Fact and Fiction about Low-Risk Investing (2020)
Ron Alquist, Andrea Frazzini, Antti Ilmanen, Lasse Heje Pedersen
The Journal of Portfolio Management Multi-Asset Special Issue 2020, 46 (6) 72-92, URL/AQR
After examining value and momentum, this week’s AGNOSTIC Paper examines some Fact and Fictions around defensive / low-risk investing. The defensive / low-risk factor captures various well-known effects like the low-volatility and Betting Against Beta effect, but also fundamental strategies like quality (a.k.a. the Quality Minus Junk factor).
Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.
Setup and Idea
Low-risk investing refers to the simple but provoking effect that low-risk securities (by whatever measure) generate better risk-adjusted returns than high-risk securities. This sounds strange at first glance, right? In a competitive market, no one should leave attractive investment opportunities on the table and higher returns should therefore (almost) always come with higher risks. Well, as we will see in this article, there are plausible reasons why this view is not quite right empirically. I mean, if it would be right, there wouldn’t be this article about the successful low-risk factor, right?
Data and Methodology
To implement the idea of buying (selling) low (high) risk securities, the authors draw on their earlier research and construct long-short portfolios from ten different risk measures (six statistical and four fundamental). They apply this methodology to a sample of US stocks starting in 1931 or the longest period with available data.
Importantly, the authors construct their portfolios with an expected beta of zero, i.e. the long-short portfolios should be market neutral. Why is this important? Because if go long low-risk and short high-risk securities, by definition you end up with a “net-risky” portfolio. With low-risk strategies, we want to bet on the spread between low and high-risk without holding a portfolio that is more risky or even short the overall market. So this adjustment is necessary.1If you go long a low-risk portfolio with a beta of 0.5 and short a high-risk portfolio with beta 1.5, the combination has a beta of -1. This is essentially a short-position on the stock market and thus a bet against the well-known equity risk premium. Not a good idea…
For most facts and fictions, the authors don’t use all ten variants for brevity. Instead, they mostly argue with the well-researched Betting Against Beta (BAB) and Quality Minus Junk (QMJ) factor. In some sense, those two are representative for low-risk strategies based on statistical (BAB) and fundamental risk measures (QMJ).
Important Results and Takeaways
Fact: Low-risk securities generate risk-adjusted outperformance
The first fact is quite obvious because without it, we couldn’t do this whole article. The authors show that all of their ten low-risk strategies generated statistically significant alpha against both the aggregate market and the Fama-French 5-factor model (plus momentum). Annual alphas are in the range of 2 to 6% and come with manageable turnover of around 40% per year (except for one signal that comes with turnovers of >200%).
By and large, these results clearly show that historically, low-risk securities significantly outperformed high-risk securities on a risk-adjusted basis. Furthermore, the turnover statistics also suggest that low-risk strategies remain profitable after costs (which is equally important for real-word investors).
Fiction: The low-risk premium is weaker than other factors
I don’t know where this idea comes from, but it is actually quite easy to test it with data. The authors compare the factor premiums of several low-risk strategies to the five factors of the Fama-French model and momentum. The result? Except for momentum, annual alphas versus the overall market were actually larger than for the other factors. For example, the Fama-French value premium stands at 3% per year while BAB and QMJ generated annual alphas of 9% and 6%, respectively. At least historically, the “low-risk premium” is thus comparable to other factor premiums, if not higher.
Fact: Low-risk strategies worked out-of-sample
As I mentioned in all of my last articles, one of the most powerful tool to examine the robustness of factors are out-of-sample tests. So let’s see how low-risk strategies score on that.
It is well-known (and quite reasonable) that factors decay after the first publication. The authors therefore compare the in-sample performance of factors with their out-of-sample performance in samples that go beyond the original publications. Except momentum and size, all factors of the Fama-French 5-factor model decayed quite substantially after their publication. However, this pattern is vastly different for low-risk strategies. The out-of-sample factor premiums of BAB and QMJ are substantially larger than for the respective in-sample periods. The authors also show that the BAB factor survived two popular scientific publications. The first paper by Black, Jensen, and Scholes in 1972, and the re-discovery by Frazzini and Pedersen in 2014.
Bottom line: there is strong out-of-sample evidence for low-risk strategies that is actually more robust than for other well-known factors.
Fiction: Low-risk profits come from industry bets
Once again, this is an important consideration and definitely makes sense at first glance. By construction, low-risk strategies bet on relatively safe securities which you often find in certain industries like Utilities or Telecommunications. Therefore, it is true that some of the low-risk premium certainly comes from betting on “safe” industries.
However, this is not the full story. You can always neutralize industry exposure (and virtually all other risks) in the construction of a factor portfolio. In most applications you should actually do this to test if your profits really come from the hypothesized effect or from other “unintended bets” like industry-exposure.
