I’ve been reading up quite a bit on factor investing recently. This has prompted me to want to share my thoughts on some of the different ways of implementing approaches based on factors and how they generate alpha.
What is factor investing?
Factors are measurable attributes of stocks that have been shown to persistently generate excess returns on average. The main factors are value, size, quality and momentum. Factors have their root in numerous empirical academic studies disproving or refining the efficient market hypothesis – the idea that with a sufficient number of intelligent well-informed investors stock prices should reflect all available information available and consequently that the only way to generate excess returns is to take more risk.
Factors arise because the assumptions underlying the efficient market hypothesis (EMH) do not closely reflect reality. The principal reason is the assumption that a sufficient number of investors are ‘intelligent’ (or ‘rational’), that is able to make the decisions that would arbitrage away excess risk-free returns, is ill-founded. In fact investors are seriously and systematically biased in a very wide variety of ways. The are numerous other market inefficiencies that likely also undermine EMH and give rise to factors including the role of information, illiquidity and constrained access to capital. By way of example, I discuss some of the underlying reasons for momentum in this post.
While the factors themselves are well-recognised, ‘factor investing’ also requires an implementation strategy that takes advantage of them.
If you read about factor investing you will pretty quickly come across smart beta. Smart beta is one implementation of factor investing.
The capital asset pricing model (CAPM), based on EMH, suggests that the only systematic source of excess returns from a stock come from accepting greater volatility than the overall market in the form of beta. The variance in returns beyond this (alpha) are empirically abundant in reality but unexplained by the model. The key insight from this model is that one can mitigate idiosyncratic risk without harming returns by diversifying across lowly correlated stocks.
Smart beta has its roots in this model but looks to also take advantage of some of the benefits of focussing on the factors that have been shown to systematically deliver excess returns. In some ways it is similar to passive investing, tilting the weighting of a broad index-like portfolio to the stocks characterised by a factor. By tilting the portfolio to multiple factors that are lowly correlated, smart beta can take also advantage of the benefits of diversifying across factors. For example, a portfolio partially weighted towards value and partially towards momentum may generate the excess returns from these factors while experiencing less volatility across the business cycle.
As long as factor premia continue to exist, smart beta seems a robust implementation to benefit from them. However, its top down and computationally complex nature make it less appropriate for individual investors and more appropriate for institutions or the creation of quasi-passive indices. More importantly, its broad brush index-like approach neglects a lot of the alpha available for investors using a more focused approach.
Applying factor based investing in a more concentrated portfolio should take a somewhat different approach. Instead of weighting an index from the top down to ’tilt’ towards factors it is more about maximising the factor premium from a selection of individual stock picks.
Traditionally generating alpha is about individual stock picking. It is about doing more detailed research and having a better sense of the market of an individual company’s prospects and its ‘correct’ valuation. It is how most investors approach things. However, the very existence of factors tells us that an ad-hoc approach to stock picking gets things systematically wrong on average. Generating alpha, much like poker, is a competitive zero sum game. One investor’s alpha is another investor’s negative alpha by definition. The unfortunate fact is that you win by capitalising from the mistakes and behavioural biases of others. Like in poker, the first step to doing this is to develop rules and discipline that get as much of the statistics on your side as possible.
So can a relatively unsophisticated amateur beat the professionals at this zero sum game? Stockopedia provides one answer through its Stockranks – I would recommend looking at these if you have not already. This article argues that fairly simple factor investing can work for individual investors and generate alpha without a great deal of effort as long as it is well thought through and applied systematically with discipline. Unsurprisingly, I firmly agree!
So how do you get the factors to work for you as best as possible?
Interactions and combinations
An individual fairly concentrated portfolio is well placed to take advantage of interactions and combinations between factors. Combinations of factors can be much more powerful than the sum of their parts. For example, ‘cheap but improving’ stocks (e.g. those with low PE but strong momentum) may do better than just generally cheap. Many stocks that are cheap are cheap for a reason.
Implementing a strategy to exploit factors in combination can get very complicated very quickly:
- There are a huge range of possible indicators to combine.
- There are correlations across different factors to account for.
- You can apply a structured approach which screens for stocks on certain factors and then applies other factors to the results of the screening. The premia attached to different factors may vary for subsets of the market.
- Implementation is not just about what you buy, but also how much and when you sell.
