Can quantitative and fundamental approaches be successfully combined?
In my estimate, this has been a top 5 industry question for a long time, including this conference at which I’ll be speaking at tomorrow
The short answer is: Yes
More-so, I believe quantitative approaches cannot work without being guided by fundamental principles and insightful questions. Even if the model is fully technical or based on machine learning techniques, there still exist underlying ideas and assumptions on which it is built.
At the same time, I believe fundamental analysis can only work over a long time-frame when expressed via some unique, structured and evolving framework.
However, I don’t think that a 50/50 approach can work because of the different nature of the two investing methods: one attempts to forecast baskets of stocks, while the other attempts to forecast individual stocks.
The following list of observations is based on my experience of successes and failures of mixing quant and fundamental approaches
Collaborative Approaches That Can Work
Fundamental portfolio managers can be great at formulating an investment framework
Quants can be great at building and testing it
Fundamental analysts can be great at asking powerful questions
Quants can great at answering them
Fundamental analysts can have deep sector and industry knowledge
Quants can be great at modeling it with large amounts of data
Fundamental analysts can recognize a strategic ‘aha’ moment for a business
Quants can find ways to measure it across time and other companies
Fundamental analysts can be great at ‘connecting the dots’
Quants are good at removing emotions in constructing the final portfolios
Approaches That Don’t Work
In cases when Quants support a Fundamental Portfolio
Asking quants to build ‘screening’ models from which fundamental analysts pick stocks (the hit-ratio is too low for this to work)
Asking quants to justify a factor definition to a fundamental analyst (it will always sound too primitive or datamined)
Asking quants to explain the model on any given stock (too much noise)
Asking quants to generate alpha by giving them an ‘off the shelf’ academic paper (great quant is based on innovation)
Looking for consensus between quant models and fundamental opinions
In cases when Fundamental Analysts support a Quantitative Portfolio
Asking fundamental analysts to rank stocks to be used by as a factor
Allowing fundamental analysts to over-ride models, change weights, impose tilts, remove individual names in the final portfolio
Asking fundamental analysts to correct their biases
Asking fundamental analysts to come up with a factor
Vague attribution of decisions, including the cost of rejected ideas (type 2 errors)
Some “Final Questions”
Can fundamental analysts respect quant’s reliance on the law of averages even if many times on individual stocks the output looks like rubbish?
Can quants respect fundamental analysts’ depth of analysis even if many times it is not correct?
Can fundamental analysts be open to mentoring and sharing their thought process with quants
Can quants be honest about the balance of data-mining and innovation that they strive for?
Can each side become great at what it does rather than try to win over the other or strive for consensus?