
5-Factor Regression Results: January 1999 to December 2015
*Data shaded in light grey have a t-stat greater than 2, which indicates a statistically significant result
Sources: MSCI, AQR Data Library
Hi Justin,
Great article! What dataset do you use to run your regression? Is it possible that the funds are using a different definition for the factors leading to an apparent alpha? Also, do you have any good resources for someone looking learn how to run their own factor regressions or other quantitative methods (outside of years of CFA training…)? Portfolio visualizer seems to be quite good for US assets, but when evaluating assets in ex-US markets, the correlation (R2) is often <70%, doesn't seem like a good fit for the data…
Thanks again for all your work on the blog!
Dan
@Dan: I used AQR’s data set for this regression. The different data sets will lead to different tilt results, but they tend to still give similar high-level results.
CFA didn’t teach us how to run regressions ;) I learned from William Bernstein’s site:
http://www.efficientfrontier.com/ef/101/roll101.htm
Not sure what you’re trying to analyze in Portfolio Visualizer – are you using the correct currency?
Thank you for the link! Amazing that that guy is also a neurologist… As an example of what I did on portfolio visualizer, I tried analyzing XFC.TO using AQR factors (4 factor) for Canada. The R2 was 64.4%. I couldn’t see any option to choose currency, but I have to assume that Portfolio Visualizer is just using the AQR dataset and XFC.TO would be only in CAD…? Maybe I’m missing something on the page?
Thanks again, appreciate the follow-up.
Dan
@Dan: That’s correct – you would need to first convert the monthly returns of XFC to USD before running a regression using U.S. return data.
Thank you for the follow-up. Where do you download market data and exchange rate data from? The best I could find was to download monthly adjusted closing price from Yahoo Finance, multiply this by the monthly exchange rate, and then calculate monthly returns using a calculator? Is there a better method? Mine seems tedious…
Thanks again,
Dan
@Dan: I use Morningstar Direct and Dimensional Returns software. Individual investors may be able to obtain some data directly from the index providers (MSCI and S&P Dow Jones post some of their historical index data in CAD terms).
MSCI published a paper titled “Deploying Multi-Factor Index Allocations in Institutional Portfolios” showing that a multi factor index consisting of
MSCI World Value Weighted Index 25%
MSCI World Risk Weighted Index 25%
MSCI World Quality Index 25%
MSCI World Momentum Index 25%
outperformed MSCI World by 2.5% CAGR from May 1999 to September 2013 (6.7% vs 4.2%)
@David: There has been countless backtests on how factor investing has outperformed a plain-vanilla index strategy. The question is whether these strategies will outperform going forward, after fees, taxes and investor behaviour are taken into account.
Hi Justin! Thanks for the good analysis!
Do you recommend to include small value and mid cap value factor tilt USA ETF in the portefolios that you oversee? I’m refering to the recent white paper of your collegue Ben Felix! It seems for me that we are in the frontier of active managing…
@Mathieu Raymond: The PWL Toronto team currently constructs portfolios without any small or value tilts. There are decent Canadian, U.S. and international equity ETF options available for investors who do want those tilts (I plan to discuss these options in future blog posts).
Well, if you had started in January of 1999 with 25% in each constituent and never rebalanced, wouldn’t you end 2015 with a return which is the average of the 4 constituents ? ( Assuming that these are total returns with the dividends reinvested back into the relevant constituent ) This average is 9.17% versus the actual return of 11.86%. Wouldn’t this imply that rebalancing added 2.69% return ?
@Phil: MSCI does not just allocate 25% of each of those indexes – I just included the additional data to help with the discussion. They create the multi-factor indices by reviewing each individual security in the parent index. So there is no relevant data available in the posted numbers that would help us determine the value-add from rebalancing.
How often is the portfolio rebalanced and how much alpha comes from this ?
@Phil – the constituents are rebalanced on a semi-annual basis, at the end of May and November. I’m unable to determine the historical benefits of their rebalancing strategy.
Interesting analysis. Thanks for crunching the numbers and for sharing the results, Justin.
Makes you curious about the secret sauce that yielded the 2.4% α. Active management trading luck?
@Holger Reusch: If you torture the numbers long enough, they’ll say whatever you want them to ;)