definition-390785_1920 edit

Problems with smart beta – part 2: Defining a factor is hard

This is the second post in a series discussing some of the problems associated with investing in “smart beta” strategies. For the previous post on the cyclicality of factors, click here

Defining a factor is hard


One of the other major problems with smart beta is that one person’s idea of what constitutes a factor may be very different to another person’s. For example, there are a truly vast number of “value” indices, all of which are constructed slightly differently, using different characteristics to define value, measuring the characteristics differently, and weighting them differently.

If we were to construct a value index ourselves, we’d have to make several difficult choices. Does “value” mean a low P/E ratio? Or a low P/B ratio? Or a high dividend yield? Or something else? If we decide to construct an index that chooses all those three above, do we use trailing or forward P/E? Do we use lagged share prices to match the lagged book value? And how do we select stocks – is it those with the top decile of our value characteristic? Or the top quartile? And if we then decide on all that, do we weight all measures equally – an even split between P/E, P/B, and dividend yield? Or alter weightings based on relative valuations? How often do we rebalance?

Defining a factor can be difficult, but is hugely important for the index’s returns.

As an example, the quilt chart below shows the performance of 8 different “value” strategies. All generated an annualized return within a band of 11.4% to 12.8% over a 50-year period ending December 31, 2017, so had similar returns over the longer term. But if you look over the shorter term, all have performed very differently:

Value factor - different returns

Source: Research Affiliates

Since 2000, you can see in another way how the different definitions of value have all performed surprisingly differently:

Value factor - different returns 2

Source: Factor Research

The difference between P/B (the traditional value metric) and P/E (an alternative value metric) has been huge over the last 18 years. The investor who’d chosen to invest solely based on P/B valuations would’ve been pretty unhappy when comparing returns to his fellow value investor who’d used P/E ratios.

Even if we manage to all agree on THE best definition of value, different value stocks may perform differently in different market environments, according to researchers at AQR. If we think about two different stocks that are both cheap on value. “A stock with steady earnings might be cheap because it has very poor growth prospects. Another stock might have high potential but might be cheap because of poor management and high debt levels. These two stocks could have very different reactions to macro data. Higher inflation, for example, could erode the value of the first company’s steady earnings, but could help the second company by devaluing the debt it holds.”

This further complicates value investing, as less predictable performance makes the strategy more difficult to stick with. If you expect your value strategy to outperform during inflationary periods, but the opposite happens, it’s going to be almost impossible to stick to your guns.

It’s not just value


Agreeing on how to define a factor isn’t unique to value, and is a problem for all other factors. There are no universally-agreed definitions for factors, and all ways of measurement have their pros and cons. The quality factor, for example, has at least a dozen different ways of being defined, including profitability, growth in margins, financial structure, earnings stability, accruals, advertising/R&D expenses – the list goes on. Research from Research Affiliates has even shown that “multiple factors commonly used in the practitioner community as a proxy for quality, including leverage and earnings growth, lack strong evidence of producing a quality premium.”

There have been similar issues defining the low-volatility factor, and there’s been considerable debate around whether or not the size factor even exists in its traditional academic form.

Differences with index construction


The fact that the largest index providers can’t even agree on how to define factors is pretty telling. Whilst both agree that P/B and P/E should be used, MSCI favour the addition of dividend yield, and S&P favour the addition of sales/price.

The differences are not just in factor definition, but also in index construction. The S&P value index ranks the parent index (the S&P 500) by value factor score, and allocates the top 33% to the value index, the bottom 33% to the growth index, and the middle 34% is split between the 2. The MSCI value index is created by assigning all stocks in the parent index (the MSCI USA) a score between 0 and 1 of how highly it scores on MSCI’s value characteristics. That number is turned into a percentage, and that percentage is the percentage of the company’s market capitalisation that’s allocated to the value index.

Comparing the returns from the MSCI factor indices to the S&P Dow Jones factor indices shows how the difference in factor definition combined with the differences in index construction result in what can be very different returns. One the face of it, the indices appear to be doing similar things, but the devil is in the details, and seemingly minor differences in definitions and construction methodologies can lead to surprisingly different returns:

S&P vs MSCI – Value indices

S&P vs MSCI - Value factor

S&P vs MSCI – Momentum indices

S&P vs MSCI - Momentum factor

S&P vs MSCI – Quality indices

S&P vs MSCI - Quality factor



For those allocating to smart beta strategies, investors need to be aware of how their factor is being defined, and whether they’re comfortable that their definition of the factor has sufficient evidence behind it. They also need to be comfortable with periods of underperformance versus other possible definitions.

A sensible approach taken by some has been to diversify by factor definition – i.e. investing in a factor fund that uses multiple definitions. For those that favour multifactor investing, multifactor funds can target factor exposure through multiple definitions – investing in several factors, each of which has a few different ways of being defined. For example, a multifactor fund could invest in value, momentum, and quality with value being defined as stocks with low P/B, P/FCF, and P/E, momentum being stocks with high 12 month and 7 month momentum, and quality being  stocks with high return on equity, profitability, and low leverage.

This adds an extra layer of diversification for those that want to avoid the drawbacks of allocating to single factors.

However, investors also need to be aware of how the indices they’re tracking are constructed, and be comfortable that the methodology accurately captures the factor exposure that they’re trying to harvest. They also need to understand that, due to different construction methodologies, there are likely to be significant differences in performance between seemingly similar indices.

The minutiae of factor definition and index construction will undoubtedly make most investors glaze over, but the crux of this point is that understanding what you’re invested in is just as important as why you’re investing in it. Investors still need to conduct their own due diligence, and dig deep into the smart beta funds they’re investing in.

The next post in the series focuses on how some factors can change over time, and how that makes allocating to smart beta strategies more difficult.

Share on Facebook
Share on Twitter
Share on LinkedIn

Past performance does not guarantee future performance and the value of investments can fall as well as rise. The information on this site is provided for information only and does not constitute, and should not be construed as, investment advice or a recommendation to buy, sell, or otherwise transact in any investment including any products or services or an invitation, offer or solicitation to engage in any investment activity. Please refer to the full disclaimer on the disclaimer page.

Notify of

Inline Feedbacks
View all comments