The fact that implementation gets complicated very quickly in one sense is very good, as it makes statistical arbitrage of the returns to the combinations of factors extremely difficult. This means that if you find an implementation strategy that works across the business cycle, it will likely persist a very long time. Approaches that attempt to identify complex implementation strategies based on combinations of factors empirically through back testing run the risk of data mining. For every strategy that works, you may find many others that are spurious temporary relationships. For this reason it is important that the strategy has a clear underlying logic to it that explains how it exploits the behaviour of other investors. Without this it is difficult to put the faith required into it.
While it is complicated I think there are techniques that make success more likely. One is to combine factors into a single weighted index as Stockopedia has done with its Stockranks. I prefer a more structured approach using screening to limit your ‘universe’ to a narrower subset of stocks to consider in more detail. This can help you by limiting the risk of you investing in the sorts of businesses where the factor indicators aren’t telling you what you think they are e.g. because they are inappropriate for the industry. It also greatly helps you focus in more detail on identifying the businesses with the best characteristics in combination rather than those that are simply better than average.
The initial stage of my process is to screen based on quality factors. My reasoning is that these factors are good predictors of long term performance, they are independent of the current share price and are more long term so require less frequent screening and this allows me to incorporate more qualitative factors in my assessment. Restricting my universe to these stocks tilts me towards this factor, mitigates risk and allows me to focus on how to best exploit other factors in combination in more detail.
Including unmeasurable qualitative assessments is somewhat anathema to a factor-based investing strategy. Remember the whole point is to systematically tilt the statistics in your favour by avoiding the ad-hoc assessments made by other investors. However, if you implement it in a more systematic way, I think qualitative judgment can add to alpha. I think of doing this as developing qualitative ‘factors’.
The reason I think this is valuable is that quantitatively-based strategies in particular (and probably many investors in general) have a blind spot. This is that all quantitative data is backwards looking, while a stock’s correct valuation depends upon its future. While the past is generally a good starting point for looking to a company’s future, there are some qualitative factors that can let you bridge the gap with more confidence: things like assessing the sustainability of a competitive advantage and the opportunities for further growth. I’ve previously discussed how I try to identify quality here. I think that incorporating these qualitative assessments has the potential to improve returns (and note that many successful investors, like Buffett, look at them closely). However, they are hard to define or test empirically so I have no way to know for sure that I am getting it right.
To be clear, for me only qualitative assessments relating to long term quality can really be successfully incorporated in an otherwise factor based strategy. In practice my reliance on them is mostly just to screen out businesses that seem to face risks of competition or limitations to growth in the future. It is tempting to try to incorporate qualitative assessment in a conceptual framework that directly looks at valuation to buy shares that look cheap. However, I find valuation assessment too arbitrary, particularly for growing businesses given it requires predicting the future, so I try to avoid putting much weight on it.
The other way to take advantage of factors is to employ a trading strategy that fully exploits momentum. I’ve discussed this previously – I principally mean holding winners and selling losers as ruthlessly and consistently as possible (e.g. with trailing stop losses).
Momentum is probably the most important factor in my view, particularly for relatively small investors who have greater flexibility to trade in and out of positions without suffering from liquidity constraints i.e. are able to trade without affecting the price. The alpha it generates is very large, particularly when you account for the fact that it is straightforward to incorporate into your strategy not only in what you buy but also in how you size your positions and when you sell. Most academic research on momentum I have seen uses momentum to indicate what to buy but then holds for a fixed time period rather than waiting for the momentum to reverse. The returns to momentum that tend to be found are large, but as the duration of a price trend can vary significantly stock to stock, they are likely to understate the returns from a trailing stop loss type strategy that only sells when momentum reverses. In my view for smaller investors this really is the ‘free lunch’ in investing.
Of course, that’s not to say there aren’t downsides or risks to momentum trading. It tends to imply substantially higher trading costs and it can come completely undone in sideways choppy markets, buying the highs and selling the lows. Following momentum can suck investors in to asset bubbles, which are obviously quite risky when they burst!
Combining with other factors can mitigate these risks (e.g. looking for less volatile high quality stocks, avoiding excessive overvaluation) though there are times when momentum investing and indeed other factor based strategies will underperform. Psychological resilience to stick to a strategy through these times is the other key ingredient required to make factor investing work.
Factors exist and arise from the systematic mistakes made by other investors. They can be exploited systematically by an individual investor, but only if care and discipline is applied. Remember active investing rather than buying a tracker is a zero sum game. If you approach investing on an ad-hoc basis or don’t understand how your investing strategy exploits factors, then it is more likely than not that you are the one being exploited.