Wednesday, July 10, 2013

...or, in case you're me, a "Noahstradumbass".

Here's a quick guide for making easy ("dumbass") predictions that are nearly guaranteed to be borne out by events, thus making you look like a sage. The key is to make non-time-sensitive predictions of things that are very likely to happen at some point in the infinite future.

For example: "The S&P 500 will eventually hit 6,000." OK, sure. Suppose that stock prices are a random walk (or if you prefer, stock returns are a fractional Brownian motion with drift, driven by tempered stable innovations with ARMA-GARCH volatility clustering blah blah blah). Well, the variance of a random walk goes to infinity as you look farther and farther into the future. So if you wait long enough, it's a statistical certainty that someday the S&P will hit 6,000. And your prediction will be "vindicated". But at (almost) any point in time, 6000 is NOT the optimal forecast for the future price of the S&P.

Lots of things in finance are pretty similar to (lognormal) random walks. Exchange rates, stock prices, etc. So you can be a Nostradumbass by making predictions about these things but conveniently leaving out the time frame (e.g. "all fiat currencies go to zero eventually"). Make sure to make up a neat story to "explain" your prediction. If the dollar is at 0.8 euros and you predict that someday it'll go to 0.4 euros, make sure to call a "bubble". Actually, the "bubble" story is an especially handy one, since you can say "No one can predict the timing of a bubble burst, but it has to burst eventually, and the fundamentals point to a true value of 0.4."

(If anyone asks you where you got your prediction, tell them you got it from many years of carefully studying the Human Action Axiom.)

OK, here's an even easier example. Some things probably can't go on forever, so according to Stein's Law, they will someday end. So just find something that probably can't go on forever, and predict that it will someday stop. For example, China probably can't keep growing faster than rich economies forever (no country can). So just predict that someday, China's growth will slow. Bingo! As soon as the inevitable happens, you will be "vindicated" and appear "prescient". It might take 10 years, but the resultant glory will be worth the wait.

Now, maybe people are ribbing you about the fact that your predictions don't come with a time frame. Never fear. A slight upgrade to your Nostradumbass strategy will silence the doubters. First, find something that is probably stationary, like real interest rates. Then predict that they'll return to their long-run level. Make sure that the time frame of your prediction is longer than the typical time frame of mean reversion for that variable - so if interest rates usually revert to the mean in 5 years, predict that interest rates will normalize within 10 years.

(Note: Be careful with this last strategy! This strategy does have a nonzero chance of failing. If the thing you thought was stationary is not in fact stationary - for example, if you think economies always bounce back within 10 years of a recession, but actually GDP is unit-root, then you're in trouble! But on average, things that look stationary will turn out to be stationary, and so on average your predictions will be "vindicated" using this strategy.)

But always, always remember the cardinal rule of the Nostradumbass: Make up a good story. Your story should have two key characteristics: 1. Originality, and 2. Truthiness.

1. Originality. If you're predicting that China's growth will someday slow, your story should not be something like: "Eventually, technological catch-up will be exhausted and the marginal product of capital will fall." These are things that everyone knows! No one will think you're a sage for citing these things. Instead, say something like "China's type of growth model is unsustainable." Originality means that instead of applying a well-known formula or just looking at the historical pattern, people will think that they need your unique insight into the quirks and complexities of the present situation.

2. Truthiness. Remember, people suffer from confirmation bias, meaning they like to hear stuff that dovetails with their existing beliefs. Tell people a story that jives with their political prejudices. If you're predicting a 50% drop in stocks (someday!), tell people that it's because America's economy used to be built on sound economic fundamentals and sound money, but that government meddling and the liberal contempt for private business have eroded our foundations. Eventually - who can say when! - the markets must realize this, and the bubble will pop. Truthiness allows you to increase the sense of pleasure your believers get when your prediction is finally "vindicated".

To sum up: The world is full of both statistical randomness and simple well-known inevitabilities. But humans are full of confirmation bias, the need for the illusion of understanding, and the need to believe in sages. You can take advantage of these two facts and become a respected sage without ever having to do any really hard thinking. The Nostradumbass life is a good one.

