Defense Against the Dark Arts in Finance: Using AI to Restore Fundamental Expectations

Defense Against the Dark Arts in Finance: Using AI to Restore Fundamental Expectations

Promotional emails are not our favorite genre. They rarely convey deep insights. But a few weeks ago, as the Q2 earnings season was nearing its peak, we sent out one that is worth revisiting today. It read, in part:

Yesterday I read in a reputable outlet that “Wall Street expects Amazon to report the EPS of $12.20 on Thursday”. That is of course nonsense, since nobody on Wall St expects anything of the sort. Anyone who’s in the know expects a number well above $14 per share, and the whispered debate is about whether Amazon’s EPS will hit $15.

And yet sell-side analysts keep churning out their biased “estimates”, news reports cite those absurd figures, and Main Street investors get led astray. But it does not have to be this way.

We at Proximilar use the tools of science to come up with the highest quality earnings & revenue forecasts in the world. We don’t put our faith in rumors or whispers. Instead, we rely on ultra-fast computers and on artificial intelligence that outperforms human analysts 70% of the time. We invite you to use our AI’s superior predictions in your work and analysis – absolutely free.

P.S. For Amazon our AI is projecting $15.01 EPS.

Now, as the new earnings season is about to begin, it’s a good time to run a postmortem.

Amazon’s surprise

Amazon.com announced its second quarter results a few days after we sent out our email, on Thursday, July 29. The reported EPS was $15.12 per share, very close to Proximilar’s forecast. Wall Street analysts were not just wrong, they were predictably wrong, and in a spectacular fashion. The standard deviation of analysts’ forecasts was about $1, which means professional stock researchers were off by nearly 3 sigmas.

But what happened to the stock after Amazon beat the EPS consensus by such a wide margin? AMZN fell about 8% the next day. It recovered slightly in the subsequent week, but as of this writing it is yet to regain its pre-announcement price of $3,600. To put this 8% drop in perspective, the options market was pricing in a ±2.8% post-earnings move. So an “outperformance” by roughly 3 standard deviations led AMZN price to drop by 3 standard deviations. Crazy, huh?

Why the tumble?

There were several reasons for the drop in Amazon’s price, including a Q2 revenue miss. But we believe the key reason is that there was no one to articulate the true market expectations. It is strange to watch this happen to a $1.6 trillion company in real time. Investors knew not to take consensus estimates seriously, but they did not have a good alternative. The crisis of authority led to an information vacuum which begot chaos. AMZN fell mainly because investors panicked, and they panicked because they had no shared narrative of what the giant company’s results were supposed to look like.

Let’s take a look at another recent example.

Apple’s success

On July 27 Apple, the most heavily analyzed company in the universe, released its earnings. The company knocked the cover off the ball in every sense: $1.30 EPS (+$0.29 surprise vis-à-vis analysts’ “consensus” estimate) and $81.4 billion in revenues (+$8.1 billion surprise). Apple beat its whisper number as well as our AI’s forecasts (Proximilar estimated $1.14 EPS and $79.0 billion revs, which translated into +$0.13 and +$5.8 billion over consensus, respectively). The announcement included new revenue records in all geographic segments and double-digit growth in all product categories. So what do you think AAPL stock did the next day to celebrate this achievement?

Reader, it fell by 1.2% while NASDAQ rose slightly.

How to beat the expectations that don’t exist?

Apple’s market cap was over $2.3 trillion, which means roughly $30 billion of shareholder value disappeared overnight. How can this happen to the biggest and the most followed company in America? This is not random noise: the decline was bigger than Apple typical daily move. So how can $30 billion of shareholder value vanish on the news that in a single quarter the company created $5 billion more value (earnings) than the market ostensibly expected?

News headlines might rationalize the AAPL stock decline by saying that investors actually expected even better results. But that explanation does not make sense. Not only did Apple’s result leave the earnings whisper in the dust, but fewer than 3% of forecasters on Estimize, a leading crowdsourcing platform, thought the result would be better than it was.

The only way this drop is possible is if the market noise overwhelmed rational expectations. To say it differently: Apple’s investors lacked conviction. They joined random selling because any expectations they had were weakly held. Again, as with Amazon’s announcement, we see that a shared narrative is missing. With no common understanding of what constitutes good or bad performance, short-term noise comes in to fill in the void.

“Contrarian reactions”

Amazon and Apple examples above are far from the most egregious cases of what academics call “contrarian share price reactions.” This happens hundreds of times every quarter: in the last year over 40% of companies who beat both EPS and sales estimates have seen their stock drop. Conversely, over 30% of those who missed both numbers saw their stocks rise. There is strong evidence that the amount of noise trading has increased in recent years as the proportion of contrarian reactions has gone up.