The authors apply such a procedure to two of their low-risk strategies and examine them within US and global industries. Within both samples, the strategies continued to earn positive Sharpe ratios. In fact, the industry-neutral factor worked even better. This strongly suggests that industry exposure is not the main driver of the low-risk premium. Instead, the profits seem to indeed come from the fact that low-risk securities outperform high-risk securities. And this is exactly what low-risk investing is all about.
Fact: Low-risk investing worked across geographies and asset classes
The authors already provided robust evidence and showed how low-risk strategies continued to work after impactful publications. For this fact, they go even further and examine the performance of low-risk strategies in international stock markets and even within other asset classes.
With respect to equities, the BAB factor generated significant positive alphas and Sharpe ratios within 24 out of 24 countries studied (each country has a sample period of about 30 years). The pattern is very similar for QMJ which delivered positive risk-adjusted returns in 22 out of 24 countries.
Regarding other asset classes, the authors cite research which documents the low-risk premium among government bonds, credit markets, and even for sports betting. The overall result is always the same: low-risk securities have a better risk-return profile than high-risk securities.
Fiction: Low-risk investing doesn’t work because the CAPM is dead
This one requires some knowledge of the Capital Asset Pricing Model (CAPM) and I assume that most readers are at least somewhat familiar with the model.2You need some nerdy interest in finance to read this blog in the first place, so I think this is a reasonable assumption. With respect to low-risk strategies, the most important result of the model is the security market line. In the CAPM world, this line summarizes the relation between risk and return in the sense that expected returns should linearly increase with higher betas.
More risk should be compensated by higher expected returns. So far, so plausible. Empirically however, and this is the academic origin of the low-risk factor, this is not the case. Most real-world estimates of the security market line suggest that it is flat or even downward sloping. This contradicts with the CAPM, but at the same time, it also opens the profit opportunity for low-risk investors. If high-beta stocks don’t offer an extra return, why bother with the additional risk when low-risk securities offer a better risk-return profile anyway?
Saying that low-risk strategies don’t work because the CAPM is dead therefore strongly suggests that the people behind such statements have not understood the idea in the first place. Low-risk investors make their money from the fact that the CAPM is dead because they bet on a too flat security market line. Of course, there may be valid arguments against low-risk investing. A “dead CAPM”, however, is actually the reason why low-risk investing (theoretically) works in the first place.
Fact: There is economic theory behind the low-risk premium
Similar to all other factors, we should have a good idea of who is on the other side of the trade and why this person is willing to take a losing bet. The empirical failure of the CAPM is a good point to start, but obviously, the real question is why the security market line is too flat. The authors present two theories to approach the issue, leverage constraints and lottery preferences.
Let’s start with leverage constraints and take the BAB factor as an example. With this strategy, we bet on securities that have a low beta because (due to the too flat security market line) these offer higher risk-adjusted returns.3Beta is of course just one risk-measure, but the mechanics are the same for whatever measure you prefer… The last part is important because to exploit the low-risk factor, you need access to leverage. Suppose you have a low-risk stock with a beta of 0.5 that gives you a return of 5%. On the other end, a high-risk stock with a beta of 1.5 gives you 10% return. In terms of “raw returns”, the second stock is of course more attractive (10% > 5%). Risk adjusted, however, it is worse. You get twice the return but take three times more risk. Pretty inefficient…
Now, suppose you want to achieve a return of 10%. The easiest way to do this is to accept the 1.5 beta and just buy the second stock. However, you could also borrow the same amount of money you already have and invest 200% of your initial equity in the first stock. This gives you the same 200% x 5% = 10% return but, and this is the catch, at a beta of just 200% x 0.5 = 1. Same return at lower risk. Any risk-averse person (and most of us are) should therefore prefer the levered low-risk stock over the unlevered high-risk stock.
Now, you may argue that this somewhat unrealistic. Absolutely! And this is exactly the point. Although it is theoretically optimal to invest in levered low-risk stocks, many investors just can’t do this. Running a levered equity strategy is not so easy in practice and requires attention to a lot of details (costs, collaterals, margins, etc.). As a consequence, many investors can’t use this advantage and must meet their return targets by simply taking more risk.4Unfortunately, including me…
In theory, this leads to additional demand for high-risk securities which leads to higher prices and consequently, lower expected returns. The low-risk securities therefore become relatively underpriced and offer better risk-adjusted returns. Said differently, if you have access to leverage and can apply it to low-risk securities, you have a competitive advantage over some other investors. The reward is the low-risk premium. The other side of the trade are unsophisticated investors who don’t have the necessary infrastructure but still want to meet their return targets.
