As we saw in a previous post, factors have provided investors with a way to enjoy superior returns over a market-cap weighted benchmark over a long period of time. But what purely looking at outperformance fails to account for, is how difficult the journey to those higher returns has been. Factors have been known to undergo long periods of underperformance, making individual factor funds difficult for investors to stick with. Value factor funds, for example, have been heavily underperforming the market for the last 10 years – not many people are able to stick with investments which can underperform for a decade or more.
Whilst funds do exist that aim to capture one specific factor, most who choose to invest in factors choose a diversified approach. Combining a number of different factors into one portfolio reduces the risk that any one factor suffers a long period of underperformance. This is known as a “multi-factor” portfolio.
Multifactor portfolios (and multifactor funds) are a good way of exposing a portfolio to multiple factor exposures, with the idea being that the portfolio will benefit from the higher returns from multiple factors, but with lower risk than betting on any single factor.
Multifactor portfolios – the evidence
Everybody’s seen the quilt chart used to show the benefit of diversification for asset classes, but it works equally well for showing the benefits of diversifying by factor. Different factors have had wildly different performance year-to-year – for example, there was an over 30% difference between the MSCI World momentum index and the small cap index in 2009.
MSCI World factor performance
Source: Morningstar
To make it even more difficult for individual factor investors, performance varies not only between factors, but between geographies too. The chart from Credit Suisse below shows how factors have performed in the US vs in the UK since the financial crisis. It’s interesting to see how there hasn’t been a single year where the ranked factor performances have been the same.
Given the varying performances, it makes sense (not only from a performance standpoint, but also from a behavioural one) to diversify not only between factors, but also between geographies.
USA vs UK factor premiums
Source: Credit Suisse
Going back further in time, using data from 1927 to 2017, the table below from Larry Swedroe shows how combining factors by simply giving each of them an equal weight in a portfolio has led to outperformance. The table shows the historic odds of outperforming the risk-free rate. In the table, P1 is an equal weighting of the first four factors. P2 is an equal weighting of the first five. And P3 substitutes quality for profitability:
Source: Larry Swedroe
A couple of interesting observations:
- Individual factors can underperform for a long time. Even over 20 years, all factors bar momentum had a chance of underperforming T-bills. (NB: This does not mean that momentum will always outperform over 20 years). After 5 years, the size factor underperformed T-bills 30% of the time. Given that most people tend to evaluate investments on a short (5 year) timeframe, sticking with the size factor would be pretty difficult if the size factor hadn’t given you a better return than treasuries.
- Diversification between factors has worked incredibly well. Regardless of the time period, any of the 3 combinations of factors provided a lower chance of underperformance than any individual factor.
Another table from the same article looks at risk and return stats for factors over the same time period:
Source: Larry Swedroe
The table shows:
- The multifactor approaches generated higher mean returns the than the majority of individual factors
- By combining factors, the risk of the combined portfolio is below that of any individual factor. Any of the 3 weighting schemes had a lower standard deviation than any individual factor. Unsurprisingly, the multifactor portfolios come out far ahead on a risk-adjusted Sharpe ratio basis.
Further research from S&P Dow Jones has shown investing in a combination of their quality, value, and momentum indices would have produced better performance than any one of the individual indices:
Source: S&P Dow Jones Indices
Consistent with the evidence from Morningstar and Larry Swedroe, S&P’s multi-factor strategy didn’t just outperform on a raw performance basis, but on a risk-adjusted basis as well. The multi-factor index provided higher returns than any single factor over 15 years, with volatility roughly in line with the quality factor, which had the lowest volatility out of the 3 factors. As a result, the multifactor strategy comes out ahead of any single strategy on a risk-adjusted basis:
Source: S&P Dow Jones Indices
Further evidence in the Journal of Portfolio Management backs up S&P’s findings, pointing to the benefits of a multi-factor approach, and shows how factor diversification improves expected returns, providing a smoother ride with smaller drawdowns.
Do multifactor funds reduce tail risk?
So it looks like multifactor has an advantage over single factor for improving returns and lowering volatility, but does it provide diversification when markets crash?
Research from the Zurich Insurance Company has suggested that factors can help reduce the maximum drawdown of a portfolio, and do provide diversification benefits when markets fall. The authors note that “overall our findings show that with the exception of the small-cap factor (SMB) it can be noticed that the returns at the Fama-French factors remain high and positive for the 10 worst cases of the market. Style factors other than SMB offer diversification benefits in tail events.” Given that of the 5 factors the authors testing, 4 provided tail risk diversification, they suggest that an allocation to a multifactor approach may help reduce the drawdowns of a portfolio:
Source: ‘Tail Behavior in Portfolio Optimization for Equity Style Factors’
However, contrary evidence from Research Affiliates has found not only that multifactor portfolios are prone to crashes, but that they in fact do not provide diversification during periods of drawdown. The black line on the graph below shows the performance of a multifactor strategy, and the red line shows how the portfolio would have performed if the returns of the multifactor strategy were normally distributed. The large crash in 2009 suffered by the factor portfolio shows that the multifactor strategy suffered a larger drawdown, and was therefore more negatively skewed, than would be expected under the normal distribution.
Source: Research Affiliates
The authors find that “forming portfolios of factors does not mitigate the risk of large drawdowns to the extent we might expect, because the large drawdowns of individual factors often happen at the same time.”
Whilst the evidence for its ability to mitigate drawdowns is mixed, multifactor investing does seem to do a good job reducing the risk of underperformance through investing in any one factor. It comes out ahead performance-wise versus individual factors and, thanks to its ability to reduce factor-specific risks, comes out way ahead on a risk-adjusted basis.
Multifactor portfolio construction – integrated vs mixed
For those that choose to allocate to a multifactor fund, just like when buying any regular fund, it’s important to know how the portfolio is constructed. There are two main approaches to capturing multiple factors in one portfolio – ‘mixing’ and ‘integrating’.
A mixed factor approach involves selecting stocks which score highly on one individual factor, but might not score highly on others. For example, a stock might be selected because it has strong momentum, and another chosen because it scores highly on value metrics. The two stocks score highly on one factor, but might have low scores for other factors.
An integrated approach involves buying stocks that are all-rounders – they might not have the highest scores on individual factors, but they score well on a combination of factors.
A mixed approach is simple and transparent, but can lead to lower overall factor exposure in the portfolio – for example, stocks chosen because of their high momentum scores may have low scores on value. An integrated approach is more complex and can result in higher active risk, but usually leads to higher factor exposures and has led to better performance in backtests (evidence here, here, and here). As we’re trying to maximise factor exposure, this approach seems sensible.
However, the jury is still out on which is preferable. There has been some research indicating that a mixed approach might not suffer from buying factors that cancel each other out, as the factors ‘mature’ (i.e. provide their benefit) over different time periods – the momentum factor might mature in one year, but the value factor might mature over multiple decades. A stock with a high momentum but low value score might therefore still benefit a portfolio if it was sold in line with its maturity date – after one year. Mixing is also the simplest and more transparent option, making performance attribution easier, and so potentially making the strategy easier to stick with.
Although there’s been little agreement on whether there is a ‘correct’ way to construct a multifactor portfolio, or whether multifactor portfolios are able to reduce drawdowns, almost everyone agrees that a diversified set of factor exposures is preferable to investing in a single factor. Multifactor portfolios have, in the past, been able to outperform single-factor exposures, with lower risk.
Whilst multifactor investing might seem too good to be true, there are several problems with factor investing, which need thinking about. The next post looks at the issues investors need to consider before committing to a multifactor allocation.