The future of factor investing…doesn’t look promising
Factor investing or smart beta investing has become all the rage, though admittedly more in the United States than in the UK and Europe. In their efforts to reduce costs while still generating above-market returns many pension funds have shifted their assets into smart beta ETFs that try to mimic one factor or another.
Meanwhile, the factor zoo has grown bigger and bigger and by now encompasses several hundred anomalies that promise significant outperformance. At least within the sample of the publication exploring the anomaly.
The typical procedure for research is to describe the anomaly, and then measure the performance of a long/short-portfolio that is long the 20% best stocks (with regards to the anomaly) and short the 20% worst stocks. You win a prize (i.e. being published in a journal or launching a fund) if the performance of that long/short-portfolio has strong performance.
Andrew Chen from the Federal Reserve Board and Mihail Velikov from Penn State recently examined 120 factors that are described in the academic literature. On average, the annual return of the long/short-portfolios formed on a factor had an annual return of almost 8%. That is impressive, but unfortunately, most studies don’t take transaction costs into account. So Chen and Velikov did it for them. Assuming that investors have to pay the typical bid-ask spread for the stocks they trade (the authors make sure to use historical transaction data where available instead of quoted prices that may have a narrower spread than investors pay in real life) the authors conclude that the average annual performance of the long/short-portfolio is lower, but a still-respectable 4.6% per year.
But investors today are very much aware of the replication crisis (and if you aren’t, read about it here). Thus, most investors expect the performance of the factor to decline after publication for a number of reasons. And that is what the Chen and Velikov observe as well. After an anomaly has been published, its average performance net of trading costs drops to 1.6% per year.
Ugh. It’s starting to look ugly.
But wait, there is more.
Stock market trading has changed dramatically in the 21st century. The rise of algorithmic trading and high-frequency trading has changed the dynamics of share prices and made it much easier for hedge funds to exploit market anomalies. And when Chen and Velikov looked at the performance of factors net of trading costs and after publication but restricted their sample to data from 2005 onwards, all that was left of the performance of the average long/short-portfolio was a meagre 1% per year or 8bps per month.
The average performance of stock market factors in the United States
Source: Chen and Velikov (2020).
But before you abandon factor investing altogether, let me stress two important caveats:
The chart above shows the average performance of 120 factor portfolios. Some factors continue to have much better performance than others and indeed, Chen and Velikov show that the two factors that appear in one study after another as the most reliable factors, momentum and value, still have better than average returns after 2005 (despite the underperformance of value in the last decade).
There is an increasing amount of studies that show that combining different factors in a portfolio can significantly enhance performance – something I will discuss in the future.