Explainable AI - A solution that will not offset hard work

Try and explain artificial intelligence to someone who is not well-versed in the mechanics. It is not easy. It is not even easy for a strong quantitative analyst. You may know the math. You may be able to effectively program the model. You may understand all of the data inputs, but you may not explain what the model is doing. The argument that you just have to live with "hidden layers" may not cut it from a risk management perspective or if performance goes wrong. 

Help is supposed to be on the way through a new area of research called explainable AI or more precisely interpretable AI. I was hoping for a new form of AI clarity when I started reading this research. No such luck. (See "Peaking Inside the Black-Box: A Survey on Explainable Artificial Intelligence  (XAI)", IEEE Access.) Investors interesting in using AI and machine learning will have to do the heavy work of learning, testing, and expending the sweat equity to understanding what is going.

So, what is explainable AI or xAI? There is no generally accepted definition but is referred to the movement to increase transparency and trust within AI. It is a broad set of techniques or approaches to reduce the obscurity associated with "black box" techniques. There is no simple model that will breakdown the drivers of a model, but looks at simplification, marginal impact and scenario examples to increase our understanding of the model and output. It has a goal of increasing explainability without decreasing forecast accuracy.

More complexity reduces explainability. Hence, interpretable models usually come at the cost of reduced accuracy. The idea behind explainable AI is to reduce complexity or show the additive value of complexity. This does not reduce the difficulties of properly using data or provide some magic easy to read procedure but tries to focus on procedures to link data relations with prediction. There is less of an issue of "black boxes" but complex boxes that focus on non-linear or deep relationships not immediately obvious. The benefit of AI is in its ability to find what in not immediately obvious. 

UnfortunateIy, I would say there are still difficulties associated with "explainable" regression. The strength of visualization software and simulation tools help, but any increase in interpretability will be related to basic knowledge and increased usage. 

Usage in many organizations is generational. Old management must be replaced with new management that have employed these techniques as part of their normal decision-making toolkit. This knowledge transfer is faster for smaller organization (hedge funds) and slower for traditional money managers. So for the near-term, crack open the textbook, learn the coding, and runs some models, there is not easy alternative or "free lunch".

Sector ETFs are an important hedging tool

It has always been known that ETFs serve as a hedging tool, but a recent research presentation explores this issue more precisely and shows that many short interest spikes are driven by hedging and not an industry view. See "Innovation and Informed Trading: Evidence from Industry ETFs" by Shiyang Huang, Maureen O’Hara, and Zhuo Zhong Darden Mayo Center Virtual Seminar April 2020.

Assume a hedge fund finds an attractive stock to purchase. Instead of buying outright, selling against another stock, or hedging against the market portfolio, the best alternative may be to sell short an industry ETF. The industry basket hedge is easy to execute and is a useful tool since there are not industry index futures available. Hence, the short interest in an industry ETF may not tell us anything the absolute or relation performance of the industry group. A spike in ETF shorts is often just a signal of hedging in order to purchase a stock that is likely to a have a positive earnings surprise. 

The researchers explore these hedging details through looking at specific times when there are combinations of stock picking and industry ETF shorting. First, they look at abnormal long-short hedge fund holdings prior to earnings announcements. The long-short trade of buying the stock and selling the industry index before a positive earnings announcement shows how hedge funds use the ETF as a hedging tool and use their stock specific skill at making earnings bets.

This is a subtle test of the value of ETF hedging, but it displays the power of ETFs as a multi-faceted tool for investors to both display an industry view and employ an inexpensive hedging tool. Industry ETFs help to make markets more efficient. 

Value has underperformed, but there is not a strong relationship with declining rates

There is the popular view that value relates to the interest rate environment. Rates have been low since the Great Financial Crisis, so it has been viewed as the culprit or at least a suspect for the poor performance of value. Nevertheless, the long-term empirical relationship between value, rates, and slope of the yield curve shows mixed results across time and measures of value.  

These conclusions on the impact of rates on value are presented in a new paper, "Value and Interest Rates: Are Rates to Blame for Value's Torments?" by Thomas Maloney and Tobias Moskowitz. A strong relation between value and rates should be stable across time and across measures of value. Unfortunately, a stable statistical link was not found in their study. While there may be a strong short-term relationship, the behavior today is not consistent with value and rate relationships in the past. There is no consistent in the relationship between rates and different definitions of value. The same applies with the slope of the yield curve.

If the value of stocks is relation the discounting of cash flows, there should be a relationship between the discount rate, when cash flows are received and value. In general, growth stocks will likely have a long duration while value stocks could be viewed as a shorter duration equity, all else equal. This difference should be seen in their performance with respect to the level and slope of interest rates. Nevertheless, a financial distress argument would draw an alternative conclusion. The empirical numbers in this research do not show a clear relationship even after studying different value measures both in the US and internationally.

The tests do not show a statistically significant relationship with the level of short rates, long rates, or slope across a number of value measures. There is a relationship with changes in short rates and yield curve slope although not for all value measures. The same conclusions are drawn from international data. An analysis of the relationship through time shows that value is more rate sensitive now, but the relationship is not stable.

Looking for rates as the rationale for underperformance with value investing is not fruitful and other reasons for the poor value returns are necessary.  

Benchmark allocations have a large impact on performance - Dispersion abounds

Investors may not predict the future, but that does mean they should be indifferent to portfolio choices. Choices matter. A review of some well-defined benchmark comparison shows the dispersion in returns over the last twelve months. We have looked at the difference between a higher return strategy versus a lower return strategy. 

The largest return differentials over the last twelve months are between growth and value, in the favor of growth, the gain of large cap over small cap, and the information technology sector over the equity benchmark. These differentials should not be surprising. Information technology generally has been immune to the COVID19 Lockdown versus classic brick and mortar or cyclical equities. Small caps have not been able to gain scale during this downturn, and growth continue to outperform versus value. 

However, there are some surprises. There is little difference between the high beta and low volatility benchmark portfolios and little difference between the benchmark SPX and a momentum portfolio. The health care sector, mainly pharmaceutical and medical device companies, has outperformed the benchmark in spite of mass reduction of non-COVID19 health care expenditures.

What is most surprising is that the 12-month total return for the equity benchmark (SPX) is 12.84. An investor who fell asleep a year ago and woke-up today would feel as though the global investment environment is in a good place and the global economy is likely stable. Little would suggest the problems we are facing.