We were recently interviewed by a major US wire house about our Data Science expertise. Below is a summary of the conversation.
At Bristol Gate, we are passionate about data science and machine learning. We have been using it in our process for over 10 years and it is a part of our DNA.
To learn more, please contact us.
How do you see growth in AI progressing from here over the next decade or so?
Digital data is the fuel that is powering AI. AI is the next level of computing power, where machines are increasingly developing human-like thinking. I believe that we will start to see AI expand the range of jobs that are possible to automate. We have seen automation take on the highly-repetitive, low skilled jobs but it is also going to start to have an influence on the medium or higher skilled jobs as well. AI will probably make the biggest gains in areas like big data processing including the processing of language and images as well as analyzing analytical results, making decisions and adapting conclusions to environmental factors. How does this affect the broader job market? Some jobs will actually become easier rather than redundant. One example is in the no knowledge-base service industries like legal services, medical services. In the field of Radiology, C.T. scans can now capture 3D images of the body which leaves you with thousands of images to analyze and process. Machines have the advantage of being able to pour over more data however, human interaction is still needed to teach the machines how to read the scans, to make a decision about patient care based on what the computer may tell you. As a result, in 2017, it is forecasted to be the best year for hiring in radiology in the last seven years. I think that really helps makes the case for big data adding to or improving enhancing the work of skilled labor rather than replacing.
Tell us a little bit about Bristol Gate’s history of using data science and how your model works.
The process was developed based on the quantitative researchthat co-CIO Peter Simmie developed. Our first version of the predictive model used Excel using the Dow 30 as the investable universe. It was a simple regression model based on cash flow growth, ROE. This model led us to two findings: 1) we figured out that you can predict dividend growth and by building a portfolio based on highest dividend growers at the best quality, you manage to make better portfolios and 2) data science became an integral part of our investment process.
Our current model is based on cutting-edge machine learning techniques. It is a tree-based algorithm that we have fed twenty-years of fundamental data history on the broad US market into to train it and identify what factors drive dividend growth. We allow the model to figure out which factors are more important so that we remove human biases. The algorithm iteratively tries to find the best prediction but also learns from the errors of the last iteration and improves upon it for the next iteration. It learns from itself and eventually we end up with 1000 trees. We then take the average of those predictions to give us a better prediction for dividend growth for the coming 12 months. We call it a self-learning algorithm based on wisdom of crowd.
What sets you apart from some of the other investment managers?
While there may be investment managers doing similar things, our differentiator is that we incorporate both fundamental data and quantitative research into our investment process. We also have a continuous feedback loop that allows us to continually improve. We are very evidence-driven. If we have a hypothesis, we try to support with rigorous back testing and show that it works before we put it in production.
We look very different than many of the dividend strategies. We are looking for the best dividend growth at the highest quality. We tend to have a lower current yield – 1.5% – than the market which has a current yield of – 2% – but we produce much higher dividend growth. Our portfolio is 22 stocks out of the S&P500, equally weighted and sector agnostic. We are typically in 7 of the 11 sectors. We pay a lot more attention to correlation analysis because we are looking for the highest prospective dividend growers over the coming year that fit well together. We want to make sure that we do not want to own too many stocks that have similar return patterns because that introduces risk.
This post is presented for illustrative and discussion purposes only. This post should not be considered as personal investment advice or an offer or solicitation to buy and/or sell securities and it does not consider unique objectives, constraints, or financial needs of the individual. Under no circumstances does this post suggest that you should time the market in any way or make investment decisions based on the content. Investors are advised that their investments are not guaranteed, their values change frequently, and past performance may not be repeated. References to specific securities are presented to illustrate the application of our investment philosophy only, do not represent all of the securities purchased, sold or recommended for the portfolio, and it should not be assumed that investments in the securities identified were or will be profitable and should not be considered recommendations by Bristol Gate Capital Partners Inc. The information contained in this post is the opinion of Bristol Gate Capital Partners Inc. and/or its employees as of the date of the post and is subject to change without notice. Every effort has been made to ensure accuracy in this post at the time of publication; however, accuracy cannot be guaranteed. Market conditions may change and Bristol Gate Capital Partners Inc. accepts no responsibility for individual investment decisions arising from the use of or reliance on the information contained herein. We strongly recommend you consult with a financial advisor prior to making any investment decisions. Please refer to the Legal section of Bristol Gate’s website for additional information at www.bristolgate.com.