Quantcon 2016: Sustainable Active Investing

Hello Community,

For the latest release of QuantCon videos, I've created a notebook that explores the predictive nature of value ratios.

This notebook focuses on a concepts presented by Dr. Wesley Gray of AlphaArchitect. Check out his presentation slides and watch his talk before experimenting with the notebook. In his talk, he explores behavioral bias and long term active investing. This notebook takes these concepts and creates a framework for quantifying the differences in short term and long term predictabilities.

Using the Morningstar fundamentals, we can select the cheapest stocks by their valuation ratios. We would like to show the short term predictability of a factor doesn't necessarily match its long term predictability. To do this, we'll create a pipeline factor and use the Factor Tearsheet to view the returns for a valuation ratio over a various time windows.

If the Information Coefficient(IC) taken at 1,5, and 7 days are near zero, then we can say that a valuation ratio doesnâ€™t predict near-term returns. If the IC taken over 1,2, and 3 years is high (|x| >0.5) , we can conclude that (for the equities sampled) a valuation ratio is a good predictor of future returns in the long term. Finally, if short-term IC is near zero, while long-term IC is high, we can conclude that a long term value strategy is more susceptible to short term noise.

The valuation ratio that I focused on is the same that Dr. Gray uses in his presentation: EBITDA/EV. Other ratios are available in the Morningstar fundamental data. In Dr. Gray's presentation, he found that using value ratios are a simple way for long term positive returns, but don't always predict short term returns.

In this notebook, I reevaluate Dr. Gray's findings with a new metric: calculating variable-window information coefficients in order to quantify this valuation ratio's predictive ability of short and long term price movements.

There are a few open questions left to tackle.

1. Is there a ratio or formulation that has a reasonable information coefficient in both the long and short term?
2. What are the minimum and maximum IC values that you can get using a combination of these ratios?
3. What would an algorithm that applied these findings look like?

Best,
Lotanna Ezenwa

117
Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

5 responses

Lotanna - this is an impressive notebook with lot of work put it into it. Good work!

I am not as familiar as I would like to be with Spearman rank-order correlation coefficient and Information Coefficent.

I was wondering if you could expand a bit on this sentence in the "Conclusion" section of the notebook : "A 'good' IC is considered to be anything greater than 0.05. "

I was under the impression that the IC could range from -1 to +1, and that the closer the value is to zero the less correlation there is but I could be mistaken. Can you share any perspective or reference on the "0.05" part of the sentence above?

Richard

Great stuff. You might want to tell people that they need to execute the bottom cell first to use the notebook.

What happens if a stock has poor or average fundamentals? Do they come to different conclusions?

@Richard

Excellent point. The IC ranges from -1 to 1, and values near zero represent a prediction that is no better than random. This correlation is of the relative percentage changes of one equity to another. So what one should be looking for is a an absolute IC that is consistently above zero at least some amount. In my opinion, a factor with a mean IC whose absolute value is less than 0.05, for this sample size and time, could be greatly influence by noise and requires some more rigorous testing. Admittedly, I'd hope more testing goes into factors in terms of sample size and time frame.

Allow me to elaborate on my conclusion. If you take a look at the 252 day IC from the second set of graphs, you'll notice the 1-month m.avg IC rises consistently to around ~0.4 then, drops sharply around July 2011. In this sense, the predictive power of the factor was very strong, then encountered some sort of regime change or market activity. The 756-day IC, however appears to be more consistently positive, showing some predictive effect.

@James

Great question, as of right now, the graphs only show the mean returns by factor quantile. In general, the mediocre fundamentals are all over the place.
I think you would need to edit the factor tearsheet to remove outliers and run the IC calculations to find that out.