In an effort to get to know the Fintech ecosystem and their representatives better, we asked them a few questions as a part of our series of interviews: The Innovators. This time we caught up with Alexandre Colin, the CEO of Behave!
Alexandre, can you tell us a little about yourself and about your company Behave?
I have been living in Luxembourg for 23 years now, started my career as an equity fund manager, running a growth fund at the end of the nineties. After the dot-com crash, I wondered if sell-side equity analysts had any value, as more than 65% of them had Buy recommendations versus 4% of Sell between 1996 and 2002 as shown in a well-known article of Boni and Womack. In fact, the change in the level of recommendations was the factor to watch. With this in mind, I started my own company a few years later and offered a first web platform exploiting behavioral anomalies like this one. Behave! was subsequently started in 2012 as Bloomberg contacted us to launch a duplicate of our existing platform on the newly created App store on the Bloomberg Terminal. We were one of the first 20 apps to be sold there, and ranked among the Top 5 apps in their charts for some time.
Behave is based on quantitative investment strategies, which are guided by sets of rules, aiming to exploit market abnormalities in order to improve returns, control risk, or diversify portfolios. The general consensus among asset managers appears to be that quant strategies can only play a complementary role alongside traditional fund management models, not a replacement. What do you think of this perspective?
I couldn’t agree more with this statement. We target the so-called “quantamentals” : asset managers combining their fundamental analysis with quant signals. We aim at generating alpha in a smart beta world. Our platform enables them to easily integrate quantitative analysis in their investment routine, generating buy and sell signals that are statistically significant within their investment universe (the normality of the distribution is another issue).
If I tell you “People, not computers, are one stock fund’s winning formula” what is your answer?
There are a lot of arguments against this statement. Never forget that 80% of active fund managers underperform their benchmark at a 5-year horizon. Not to mention survivorship bias, and not mention that this figure has been a lot worse over the last few years. Next to costs and regulation, this has been a catalyst for the growth of ETF and systematic strategies. Also, humans are subject to behavioral biases. The main goal of an asset manager is not to generate the best returns for his clients, it is to keep his job. That’s the agency theory. This leads to different investment decisions, and by the way explains the Value anomaly to some extent. Truth is, the goal is to make better informed decisions. Hence I will tell you they are both part of the winning formula. Systematic strategies are built based on the knowledge of the market as of today. If I build a model on our behaveSFI platform, I can exclude the survivorship and forward-looking biases – and I could tell you that they are still caveats to that statement – but it is based on my knowledge of the market as of today. So in a way, there is a forward looking bias by design. And it certainly would not be smart to ignore our market knowledge as we build models. Going forward, we could expect artificial intelligence and deep learning to generate investment strategies that will evolve through time and spontaneously take into account newly relevant data. But we are not there yet.
On the 3rd of September you will attend the European AI for Finance conference as part as the LHoFT delegation. How would you qualify the use of Artificial Intelligence technology in your models?
Moderate. I saw a cartoon the other day where a guy was asking his friend “Have you heard of how much progress was made in artificial intelligence?”, and his friend replied “And what about real intelligence?!”. This tells it all. We use machine learning to forecast how investment styles are going to be rewarded over the next few weeks. Feed your model with several dozens of well-known factors and you get nothing. Now think about which factors could influence, be correlated or be useful indicators in predicting how a style is going to be rewarded and you start seeing more interesting forecasting signals. On our platform, we offer our client style forecasting signals on various investment universes, which can subsequently be tailored on demand to their investment universes.
Behave was ranked 1st among all quant providers over the last 2 years for your long-short model performance (Investars, 4th of June). What does this mean for you and Behave?
And even better, our long only portfolio (first decile of the S&P 500 based on our proprietary ranking) is currently first over the last 6 months, hence including the recent market turmoil, as the longest bull market on record is showing signs of weakness. So it means that our clients can safely rely on our scoring if they are looking for turnkey models to screen their investment universes, or build on our scores to add new or custom factors.
With the Bloomberg App, you had clients in various countries around the world. What are the next steps for Behave with this new platform?
With the Bloomberg App, we only offered our proprietary models. With this platform, our clients can access the Refinitiv databases, formerly known as Thomson Reuters, including ESG data, a world financial data leader, to build their own models, use our proprietary scores derived on their data, and integrate our AI signals. No coding capability is required to build your models, and last but not least as margins are pressured : it is cost effective. And while we target quantamentals, strategies can also be applied systematically.
Find Alexandre and his team at the European AI for Finance Conference on 3rd of September in Paris on the LHoFT stand
