Monday, December 23, 2013

Noah smackdown watch: Bayesian updating edition

Either I manage to stay on most people's good sides, or most people just don't care about me that much, but for whatever reason, Noah smackdown attempts are rare. But today, J. Bradford DeLong has delivered a smackdown on Yours Truly. Last year I wrote a qualified defense regarding John Cochrane's warning of inflation, saying that basically, well, these things are hard to call. If there's a 1% chance of a catastrophic hyperinflation and a 99% chance of a continued economic stagnation, is it right to say that inflation is "the danger"? Brad says no:
Naughty, naughty Noah!... 
When you say something is "the danger", you are saying that other things are not dangers. "The". Definite article. You are saying that there is one and only one danger... 
And if other things are in fact dangers? Then the fact that you did not say that "the danger" would arrive by a date certain does not rescue you. You are wrong in your dismissal of those other things as not dangers. 
You can plead hyperbole--that you did not mean that this was really the only danger but that it was the only significant danger, or the only truly worrisome danger, and that it was the only worrisome or significant danger. 
But then, when asked about what actually happened, and whether your dangers was in fact the only worrisome or significant behavior, all you can then do is say something like: "LOOK! HALLEY'S COMET!!" 
I still say that while it is true that predictions and scenarios are not necessarily forecasts, a willingness to mark your beliefs to market is the first virtue of a wannabe academic. 
I say a failure to mark your beliefs to market is a fatal vice for anybody who wants to be taken as an intellectual.
As the wannabe academic in question, I think I'm fairly willing to mark my beliefs to market. But my own beliefs didn't really enter that much into my earlier post.

But how about Cochrane's beliefs, and terminology?

Suppose that you decide that A) a catastrophic hyperinflation is the worst feasible scenario (which seems reasonable, given that the U.S. borrows in our own currency), and that B) there is a 1% chance of that happening, and that C) 1% multiplied by the cost of (A) is considerably greater than 99% multiplied be the cost of a continued grinding stagnation. In that case, I think you're justified in saying that inflation is "the" danger. Yes, as Brad says, you're being hyperbolic, but the phrase "the danger right now" is common enough that I'm not going to lose any sleep over it.

So how about marking beliefs to market? With very unlikely events, this is hard to do. Suppose you believe that you have a 1% chance every year of getting shot. Suppose three years go by and you don't get shot. Well, even if your prior belief was correct, there was a 97% of that outcome happening. So you're not going to update your belief very much. 

So that, I think, is why I didn't get on Cochrane's case about the inflation warning, the way I got on, say, Steve Williamson's case for his 2012 prediction of imminent inflation.

If all this sounds like excuse-making, well, it is. It's difficult to go back and read something you wrote in the past and try to figure out if it was dumb or not, because there is always the nagging feeling that you must have had some good reason to write what you wrote. So here I am reading my earlier post and wondering what I was thinking. 

I think I was thinking something like this:

Before the 2008 crisis, many people were worried about a dollar crisis, in which China would decide to sell off dollar assets, triggering a flood of selling that raised interest rates. Those who worried about that danger included Brad DeLong (see also here). A spike in interest rates would tempt the Fed to engage in extreme monetary easing or even finance the deficit directly, which many theories predict would lead to inflation.

I think that this is probably what I thought Cochrane had in mind when I read this:
Our danger now is a run on Treasury debt.  It's not just can the Fed soak this stuff back up again, but can it soak this enormous amount of debt back up again when people don't want either money or Treasury bills or anything labeled "U.S. Government."  The danger is not 1932; the danger is Argentina, a massive run from Treasury debt.
Now, that scenario is not at all what happened in 2008, causing many of the people who had worried about it - including Brad DeLong - to issue mea culpas. But just because a dollar crisis didn't happen in 2008 doesn't mean that it wasn't possible in 2012. It is not even necessarily true that what happned in 2012 proved that a dollar crisis was significantly less likely in 2012 - remember the example of the 1% chance of getting shot. 

So I was probably thinking in those terms. I was probably thinking "Cochrane thinks that the dollar crisis scenario is still a danger. That seems unlikely to me given that it hasn't happened already, but maybe Cochrane sees things differently for reasons I don't understand. And maybe he's talking about small-probability events with catastrophic consequences, and that's why he calls it 'the danger' while not noticeably changing his belief in this regard."

