Monday, May 29, 2017

Avoid atheoretical data analysis - a rule for any data specialist

The process of scientific discovery even within finance is important. One approach to finding new strategies could be to generate observations and then provide explanations, an inductive approach. The other is to first form a theory and then test a hypothesis, a deductive reasoning.  Much of machine and statistical learning is inductive reasoning where data are used to suggest general hypotheses. Deductive reasoning is used with experts systems where rules are made and then tested against the data.

The surge in data analysis is based on a belief that inductive analysis will be able to identify new relationships that may not have been previously hypothesized. The danger comes when the analysis is atheoretical. This has been given a name, HARKing (hypothesizing after the results are known). 

There also is mining of data for anomalies or risk premium that may not exist. Data can be tortured until it generated some level of significance, "p-hacking". A relationship is found, but only after the fact is it given an explanation. From data come stories not the testing of ideas. Call it meaning without structure. Now, we don't want to put all of these techniques on a trash heap to be ignored, but there is important room for experts and practitioners to guide and interpret what the data mines are producing. 

If data suggests a relationship but there is not an easy story to tell, the relationship should be suspect. Good modeling and data analysis tries to test hypotheses and stories and has an idea of what could be possibly found in the data. Can there be new surprises in data? Of course, but those should be the exception no the rule with machine learning.

Friday, May 26, 2017

Looking at asset class risk through boxplots


Investors are so used to looking at standard deviation to define risk that they forget some easy exploratory data analysis tools that can be very helpful. The boxplot focuses on a greater description of the data through a simple display of a brand array of information. The box is formed by the first and third quartiles, the whiskers are 1.5 times the interquartile range, the green diamond is the average, and the black boxes are the outliers. 

We have looked at the last three years of monthly data and can make some simple observations. This list can be much larger. First, most asset classes are subject to some outliers and there is a lot of movement going on outside the box. Second, risks can vary greatly across asset classes even between US and international stocks. Third, there are significant differences in fixed income risks. The diversified lower duration Barclays Aggregate index is much less risky than holding the long bond. Finally, a portfolio of managed futures as a representative alternative asset class shows a much tighter distribution of returns. Portfolio risks may be more concentrated in the traditional assets. We are not making a case for managed futures, but rather highlight the richness of data that can be gleaned through simple tools beyond standard deviation.

This follows the old adage - before you model it, plot it. If you don't have a feel for the data, it will never talk to you or guide you.

Wednesday, May 24, 2017

Size matters with managed futures but not just with performance


There has been a preference for large managers within the managed futures space as measured by money flows, but it comes at a cost. Looking at year to date performance from the CTA Intelligence performance database shows the differences in performance based on size. What is clear is that the average performance of a set of 40 small managers is significantly higher than the performance of the largest 40 managers. Going down in size will allow clear increases in average return and with median returns; however, a closer look shows that the price of obtaining higher returns can be high in terms of regret. 

The standard deviation as one measure of performance spread shows that as you move down in size the range of performance gets drastically larger. One standard deviation on the downside will wipe out the average return for all categories. However, performance does not seem to be normally distributed. There is skew to the upside. More importantly, the range between the high and low manager within a category is multiples larger than the average. 

While many may look at managed futures benchmarks, those returns are not usually representative of what you may get from choosing any one manager. There can be large discrepancies between an equally weighted portfolio and an individual manager. This is the price or cost of due diligence. If the wrong manager is picked, there can large downside costs. If right, the gain can also be high. Even in one quarter, a bad pick can be multiples away from the average.

The advantage of buying a large manager is with the clustering effect with performance. The cost of being wrong is minimized. So investors who pick from the large managers may not asking for something better but something that is less likely to be out of step with the rest of the group. Call it safety-first and performance second. Going down is size allows for better due diligence to be rewarded. It is a trade-off. 

Trading gold - CME vs. LME - This could be an interesting battle


With a new regulatory environment, the traditional gold market is up for grabs.  While many of the changes have not been direct to the gold market but based on a broader environment, regulation is changing the market structure for trading in gold.  Regulations which increase the cost of trading over-the-counter has and will push gold trading onto exchanges, centralized counterparts (CCP's). This environmental change is the opportunity for new exchange entrants in the gold market. We have already seen new trading through ETF's over the last few years change gold dynamics and now there is the opportunity for changes in futures trading.  

For many hedge funds and managed futures traders, the main point of access for trading is the futures exchange. In this case, the COMEX exchange of the CME. However, the LME wants to become a rival through new gold futures contracts. The question becomes whether the move to exchange trading gold around the globe will allow for another exchange to take advantage of the opportunity and co-exist with the CME or take market share. History has shown that there are few cases of successful entry into the futures markets relative to an established marketplace. Regardless of the regulatory environment, it is hard for new exchanges to break the network economies created from a successful trading forum. There has to be something different to attract traders. 

What makes this time different is the changes in regulation that will increase the costs for major banks to trade gold OTC. With higher risk weights, credit valuation adjustments, higher margin requirements for non-CCP trades and for longer margin period of risk (MPOR), exemption of leverage ratios for CCP trades, standardization, and best execution standards all affecting OTC trading, there will be movement to exchanges. The issue is how much, how quickly, and where. The COMEX/CME has served many participants well but is a standardization model different than what has been used by the LME. 

The LME would like to threat the needle and provide something that has the concentrated liquidity of the COMEX and the features well-known to industry LME users. Hence, the "hybrid" approach of LMEprecious will allow for more flexible trading.  With daily dates tradable from T+1 to T+25 and third Wednesday monthly prompts for 24 months and another 12 quarterly dates, the LME contracts will provide more flexibility than the COMEX. The question is whether the cost can be reduced enough to facilitate the switch from the OTC market.   

This will be a battle worth watching.