A new Harvard study found that AI could predict 71% of stock trading activity made by active fund managers from 1990 to 2023. Basically, machine learning could tell whether a manager would buy, sell, or hold a given stock or fund in a given quarter. And that spells trouble for the pro-active management crowd. And puts another feather in the hat of passive investing, index fund proponents.

Because the whole thesis of active investing is that managers can generate alpha, or above market returns, through analysis, skill, and innate talent. However, this study suggests that, in reality, active trades more so reflect industry standards and systematic, habitual behavior.

And all this means that the stock market is maybe even more efficient than we give it credit for. So trying to beat it is a fool's errand. So, did AI just end this debate?

Let's explore.

A Brief Overview of the Harvard Study

Basically, researchers looked at quarterly mutual fund trading decisions from 1990 to 2023 using AI machine learning models. They added in factors including fund size, investor flows, stock characteristics, and economic conditions.

And the machine model predicted around 71% of all trades.

So, what about the remaining 29% of trades that it didn't predict? Well, those were the trades that were associated with outperformance, or alpha.

And what is the takeaway here?

Well, it means that your routine trades are actually really predictable. But the ones that would create outperformance are rare and hard to model. Even though this is just a first crack at such a model and we would expect that future, more developed ones would do an even better job at such predictions. Essentially, this model, the first draft of it, became your average active fund manager and learned to act accordingly to market flows, signals, and the behavior of other managers.

What Does This AI Study Mean for Active Stock Trading?

That's the real question, right? Well, it means a few things.

Most importantly, it means that most “active” stock management decisions aren't really unique. Again, the model was able to predict these trades because they were predictable. I know that sounds like a Yogi Berra quote but it is worth emphasizing.

So, why are they predictable?

There's a bunch of potential reasons. Most fund managers are going to respond similarly to the same signals. Most analysis is going to generate the same signals. And, as we know, neither those signals nor the managers reacting to them can accurately predict future returns. But it is still human nature to try. And this leads to active fund managers largely following similar strategies and adjusting their (and their clients') portfolios in similar, predictable ways.

Moreover, fund managers face a lot of pressure that essentially leads to professional herding. They face quarterly benchmarks, expectations from clients, and the like. The result is that most fund managers tend to follow each others' actions, whether consciously or not. Remember, as humans we are loss averse. We feel losses way more than similar gains. So any manager worth their salt will know that, while their ultimate goal is to generate alpha, failure to do so is not the worst outcome, losing your money is. In fact, they get compensated pretty similarly regardless of what happens…another conflict of interest and herding pressure.

All of this adds up to one rational conclusion, if most active trades are predictable, then, as one Harvard researcher put it, it becomes (even more) difficult to justify active management fees.

Why Predictable Behavior Implies Efficient Markets

The whole tenet of passive investing strategies is that the stock market is so efficient that it is effectively impossible to outperform it. Any nugget of information that could lead to a trading advantage on a stock or fund is nearly instantaneously absorbed into the price of that security. This happens because information spread quickly and market managers and investors react quickly. So any opportunities disappear just as quick as they appear. Blame the Internet, your smart phones, and instant trading apps.

Remember, this doesn't mean that some active trades don't generate outsized returns. Some do. Even if most don't. It's just that no one or thing has been able to predict which ones will hit and which ones will bust more than plain, dumb chance would dictate.

And I can hear you asking, “But what, doesn't this study show that AI can predict the market?”

And that's the big difference. No, it does not. It shows that AI could predict how humans would react in their attempts to predict (usually incorrectly) the market.

The study shows exactly this dynamic. If trading patterns are visible to other market participants, competitors, and (yes) AI algorithms, then those same people and systems can anticipate and arbitrage them nearly instantaneously.

And we are back where we started. With efficient market theory and the concept of an extremely efficient market where any behavior gets priced into the markets.

Why Active Trading Is a Losing Game

The study highlights several reasons active trading is so difficult.

Any competitive edge is tiny. Only a minority of unpredictable trades generated excess returns. And remember, the trades were unpredictable because AI could not predict any human behavior that efficiently or effectively predicted them. This tiny advantage holds for all investors, institutional and individual.

But the odds are stacked even higher against individual investors. Every move is analyzed by hedge funds, quant banks, large trading platforms and brokerages. Add in machine learning and AI.

We are not going to win.

The Real Lesson: Investing Should Be Boring

And even that is a misgiving statement. Because all of the aforementioned are also going to lose. Because active stock management has not ever been able to consistently beat the market return. And then add in the additional taxes and fees generated through active trading strategies.

If AI can replicate most active stock trading based on its analysis of humans trying to predict the market, then active stock trading isn't creating alpha, it's just creating expensive portfolio maintenance. Because when 71% of active trades can be predicted by an algorithm, it means the industry’s “secret sauce” is often just a recognizable pattern. And patterns don’t survive in competitive markets. They get arbitraged away immediately.

That is why passive investing is the winning strategy. Invest broadly with low costs for the long term. Don't try to predict the stock market. Instead, just buy those whole market so you don't need to predict it to be successful.

It's all very zen when you think about it.

For more on market theory and the wise way to invest for financial freedom, check out these resources:

• 5 Active Investment Decisions Physicians Are Making
• 3 Reasons We Can’t Exploit Pricing Errors in the Stock Market
• How To Buy Index Funds in 4 Steps For Beginners
• 7 Things To Do When Index Funds Get Boring

What do you think? Is AI predicting active stock trading a good or bad thing? Or does it really mean nothing at all? Do you still trade actively? Let me know in the comments below!