1. Anonymous9:24 PM

Ah, the business strategy of Fox News. 'Tis a good one.

2. Oh no, you have NO idea the potential of The Nostradumbass™. And I've already used it no less than 6 times so far:

"Going to pick up take-out, but I need to fill up the tank first before I run out of gas." Good thinking, Nostradumbass.
"Hurry up, the food's going to get cold!" Thanks, Nostradumbass.
"We should...before the sun goes down." Amazing, Nostradumbass.

I look at the world differently now, and I've only JUST begun to really live. ;)

3. But there is a way to fail with no qualifications: announce, like Krugman in the 2000s that the US will have a "funding crunch". "Some day" of course. But this day will never happen, technically impossible for a currency issuer to have a funding problem.

1. Anonymous11:06 AM

Restating your priors does not amount to a refutation of views that do not rely on your priors. Please, this was a nice, funny poke at prognosticators-for-hire until you messed it up.

2. Nathanael1:06 AM

It is, of course, impossible for a currency issuer to have a "funding problem". This is really part of the definition of "currency issuer".

A currency issuer can have *other* problems -- for instance, a total collapse in the legitimacy of the government -- but never a funding problem.

4. Noah,

" it's a statistical certainty that someday the S&P will hit 6,000"

Well, no. An upward drifting brownian brownian motion crosses a lower boundary with less than probability one. The point that variance goes to infinity is irrelevant. If variance grows linearly, expected deviation goes as root time, so it's not hard to see that that a linearly vanishing boundary might quickly become unattainable. In reality, I don't think it's a safe bet a all that the s&p will hit 8000.

1. Yeah that's true, the downward prediction isn't as certain, just the upward one. If the drift were zero then any prediction would be a sure thing.

2. At very short time scales (i.e. intraday) drift is negligible. The asset prices move by Brownian motion + jumps. What type of jump process, its intensity, proportion, etc. depend on the asset class. Bollerslev, Andersen, Ait-Sahalia are those who work on this stuff yet they do not study prediction side of the story (to my knowledge).

5. Oops. I meant 800. If its drifting upward it will hit 8000 almost surely. It was your the "nostradumbass" prediction about stocks going from 1500 to 800 that I was objecting to.

1. You're right, to make the example better I changed it to exchange rates, which are closer to a non-drifty random walk. The case of the falling stocks prediction getting too much respect is still OK, but a little too subtle for a humor post.

6. Did you get too much blow back from your last Keen post, Noah?

1. No, I didn't get any blowback actually (though I'm guessing Keen wasn't too happy with it). If you think I was mean to Keen, wait til you see me take on Schiff!!

7. Anonymous7:03 AM

^_^

Nothing else need be said.

9. Perhaps you can follow up with a post on how to become a respected business journalist. Step 1, state which direction the stock market moved in the past 24 hours. Step 2, find something else in the world that moved the same direction. Step 3, write an article claiming that the latter caused the former. WOO! You're a journalist!

1. This post is coming at some future date...

2. My joke was that when something happens, just take your new feed (formerly the stock ticker tape) and scroll back to the last good/bad thing that happened in the world. Use that.

3. 'news feed'

10. Anonymous8:21 AM

There's such a thing as fractional Brownian motion and also geometric Brownian motion. But there's no such thing as "fractional geometric Brownian motion". :P

You also didn't mention to not make too many predictions. Otherwise, you'd end up looking like those cranks who predicted 10 of the last 2 recessions over the past few decades.

1. There's such a thing as fractional Brownian motion and also geometric Brownian motion. But there's no such thing as "fractional geometric Brownian motion". :P

?? What do you call the price process when the return process is a fractional Brownian motion???

2. Fractional Brownian motion is not a semimartingale. Thus, such a process does not avoid arbitrage opportunities. Brownian motion (i.e. standard Weiner process) is a semimartingale.

3. That's right, fractional brownian motion is what they now are using to try to describe "long-range dependence"...

4. Anonymous4:06 AM

Noah, you make a good point about what we should call the price process when the returns are fBm. I haven't thought about this and haven't seen a standard term for such a thing. In Biagini, Hu, et al., a book about fBm, they call it a "generalized stock price". That's a pretty bad name.

