AI and the Value of Information

These are interesting times, indeed. Volatility in the stock market has certainly been exacerbated by geopolitical developments in Iran and elsewhere, but there is something else afoot too. If there is an overriding theme today across all sectors, styles, and “investing” approaches, it is the effect, and affect of AI on analysis, processes, price discovery, share prices, and alpha.

And, in our view, it appears at some level that this can exacerbate the concept of overreaction. Sure, recent macroeconomic news hasn’t been great at all, and has created significant uncertainty. But AI seems always to have an immediate suggestion of what must to be done instantaneously by those plugged-in traders out there. And as far as many of these specific market participants are concerned, the good today are great and the weak are awful, at any price.

This of course leads to security mispricing, and an opportunity.  

Importantly, just because someone is using AI or some machine or code to process information, a simple, mathematical fact still holds: The net excess return of all active investing in the aggregate is zero. If you don’t believe me, just read about Bill Sharpe. As always, there will be losers and winners, and the latter will be the ones who process information thoughtfully, and without bias. This is true whether you use Claude, Gemini, or a slide rule.

AI is no panacea.

And, unless your AI is cheating, there is no such thing as an informational edge. There hasn’t been for some time. AI hasn’t changed anything here. In this excerpt from The Analyst’s Code, we reference a piece we wrote over 25 years ago, when I was just beginning my career:

Information is becoming cheaper and more accessible, and this paradigm will only accelerate with advances in technology. Thus, it will become more difficult for many to add value due to the increasing scarcity of proprietary information. In this environment, it is the “intelligence” on the margin that will separate good analysts from bad. A stock analyst should of course always ask “who might have the information I need?” but perhaps the more relevant question will be “is there information already in public hands that is being misinterpreted?”

And as we wrote twenty years later in James Harden and Alpha, stock-picking intelligence is primarily about being objective with the information that everyone already has.

Do you meet with company management? So do 30 sell-side analysts and 100 buy-side analysts and PMs every quarter. Do you build your own, bespoke, earnings models? So does half the buy-side. Do you interview competitors, customers, and suppliers? So does half the buy-side. Do you pull credit-card history, satellite images, and other big data in real time? So does half the buy-side. If you think you have a sustainable informational edge, you’re either deluding yourself, or you have inside information. We are not playing an information game. Everyone has the information. The question is how objective can we be when we process it, and might Mr Market see the same information with extreme bias?

Coalescing and uniting all of this, as we wrote in our Hunting for Alpha blog post:

A winning strategy will be one that avoids where the consensus investor is prejudiced positively, and be attracted to securities where Mr. Market is prejudiced negatively.

That’s it, that’s the recipe for long term alpha.