Unreliable earnings estimates may not be the sole cause of the problem, but they have contributed to the investors’ nihilism. Today fewer investors think, “if AAPL reports $1.20 or higher, the stock will go up.” Instead, they think, “Can’t trust anyone’s judgment on fundamentals. If Apple starts falling, just ignore the damn EPS number and sell the stock.”

This is a rotten dynamic, but a recognizable one. The investors’ understandable loss of faith in finance experts leads to the same result we have already seen in US political life: the disappearance of a shared reality.

Greed has no time for fundamentals

There have always been Ponzi schemes in America. Fraudsters and loud charlatans always preyed on the people who wanted to get rich quick. But market experts and trusted intermediaries used to lend an aura of legitimacy to some of the sounder investments while making others more dubious in the public’s eyes. Now, whether through greed or ineptitude, many of these experts have lost their credibility.

Much as J.P. Morgan or Goldman Sachs might like to think that they are steering U.S.S. Finance, the proliferation of crypto assets, speculative SPACs, Robinhood options trading, and meme stocks shows otherwise. In most cases respectable financial institutions are not leading the way – they are just jumping on the bandwagon of a snake oil boom they don’t control.

The US markets are in the midst of multiple speculative manias driven by short-term greed. This greed is rooted in the fear of missing out and its primary tools are sentiment and momentum trading. One thing greed has no time for is fundamentals. If history is any guide, these speculative manias are unlikely to end well. But whether or not they lead to a major market downturn, the only way forward – and the only antidote against snake oil – is to bring long-term thinking and responsible decision making back to finance.

Rational projections

Charlie Munger once said,

It is an unfortunate fact that great and foolish excess can come into prices of common stocks in the aggregate. They are valued partly like bonds, based on roughly rational projections of use value in producing future cash.

But they are also valued partly like Rembrandt paintings, purchased mostly because their prices have gone up, so far.

Is there a way to restore a sense of responsibility to the frenzied market? To de-emphasize the appeal of assets whose prices have gone up? Can we help investors refocus on “rational projections of use value in producing future cash”? Proximilar believes that we can and we must.

Things that we don’t measure

There is a well-established idea in business and in social sciences: if you can’t measure something, it does not exist. This is true for high-school students’ knowledge of math and for Olympic athletes’ performances. And it is also true for investors’ decision making. If the quality of analyst forecasts for the next quarter’s results is dismal, then their predictions on a multi-year horizon are even worse. So any lay person who tries to base their investment decisions on “rational projections” is in for an unpleasant surprise. They will find that the ingredients and tools available are not suited to the task.

Suppose ten finance professionals are asked to build ten fundamental models of a company. These models may all make different assumptions and arrive at different conclusions. But they will all rely upon similarly flawed earnings estimates we talked about above. What’s more, they will all use the same antiquated analytical tools. Instead of using the most advanced techniques that the 21st century has to offer, like AI and simulations, they will fall back on valuation models that have not changed in decades.

The ten resulting models may yield different answers, but they are all the fruit of the same sickly, aging tree. Instead of true diversity of opinion it yields groupthink.

AI can do what experts never could

There is a way out of this predicament. To help investors focus on longer investment horizons and on financial fundamentals again, we need to leverage new scientific tools. We are talking about the tools of AI that mathematicians and data scientists developed in the last two decades.

We need to build operating models of companies that realistically represent a broad range of possible ups and downs over the long term. This is not easy, but it is not impossible. Manufacturers and engineers do this for each new aircraft and for every nuclear power plant they build.

This road may be long. New ways of doing things do not catch on overnight, especially in an industry as conservative as finance. But we have already taken the first several steps – big steps – and we are not stopping. Our EPS and revenue prediction engine is state-of-the-art and one of a kind. It is the foundation on which we are building a deep forecasting toolkit. It will make sound decisions easier for investors everywhere.

Our vision: evidence-based finance

What is our goal? It’s to empower investors with science and objectivity. A couple of years from now, an investor will type a stock ticker into a browser window, and in addition to the data they get today, they will see an objective estimate of what the stock is worth. They will get the probability distribution of the future value the company will generate for shareholders. The discounted sum of future free cash flows, the liquidation value, and the acquisition value will all be combined in a probabilistic fashion. The way we analyze companies now will seem quaint, like a guy reading a stock table in a newspaper in an ’80s movie.

Before this future inevitably arrives, a lot of work will need to be done – both in the technical sense, and in terms of outreach to the investor community. Still, this is the future. We’d love for you to join us in it.

SEE ALL FORECASTS

Leave a Reply