Can robots invest better than you or I can?
That’s the fundamental question asked — and only partly answered — by six researchers who built an investment algorithm. The researchers compared the algorithm’s performance with the returns of 255 angel investors. The algorithm used the same data — pitch material, social media profiles, websites, etc. — that was available to the angel investors at the time.
The study, whose results were published in the Harvard Business Review, divided angel investors into two main groups. The inexperienced group averaged five investments each. And the experienced group averaged 12 investments per person. The algorithm did much better than the inexperienced group — not so surprising. But it did much worse than the experienced angel investors.
The results are even more interesting when you factor cognitive bias into the mix. Elite experienced investors, which the study describes as those “experienced investors who can suppress their cognitive biases,” performed the best in this study.
|Average internal rates of return (IRR)
|Elite experienced investors
|Experienced investor with cognitive biases
|Inexperienced investor with cognitive biases
It turns out that cognitive bias really hampers investors — even experienced ones. Experienced investors who showed a high level of cognitive biases only averaged a 2.87% IRR. That’s significantly worse than the algorithm’s 7.26% IRR return — even worse than the 3.51% the less-biased novice investors achieved. And novice investors with higher biases had an IRR of -20.52%.
One of the main things the algorithm did was try to eliminate the five major “cognitive biases” identified by the study’s authors: They were:
- Bias toward investing locally
- Sensitivity of incurring loss over making gains
- Overconfidence leading to investing too much in a single startup
- Gender bias
- Racial bias
It’s hard to say whether the algorithm successfully eliminated bias — though it did clearly outperform experienced and inexperienced investors with biases. But it is clear that cognitive bias plays a critical role in investing — and that experience does not eliminate it.
And that’s really counterintuitive. In theory, as you gain experience, you should lose your bias. But this study shows that’s not necessarily true.
More than anything, the study reveals how important it is for early investors to have an open mind and a willingness to learn from their mistakes and biases.
If you’re from Cincinnati, after looking at 1,000 startups, you should realize that not all the best startups come from Cincinnati. Counterintuitively, early investors should become less confident as they gain experience — because their portfolio will have more losers than winners. As investors learn through experience, their other biases should gradually fall away.
Algorithms can have a role in startup investing. They can help point investors — especially inexperienced ones — to startups they would not have otherwise considered. For experienced investors who have learned to compartmentalize their biases, today’s algorithms may not be as useful.
But the algorithms will improve over time. I’d be shocked if I don’t have an algorithm or two in my toolbox a couple of years from now. I expect they’ll help me make decisions — but not take over the entire decision-making process. Some people might appreciate the ease of having algorithms choose all their investments. For me, that day will never come. It’s the process — the journey early investors take — that makes it fun. I doubt I’ll ever give that up.