At least, that's what I think I was thinking...

Update: On Twitter, Brad pointed out that an "Argentina" outcome probably isn't so bad that we should consider a 1% chance of it to be "the big danger". Actually, my instincts agree with that. But since the dollar is the world's reserve currency, I think it quite likely that a dollar crisis would be a much worse calamity than a run on the Argentine peso.


  1. There have to be worse sins than the sin of being too nice . . .

  2. While I like to disagree with Noah about all things macro, I think DeLong's smackdown is pretty weak sauce.

    It is true that high inflation (or even target inflation or even a pick up in inflation) has not come to pass but that does not mean it was a risk that we should now slam someone for having considered.

    This whole back and forth makes me appreciate the forecasting procedure at the Board. We have a staff forecast (which is published in the Tealbook/Greenbook) and it is the "modal" forecast or the one we view as most likely (be it inflation, GDP, employment, etc.). Economists usually think in averages as a central tendency and most of our models spit out averages but we specifically try to judgmentally (with lots of model help) to write down the most likely path for these variables, conditionally on policy. Now of course, that is not enough information. We also think about the risks, to the upside and to the downside and whether those risks are well balanced.

    I believe DeLong was splitting hairs on the article Cochrane used for "the" danger. Maybe to Cochrane that was the relevant one to HIM (and he had a whole story to back it up related to unconventional monetary policy and economic relationships I suspect). But that does not mean it is the only risk. Not sure about the timing but there have been a few deflation concerns in recent years too so I am sure Noah could have (and should have) pointed to an opposite danger too. Ok so why is it important to think about these risks (even if they later don't come to pass?), well over time data comes in and other shocks hit the economy and that modal forecast may not look so smart and so it starts to morph toward one of the risk scenarios. Or maybe not, but it is good to have alternate scenarios in place ahead of time. (And valuable alternate scenarios often come from rather different approaches than the baseline.)

    In the end, I would say the smackdown is no good if it discourages others from thinking through risks, even ones that in this run of history don't come to pass. (Our models are too simple for there to be one that always works best.) And for those who want to get into forecasting (Noah?) you should be very clear about what you think (not so important what you think what some other economist thinks) and why. It makes the ex post evaluation easier. It is fine to pick Cochrane's forecast apart ex post but there is no shame (and some virtue) in having thought about its plausibility ex ante.

  3. 1% multiplied by the cost of (A) is considerably greater than 99% multiplied be the cost of a continued grinding stagnation.

    You can't just throw off a statement like this. To Brad's point, you have to have some way of quantizing those costs, then multiplying by the probabilities.

    And where do 1% and 99% come from? Are they picked from air? [I'm being kind.]

    What if it's 0.1% or 0.001%. That certainly changes the picture.

    And I thought both 1) having our debt in our own currency and 2) the dollar being the world's reserve currency were huge deterrents to hyperinflation.


    1. What if it's 0.1% or 0.001%. That certainly changes the picture.

      What if it's 5%? Or 10%? Still a small chance, right? ;-)

      And I thought both 1) having our debt in our own currency and 2) the dollar being the world's reserve currency were huge deterrents to hyperinflation.

      (1) no. (2) yes, but that is exactly what would suddenly stop being the case in a "dollar crisis".

    2. It is excuse making. Inflate the cost of any possibility enough then any triviality becomes significant. Do people under or over estimate remote chances? Do they under or over estimate the cost? They do both depending on the situation and framing. And the cost must be compared to the costs of it not happening, which in this case are likely even greater.

  4. I'm totally going to call you "Naughty Noah" from now on. That is an awesome nickname/Twitter Persona.

  5. Your justification that a 1 % probability of an event multiplied by the estimated costs of such an event does not acknowledge the very high systemically inherent uncertainty of each estimate. How can a confidence level of 1 sigma, 2 sigma, etc be assigned to an extremely unlikely event be rigorously supported. Is it 1 % chance of occurrence or .05 $ or .01 % ? As for the costs of catastrophic scenarios they are also, by their great distance from both the current situation and largge difference from prior catastrophes also difficult to estimate. I think there is a .001% chance a meteor will destroy Fort Knox. Prove me wrong. Prove the costs to the country and world of that event?