The other theory are lottery preferences. The mechanics are exactly the same, but the reason why high-risk securities are overpriced is now different. Research in behavioral finance shows that some investors are not perfectly rational and exhibit a preference for lottery stocks. In simple terms, this means you don’t like the boring and stable companies with steady profits. Instead, you want to invest in the current hot start-up which has a tiny chance of becoming the next Google and paying off really big. If enough investors suffer from this bias, they bid up prices of high-risk stocks and low-risk securities become again relatively undervalued. Although still debated in academia, I personally believe that this theory is very applicable to the meme stock mania of 2021.
No matter which theory you like more, the important point is that there are two plausible ones out there. In reality, both of them (and maybe even more) are probably active at the same time anyway.
Fiction: Low-risk investing does not survive trading costs
This a common and important point about any strategy. Does it work after costs? A strategy that looks good in a hypothetical backtest but does not survive real-world transaction costs may be a simulating intellectual exercise but practically, it is still worthless. So let’s see how low-risk strategies score here.
The authors show that most of their low-risk strategies come with implementable levels of turnover in the range of 40% per year.5For comparison, Fama & French’s value and momentum factor had turnovers of 26% and 100%, respectively. Moreover, the authors also name certain implementation details which help to reduce transaction costs of low-risk strategies even further. Finally, they estimate that given the historical premium of the BAB factor, transaction costs would have to be well above 100 basis points to turn the strategy net-negative. Needless to say, this is way more than what institutional (and nowadays also retail investors) pay. So this is strong evidence that low-risk strategies most likely remain profitable after costs.
The authors also note that low-risk strategies are not limited to illiquid or small cap stocks. Although they historically worked better among small caps (most factors do), the results for large caps are also significant. In addition to that, the authors cite research that finds low-risk premiums within very liquid assets like treasury bonds or equity indices.
No matter how you look at it, low-risk strategies are not limited to illiquid securities and have manageable turnover. The consideration that they don’t survive trading costs is therefore generally a good one, but in this case just a fiction.
Fact: Low-risk investing can lose money in bear markets
It is quite tempting to equate “low-risk” with “don’t lose money”. While it is certainly an advantage of low-risk strategies to have (wait for it) lower risk, this still goes a little too far. Except you are Renaissance Technologies, every strategy loses money at some point in time and the timing ultimately depends on portfolio construction.
A beta-neutral version of low-risk is equally likely to lose in bear and bull markets. That’s the idea of being market-neutral. In contrast, dollar-neutral low-risk strategies have a negative net-beta by construction and should therefore make money when the market is down. Long-only versions of low-risk strategies overweight securities with low betas and therefore tend to have a beta that is positive but smaller than one.. In a bear market, such strategies therefore still lose money but (hopefully) less than their benchmarks.
Bottom line: “low-risk” is not equivalent to “not losing money when the market is down”. Different portfolio constructions yield different results. That said, many investors still use low-risk strategies as a downside protection to lose less in bear markets.
Fiction: Low-risk factors became too expensive
I concluded last week’s post with the notion that the value factor currently looks very cheap. You can of course do the same exercise with low-risk and look at the valuation spread between low and high-risk stocks to determine how cheap or expensive the factor currently is.
The paper covers data until September 2019 and back then, the low-risk factor looked somewhat expensive relative to its own history. However, and this is a general theme of these articles on the popular factors, low-risk strategies worked even better when combined with other factors like value and momentum. So given all the evidence that (hopefully) came through in this article, the low-risk factor deserves its spot in the investing tool-kit even when it looks a little more expensive than usual.6Unfortunately, I don’t have the data resources to update the valuation spread of the factor. But given that the low-risk factor has not spectacularly outperformed over the last years, I guess that the factor’s current valuation is still around its own historical average.
- AgPa #74: Peer-Reviewed Research is Not Helpful to Predict Returns – Really?
- AgPa #73: Country and Industry Momentum
- AgPa #72: Machine-Reading of Private Equity Prospectuses
- AgPa #71: Go Where the Earnings (per Share) Are
This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). All content that is not my intellectual property is marked as such. If you own the intellectual property displayed on this website and do not agree with my use of it, please send me an e-mail and I will remedy the situation immediately. Please also read the Disclaimer.
|If you go long a low-risk portfolio with a beta of 0.5 and short a high-risk portfolio with beta 1.5, the combination has a beta of -1. This is essentially a short-position on the stock market and thus a bet against the well-known equity risk premium. Not a good idea…
|You need some nerdy interest in finance to read this blog in the first place, so I think this is a reasonable assumption.
|Beta is of course just one risk-measure, but the mechanics are the same for whatever measure you prefer…
|Unfortunately, including me…
|For comparison, Fama & French’s value and momentum factor had turnovers of 26% and 100%, respectively.
|Unfortunately, I don’t have the data resources to update the valuation spread of the factor. But given that the low-risk factor has not spectacularly outperformed over the last years, I guess that the factor’s current valuation is still around its own historical average.