So "geometric fBm" is arguably the best name for it, although I don't think it's standard.

5. Technically it's not even Brownian motion if the driver isn't normal, right?

I am not an expert with these terms.

For the "fractional geometric Brownian motion" return process with normal innovations, i.e. real fBm, it shouldn't be that hard to sit down and work out explicitly what the price process is...we could just call that "fractional geometric Brownian motion"! ;-)

11. Anonymous10:00 AM

Can't help to fell the target here is Mr Pettis. However I felt it is a little bit unfair for him. Say you want to figure out when the China grotwth will slow. Based on "Eventually, technological catch-up will be exhausted and the marginal product of capital will fall.", you will probably guess the upper constrain is real GDP per capital at US or even Taiwan level. But based on Mr Pettis’ analysis on trading balance, you can probably argue the total nominal GDP at US level is a constrain. So there is some new information provided.

1. Actually I respect Pettis' predictions a lot, because he predicted not just a China slowdown, but a shift toward services and consumption that would accompany that slowdown. Furthermore he predicted that consumption growth would accelerate as growth fell, which is counterintuitive but turned out to be true. And finally, he gave a time frame for his predictions, which turned out to be pretty accurate. So I'd say his predictions have been pretty non-dumbass.

"This irrationality will be corrected one day, by market forces, and the consequences will be dire."

http://inteligenciafinancieraglobal.blogspot.mx/2013/06/the-wars-against-gold-roubini-wrong.html

1. buahahahaha

13. The fastest way to being a Nostradumbass is to predict things that are already happening. Seriously. It's a favourite ploy of investment advice rags.

14. Nathanael1:03 AM

I have to say that my prediction of the collapse of the US federal government falls into this "Nostradumbass" category.

* Very vague timeframe.
* No government has ever lasted forever, and the US government has already had an exceedingly long run compared to most. So I'm just predicting that something which can't go on forever will stop.
* There are already major signs of dysfunctionality and non-functionality, and people are already "routing around" the government, meaning that I'm doing a "predicting the present" to a great extent. (Making the prediction before the 2000 coup would have been more of a prediction.)

Now, here's the thing: I think my prediction is still useful. Why? Well, although it's obvious to any *educated*, *informed* person that the US has a spectacularly broken federal government (exhibit A: the US Senate) and a defective democracy... and while it's also obvious that a government which simply doesn't bother to keep people fed (exhibit A: cuts to SNAP) is going to be overthrown sooner or later.... that's only obvious to *thoughtful* people.

The "America Fuck Yeah" mentality means that lots of *uneducated*, *ill-informed* people take this prediction as if it's a surprise. " 'Murka will last for 20,000 years!" is their mentality, so to them, this is news.

15. The Nostradumbass strategy has worked quite well so far, but many people will wise up to it. The Nostradumbass bubble will burst.

16. You mean "jibes," not "jives." A common misconception. Nobody's really jived since the '70s.

17. If you want to preserve your legacy, you have to move on to the pure "Nostradamus" strategy. In addition to loose time frames, you have to add obscurity:

The river that tries the new Celtic heir
Will be in great discord with the Empire:
The young Prince through the ecclesiastical people
Will remove the scepter of the crown of concord.

In a short time the temples with colors
Of white and black of the two intermixed:
Red and yellow ones will carry off theirs from them,
Blood, land, plague, famine, fire extinguished by water.

The enemies of peace to the profligates,
After having conquered Italy:
The bloodthirsty black one, red, will be exposed,
Fire, blood shed, water colored by blood.

Once people start trying to figure this stuff out, you're bulletproof.

18. I once argued with a theologian who defended prophecy on the grounds that it had to be judged on a case by case basis-- apparently on his say-so since he offered no reason to believe so. I think dubious economic forecasting is open to all the same criticisms as divine prophecy. I can go on uttering near tautologies all day concerning future events, and I would have a 100% succesd rate, but this does prove that I posess any special insight.