Money Sundays: On Making Overconfident Predictions

In some domains, there's a clear and direct relationship between a prediction's accuracy and the evidence supporting it. Example: You want to know if Hillary's going to win the 2016 election. You look at polls, and if they show a high likelihood that she'll win, you can thus have high confidence that she will indeed win.

Or: You want to see if the latest iThing is a good product or not. You read many reviews, and all of them are positive. This leads you to believe--again, with high confidence--that the product is a good one.

In those kinds of domains, a prediction, as long as it's backed by a lot of evidence, can actually be "confident." And it's likely to be right.

However, in other domains, there is much less of a relationship between a prediction, the evidence and your confidence. As an example, take investment markets. You think XYZ stock is a great stock. You look at all the evidence: it has amazing profitability metrics, a growing market share, a juicy dividend, rapid growth, whatever. You therefore predict that this stock is a good stock to own, that it will offer you very good returns. And because of all of this compelling evidence, you are highly confident in your prediction.

Except that here you are likely to be wrong with your prediction, and your wrongness will likely be proportional to your confidence.

Why is this? Because you are not acting alone in this domain. Others are looking at the same information you are and have likely acted accordingly, perhaps by buying the stock and driving it up to a price where all of this compelling evidence is already fully discounted in the stock price. In other words, information about the domain directly impacts the domain itself, because others are taking action alongside you based on that information.

The environment therefore becomes recursive and it leads to, at times, painfully counterintuitive outcomes. Information that makes a stock look attractive actually causes that stock to become unattractive, because too many other people already agree with you.

So, what are other examples of domains where we see this? We see it in bond markets, currency markets, commodity markets and in interest rates, which are essentially a market for the price of money. We also see it in many forms of economic forecasting.

In each of these domains you must be mindful of the second-order (and third-order, and nth-order) effects at play. What do I mean by this? Simply that you have to think about not just the information available, but also what other people think about that information. And what people think about people's thoughts about it. And so on.

This is not to say that evidence doesn't play a role in decision-making and prediction-making in domains like the stock market. It does. But you must also consider to what extent the opinions, information, evidence and predictions you hold are shared by or already acted upon by others. Admittedly, it can be difficult to know this part of the equation, but one general heuristic you can use in these domains is this: the widely-held obvious conclusion usually isn't the right one.

This is an important distinction between two different types of decision-making domains, and it's why it is dangerous to claim rock solid confidence with any prediction about interest rates, stock prices, inflation rates, bond prices/yields, etc.--particularly over the near- to medium-term. The perfect example here would be the widely- and confidently-held view from say, two years ago, that we were in a soon-to-pop stock market bubble and that terrible inflation and significantly increasing interest rates were right around the corner. So far, that "obvious conclusion" has been quite painfully wrong. The irony, of course, is that this conclusion will probably become likely only when people start to seriously doubt its obviousness.

To put it bluntly: No one is ever on rock solid ground making predictions in these domains. And your confidence in your prediction is much more likely to be a reverse indicator!

Read Next: Money Sundays: How To Get Balanced, Consistently Useful Expert Advice

How can I support Casual Kitchen?
Easy. Do all your shopping at via the links on this site! You can also link to me or subscribe to my RSS feed. Finally, consider sharing this article, or any other article you particularly enjoyed here, to Facebook, Twitter (follow me @danielckoontz!) or to bookmarking sites like reddit, digg or stumbleupon. I'm deeply grateful to my readers for their ongoing support.


Lauren said...

Wasn't it Heisenberg who said that looking at a thing moved it, so that the act of observing its position changed its position? Something like that.
There's math, which is tidy, and then there are the people who use it, and people are inherently messy. All this business around trying to predict people is about trying to sanitise the messiness that is humanity. There's a hint of the Obermensch to it. It's futile. But it keeps a lot of people busy, and the wheels humming along.

Lauren said...

I stand corrected: it was Schrödinger's cat. From WikiHow: "the observer's paradox: the observation or measurement itself affects an outcome, so that the outcome as such does not exist unless the measurement is made."
Heisenberg said you could either know the position or velocity of an object in motion, but not both. Or something.
I don't really do straight-line thinking, I just happen to be married to a German engineer.

Daniel said...

Lauren, I think that does apply to an extent to investment domains: if enough people are financially positioned for a specific outcome, the outcome tends not to happen in the way it was expected to.

And, heh, thanks for the physics lesson. :)