Jan 07, 2021
Meritage Security Selection Process Update
We are pleased to announce what we believe is a significant enhancement to our security selection process. As you may recall, Meritage uses a combination of fundamental research with quantitative modeling in our investment process to identify attractive companies. We believe this approach enables us to efficiently evaluate a large universe of investment candidates with consistent criteria and avoid the influence of emotion and our own natural biases.
Over many years, the equity models that we use have evolved as the markets themselves have evolved. Our quantitative models use ‘factors’ which measure how well a stock’s performance can be explained or predicted by a particular financial characteristic or data formula. The development of our models has been a combination of our own expertise and experience and our work with trusted external partners whose primary focus is supporting quantitative stock modeling.
Over the past year, we have evaluated several opportunities to take the ongoing research and development of our models to a higher level. While our firm was one of the early adopters of quantitative modeling, this discipline has grown significantly with the enhancement of computer processing powers and the more recent use of dynamic modeling and artificial intelligence.
We have found this opportunity in a new research partnership with the firm, Empirical Research Partners, led by Michael Goldstein. Empirical is very well-respected in the industry and we know Michael well from working with him many years ago in his previous role as the Chief Investment Strategist at Sanford Bernstein.
Empirical has a team of 10 quantitative modeling specialists focused on researching and improving stock factor models. The models they have built align well with our investment beliefs, making this an excellent fit with our modelling process, while offering a more robust capability and broader utilization. This enhanced modeling has direct application for each of our equity investment strategies.
Without getting too deep into the arcane details of modeling, here are several of the key enhancements Empirical’s work will bring to our investment process:
- Among the key factor groupings (Valuation, Earnings Quality, Capital Deployment, Market Reaction, and Machine Learning), Empirical employs what they call a market regime model, which indicates where the current market environment lies along a value-to-growth spectrum. This provides a more dynamic capability to the models as the weights of each factor group vary based on the market regime indication. This is an important addition to the modeling process as the different factor groups perform better in certain market environments than others.|
- Quantitative modeling is largely backward looking and therein lies its shortcomings. A prime example of this is the sustainability of the large free cash flow margins of some mega-cap tech companies like Apple, Amazon, and Google. Traditional models will expect those margins to fade to the industry mean over time. The market has judged these margins will prove more resilient. The Empirical models take into account views of how the world should work as well as how things are actually playing out, highlighting where judgment matters.
- The Empirical models make better use of intangible asset factors. Intangible assets, such as R&D, intellectual property and licensing agreements have become a significant part of many companies’ market value in the last decade. The use of Artificial Intelligence (AI) and Big Data are also components of the Empirical models. AI is used to recognize trends inside very large data sets that are not obvious to human observation enabling factor emphasis to shift accordingly. During periods of low volatility, AI has been effective in identifying what is currently working. Big Data has been helpful in providing new factors on investor sentiment.
We are highly encouraged by what we have seen so far from the integration with our new research partner. Outsourcing the R&D component of our factor model also frees up our own time to spend on the output of the model. We will continue to keep you apprised of our ongoing work to improve our security selection process and as always, we are glad to discuss any aspect of this with you in more detail.