You are well-known for your research on the equity risk premium and its predictability. Can you shed some light on your findings as well as your views on market timing?

“In our research paper,¹ we found that it is very difficult to time investment allocations between stocks and bonds, for example. It is challenging to figure out which asset classes will do well in the next month, quarter or year. We are now in the process of updating that study² and have observed that our findings still hold, even when using new variables. In essence, we conclude that market timing does not work.”

“But I do want to emphasize one point. Some readers interpret our paper as stating that predictability does not work, but this is a larger and wider concept that is not fully covered in our research. For instance, the scope of our study does not include how predictability works within bond, commodity or equity markets in terms of determining which securities might perform better than others. In fact, I believe there is a lot of predictable variation in the cross section of bond, commodity or stock returns.”

How does market timing apply to factors?

“I am aware there is a debate on this topic. I have not personally conducted any research on the subject, so it is hard to give you a concrete view on the issue. That said, I am inclined to believe that it is also tough to figure out if, for example, value will continue to do well in 2023 as it did in 2022, or struggle as it did in the second half of the 2010s. The same applies for momentum or any other factor. So in singing from the same hymn sheet, I generally believe that timing is difficult to pull off.”

What do you believe are the main opportunities for factor investing?

“Factor investing is a mature concept in the realm of equity investing. But this is not to say that this field of research is at a standstill. A lot work is still being done in the space, new factors are still being documented, and maybe there will be a synthesis of similar factors. However, it is still a nascent investment style in other asset classes. There is no widespread understanding of factors in bonds and certainly less so in commodities or currencies. In my view, this is the next frontier of factor investing.”

So you believe factors will become mainstream in fixed income as they are in equities?

“I certainly believe so. Academics and practitioners typically view fixed income through the lens of several dimensions such as the type of instrument (sovereign versus corporate bonds), the associated risk profile (investment versus speculative grade), and the related maturity profile (short term versus long term). This makes fixed income ripe for factor investing in my opinion.”

“The reason why this has not happened sooner is probably because, compared to equity markets, there was less fixed income data for academics and practitioners to work with, and therefore fewer insights that could be gleaned. Nowadays, however, there is an increasing number of valuable databases, such as TRACE. Academics and practitioners are therefore starting to pay more attention to fixed income factors and more research is being done.”

What do you make of the rising wave of machine learning in the asset pricing literature?

“In principle, machine learning appears to be a good development. Neural network or random forest algorithms, for instance, can detect patterns in financial data that a human brain cannot. These can then be used as signals to determine which stocks to buy and sell.”

“But I am not a big fan of blindly following a machine in determining investment decisions. It will find patterns in the data because that is its strength, but it will not tell you why the signal is proposing the purchase or sale of a stock. That is just the nature of unsupervised learning.”

“Applying machine learning in academic research is a more fruitful avenue, in my view. There is scope to use machine learning techniques to design factors that can better explain the cross section of stock returns or get an investor closer to the tangency portfolio.”

“From a practitioner perspective, machine learning can also be used to analyze big data and different types of data, such as the cliché example of parking lot satellite images to determine retail footfall. In this context, machine learning used with big data can lead to better or faster signals. But again, I believe there has to be some element of supervised learning.”

Unlike many finance researchers, you have exhibited an interest in market frictions such as transaction costs. What can factor investors learn from your research?

“There are two points I would like to mention. The one is related to research that I am currently working on regarding trading venues. For a long time, people only thought of actioning trades, breaking up orders, and so on during regular trading hours, more or less between 09h30 and 16h00. But over the past ten years or so, closing auctions have also become an important venue for trading.”

“For instance, both the New York and Nasdaq stock exchanges hold closing auctions at 16h00 and these account for about 10% of the day’s volume. This is significant, especially since our ongoing study has revealed that transaction costs are much lower during these intervals. As such, we suggest that investors should look into this as an additional option in terms of forming a trading strategy.”

“The other point is related to a study based on work other researchers have done on integrating transaction costs in portfolio optimizations. While this concept has been discussed in the academic literature since the 1980s, recent research is exploring new implementation methods. Factor diversification is one area that has drawn attention in this regard.”

“Using a simple example, one factor could propose a purchase of Apple stock while another factor could propose the sale of the same security. In a multi-factor solution, these trades could be crossed to minimize transaction costs. Such techniques can potentially yield significant benefits and enhance the capacity of multi-factor strategies. Indeed, there is now increased awareness and research on how to integrate transaction cost minimization techniques within portfolio construction processes.”

What are you most excited about these days in terms of research?

“Broadly speaking, there are two main themes in my current research. The first is to look at how endowment funds, pension funds, etc. invest. The second focuses on factor investing across asset classes.”

“Regarding the first, I think we need to get a better understanding of how asset owners go about their decision making. Although asset allocation theory extends back to the 1970s, I believe we do not know enough about how asset owners allocate capital across asset classes. There are many more questions, and not just limited to equities and bonds, but to other asset classes. For instance, we can study how asset owners award private equity mandates.”

“On the second point, I believe factor investing across asset classes is the next El Dorado. This is often referred to as ‘alternative risk premia’ in the industry. I believe a lot of research can be conducted on this front and some academics are currently exploring this area. So maybe we will come across some interesting insights in this space over the next few years.”

Footnotes

¹ Goyal, A., and Welch, I., July 2008, “A comprehensive look at the empirical performance of equity premium prediction”, Review of Financial studies.
² Goyal, A., Welch, I., and Zafirov, A., September 2021, “A comprehensive look at the empirical performance of equity premium prediction II”, Swiss Finance Institute Research Papers Series.