Factors come and go…

The investment industry is full of fads and trends that come, lead to disappointing results, and then disappear again. Curiously, some of the investment trends are regional in nature. Structured products have been popular in Europe but much less so in the US. Factor investing and smart beta, meanwhile, have been quite popular in the US in the 2010s but never really caught on in Europe. All the better, because factor ETFs tend to disappoint on average.

One reason why factor investing is disappointed is the publication bias shown in the link above. But another one is that market dynamics change all the time. Value works one day and then stops working for a while before it comes back again. Same with small cap investing and all other factors.

Investors can deal with this ever-changing environment by combining different factor exposures in one portfolio. This approach marries the benefits of factor exposures with the benefits of diversification. But it has a significant drawback. What if a factor stops working for a long time or stops working altogether? What if there is a new, so far unknown factor emerging?

A more sophisticated approach to factor diversification is to use regime-switching models that change factor allocations depending on the prevailing macro environment. However, these regime-switching approaches have drawbacks of their own. For one, they don’t allow for new factors that aren’t included in the model calibration to emerge. They tend to be rather inflexible in their linkage between the macro environment and the factor performance.

If high inflation leads to factor X outperforming in the 1960s, then the model will assume that factor X will again outperform in the high inflation environment of the early 2020s. Yet, factor X may have worked in the 1960s while labour markets were more rigid and wage growth closely tracked inflation but may not work today when wage growth is much less correlated with inflation. The regime-switching model does not account for changes that are not reflected in the model, so it may make mistakes.

Siddharta Chib and Simon Smith have developed a new, more flexible approach that tries to improve on these shortcomings. They use the six factors described by Fama and French to construct the optimal portfolio with the highest risk-adjusted returns between 1963 and 2024. The six factors they consider are:

• Market factor (i.e. stocks with a higher market beta outperform stocks with a lower market beta)

• Small cap factor (i.e. smaller companies outperform larger companies)

• Value factor (i.e. cheap stocks outperform expensive stocks)

• Quality factor (i.e. profitable stocks outperform unprofitable stocks)

• Investment factor (i.e. companies with more conservative investment spending outperform ones with more aggressive investment spending)

• Momentum factor (i.e. past winners outperform past losers)

Then they try to identify break points in the last sixty years when the market environment for factors has changed. They identify three important breaks:

• 1975 or roughly during the oil crises that triggered the high inflation of the 1970s and the subsequent disinflation in the 1980s.

• 1995 when the internet emerged, and finance and business were digitised.

• 2005 or just before the onset of the financial crisis and the era of zero interest rates.

They calculate the ideal factor exposure for the periods between the three break points and the ideal factor exposure for an investor who holds the same portfolio for sixty years from 1963 to the end of 2023. The chart below shows the results.

Ideal factor exposures to maximise Sharpe ratio

Source: Chib and Smith (2024)

First, I want to stress that the results above are for US stock markets. Factors work differently in Europe and Asia. For example, the underperformance of value stocks has been much less pronounced and shorter in Europe during the 2010s than in the US. So, don’t translate the chart above to markets outside the US, please. You will make big mistakes.

What I find notable about the results is:

• If you optimise a portfolio across the different regimes and ignore the breakpoints (the portfolio on the very right in the chart above) you will end up with a portfolio that can be very far from optimal. Just look at the quality factor, which only enters the optimal portfolio until 2005 and with a weight of 20-25%. But if you optimise the portfolio for the whole sixty years, the quality factor gets a weight of 40%. Weird, isn’t it? Even at the best of times, you are some 15% overweight on the quality factor.

• The value factor seems to have stopped working a long time ago and hasn’t entered the optimal portfolio since the 1990s. This is something I have talked about before, though there are a lot of other results that show that value still works and is one of the key factors to have in a portfolio.

• Since 2005, we have been essentially in a lean factor world where only two factors still work: Quality and the market factor. This is a result that I find particularly doubtful. First, it seems to me this result is heavily distorted by the long-term outperformance of US tech stocks which has not been seen outside the US (or to a far lesser extent). Second, it seems to me that with the inflationary bout of the recent past, we may be back to a nonzero interest rate environment again (I am in the camp that expects interest rates to average much lower levels than what we see today, but still well above zero). And if the latest regime change was triggered by the zero interest rate policies of central banks, we might have entered a whole new regime in 2022 or 2023. In this case the results for the period 2005 to 2023 are completely useless for the future.

If that sounds frustrating to you, don’t despair. This is what makes investing fun and interesting. The world changes all the time in wholly unforeseen ways that are not captured by any theory or model. Just look at how market reactions to macro news have changed in the 2010s and you know why I am excited about the next ten years. It is going to be fun to discover what works and what doesn’t. And I for one believe we are going to have many surprises. Let the fun begin…