# Bubble Detection: The Challenge

Damned! It seems Flo and I will never truly disagree about anything.

Let me summarize our agreement: A fully rigorous definition of fundamental value is “the expected value, conditional on the information available today, of the discounted sum of future cash flows”. Fundamental value is equal to the equilibrium price we would get in an efficient market. That’s the efficient market hypothesis (which is a misnomer, because it’s not just a hypothesis but a true theorem). Hence what we mean, conceptually, when talking about a bubble is the deviation of the actual price from this hypothetical equilibrium price. And these deviations can come from a) people being stupid in the sense of not taking into account all the available information and/or b) market imperfections like transaction costs or costs of information gathering.

Our disagreement is that, while Flo thinks there could be reliable ways to detect bubbles in the real world ex ante, I don’t. There’s only one way to find out. I can “predict” 50% of all bubbles merely by chance (e.g. by flipping a coin). So this must be the benchmark for any bubble spotting strategy. Let’s see whether credit growth can successfully predict the next 5 bubbles. Under the null hypothesis that credit growth has no better predictive power than a coin, the probability of guessing 5 out of 5 bubbles right is 3,125%. If Flo can do it, we can reject the null at a reasonable level of significance. Or, if you prefer, we can also reject the null if credit growth is able to predict at least 15 out of the next 20 bubbles (p-value 2%).

Fancy a bet, Flo?

## 4 thoughts on “Bubble Detection: The Challenge”

1. Florian Kohlfürst says:

Oh, we’re on! Economists should put their money where their mouth is way more often anyways.

Some thoughts though on the practicability of the exercise – or the precise parameters of the bet. I’ll try to come up with something useful, but mainly I see the problem to be: How do we determine ex-post what a bubble was?

On the “reliability” of indicators: I’d quibble with your wording. I don’t think there are truly reliable indicators, just indicators that are useful enough to base policy decisions on them, even if sometimes they will be wrong.

2. My test of reliability is whether an indicator has a predictive power higher than 50%. So if we agree that there are no bubble spotting techniques out there which outperform a fair coin, why then should monetary policy be based on them? Why don’t we let Draghi and Yellen flip coins? Call that the random walk theory of monetary policy.

• Florian Kohlfürst says:

By “indicators that are useful enough to base policy decisions on them” I meant indicators that significantly outperform a coin. In any case, a central bank would have to compare the certain costs inflicted by a potential policy action (e.g. interest rate rise) with the uncertain benefits of limiting the scope of the bubble (which include the possibility of there being no bubble at all). There is no reason why an indicator has to be perfect to make this a very reasonable way to go about it. Central bankers already operate in an area with considerable uncertainty, there’s no inherent reason why uncertainty with regards to the presence of bubbles should disqualify these from being targeted, while the equally real uncertainty in economic forecasts regarding inflation and unemployment/output gaps should be considered perfectly normal.

3. That reminds me of James Tobin’s only slightly exaggerated characterization of monetarism: we don’t know what money is, but whatever it is, its supply should grow at 4 percent annually. We don’t know how to spot bubbles, but the central bank should fight them. Or perhaps central bankers should identify bubbles like the US Supreme Court identifies porn: they know it when they see it.