A qualitative guide to the quants

The writer is chief executive of Martlet Asset Management, an adviser to the FDP Institute and retired co-founder of PAAMCO

It has been a humbling few years for many quantitative fund managers. The 2018-2020 period has been described as a dark winter for the sector as many of the investment strategies pursued by these data-driven fund managers struggled.

Despite a pick-up this year, I am often asked if I still have faith in these strategies after having spent more than 35 years of my life designing and analysing them while investing in the sector. My short answer is yes, but with certain caveats.

First, not all managers are created equal. Quantitative strategies encompass a wide variety of styles, markets and philosophies. Unlike in the 1980s and 1990s where merely the introduction of structured data and computing power offered a competitive advantage, today a firm labelled “quantitative” doesn’t give much insight into how the money is actually being managed.  

One type of quantitative manager could be called the fundamental quant on steroids. This manager is philosophically a close cousin of his non-quantitative counterpart but brings significant “enhancements” to traditional processes. These include being faster and having a larger scale.

A second type would be loosely based on artificial intelligence or more narrowly on machine learning. This type of manager often uses incredibly large data sets with complex algorithms guiding positions rather than economic theory. In essence, this is akin to sophisticated data mining. Performance in these strategies tends to struggle when there is a shift in the economic backdrop. 

Finally, another type of quantitative manager is almost a fund of funds of individual algorithms combining dozens of different styles and anomalies. Within reason, whatever makes money on a risk-adjusted basis is added into the blend. For example, there may be one algorithm that uses alternative data sets such as the number of cars in a supermarket parking lot. Another might focus on models of market microstructure. Given the number of algorithms, these firms have tended to produce more consistent returns, albeit often with significant levels of leverage.

Given the diversity of approaches, what are the things to look for in evaluating these quantitative managers?

The first question involves the true robustness of the investment process. While all investment managers are liable to reinterpret past events incorrectly, this mistake poses a key risk with quantitative strategies. 

Any halfway decent data scientist given large data files and enough time should be able to find “relationships” among the variables. The issue is whether these relationships are just the natural consequence of having so much data that spurious results arise. Remember the celebrated theory linking stock market returns with the winner of the Super Bowl? Convincing oneself and one’s colleagues that there really is a relationship has happened to all of us. 

The second question focuses on what percentage of the portfolio is driving the returns. In traditional security analysis, it is common to see a few positions drive most of the return. However, almost all quantitative strategies are built on portfolios where the return is generated from a repeatable process applied over and over to many different securities. Assessing how the returns are distributed not just by strategy but by the number of positions is vital.

The third area of focus is the size of the firm versus the capital base. Some of the best managers I have hired are relatively small and trade narrow niches but also limit their capital. While larger firms can clearly throw more against the wall, they also have the challenge of a larger asset pool to invest.  

Finally, the largest leap of faith or judgment that any investor must make goes back to the concept of equilibrium. Most quantitative strategies involved purchasing securities which are “mispriced” and waiting for them to return to equilibrium.

But a traditional manager might invest in concept where the equilibrium is not stationary. For example, one investment theme might be that the world is moving towards more online payment systems. The manager in this instance likely does not graph the historical average of online payments and extrapolate. Such an approach is much harder for most quantitative managers. 

So, where does this leave us? Quantitative managers and the techniques they use are not broken. For every quantitative manager that struggles (and most do at some point), there likely is another somewhere taking advantage of it.