    1. Well, overlooking the fact that Fort Knox doesn't have any gold in it anymore, you're right, these rare events are inherently hard to predict from past data, you have to go by theory. And we have precious little theory about the motives of the Chinese government.

    2. The motives of the Chinese leaders are pretty clear: (1) they don`t want to be murdered by their own people; (2) enhancing China`s economy and prestige reduces the chances that they will suffer a Ceaușescu; (3) one of the risks for China is rising to an unsustainable standard of living; (4) accumulating foreign reserves provides employment and enhances China`s access to resources, prestige and future influence while suppressing the current standard of living.

      If China wants to challenge the United States it is much better off investing heavily now and not building carrier strike groups until it is much richer than it is now. China could not match the US militarily now but if it keeps military spending down and invests heavily now it can expect to be able to challenge the US militarily in the future.

  6. When I read Brad's latest smackdown,it was obvious to me that he was smacking Cochrane rather than Noah. Then why did he put Noah's name on it? Beware, Noah. Be very much aware that if you defect from the Delong, Krugman, Baker line, even for one small paragraph, you will quickly cease being their junior darling of the economic blogosphere. Your blogs will then cease getting their frequent links and reblogs which are driving your public as well as salt water academic notoriety. Maybe they will even take Kimball down with you.

    Whatever happens , I will keep keep on reading all of you because you are among the few making sense, and it's rather amusing how Brad tries to throw his inconsiderable weight around.

    1. Be very much aware that if you defect from the Delong, Krugman, Baker line, even for one small paragraph, you will quickly cease being their junior darling of the economic blogosphere. Your blogs will then cease getting their frequent links and reblogs which are driving your public as well as salt water academic notoriety. Maybe they will even take Kimball down with you.

      I think if Brad, Paul, etc. merely wanted to reward yes-men, they could find much more willing yes-men than Yours Truly.

  7. a dollar crisis would be a much worse calamity than a run on the Argentine peso.

    A sudden devaluation of the dollar might take out the world financial system through the derivatives markets but apart from that it is hard do see how you could even have much a dollar crisis.

    What would a dollar crisis look like? Where would the money that is flowing into US treasuries go? Would it go to the Detroit or New York real estate markets - that would be no big deal. It might go into physical inventories of gold and copper - but that is already happening. If the US dollar drops, the price of gasoline in the US goes up and German and Chinese manufacturing do face plants - not a huge crisis for the US.

  8. I call this for Brad. "the danger" means all other dangers are insignificant in comparison. The dollar being the world's reserve currency affects not only the costs of the US Argentining but also the probability. The US can't have an experience like Argentinas as US contracts are in dollars and Argentine contracts were in dollars. Cochrane did not come up with an absurdly bad analogy due to mere carelessness. Nothing like that which he considered "the danger" has occured in world history (as far as I know).

  9. The trouble I have with your defense of Cochrane is that you've more or less given him a blank check. After 4 years of being dead wrong he's lost no credibility-and as it's still only a small probability event the same rule applies. So in 10 years if empirically he's still dead wrong your defense still applies.

    Essentially he's made a totally unfalsifiable claim by your argument. It will take about 100 years before we can say 'Wow, Cochrane sure made a bonehead claim there.'

    Also kind of along the line of what Delong said, I'm not so convinced that hyperinflation is a worse fate than permanent recession. More or less a choose your poison choice.

    Although my definition of 'hyperinlfaiton' is a little different form the mainstream I think that in time the inevitable result of permanent recession would in fact be hyperinflation.

    Again, this is because I interpret hyperinflation as not the result of wild-eyed money printing but political instability.

  10. You never seem to answer my comments-is it personal? Or is it that they are never worth answering? In that sense I think I like Williamson better-he often answers my comments. One the other hand Sumner always answers my comments-he can't bring himself not to-and we clearly aren't friends either.

    My guess is you don't like polarization and sense I'm a polarizing guy.

    Meanwhile I think that Krugman pegs you here. You clearly are enabling Cochrane:

    What did happen was a significant change in what the usual suspects — the people who have been predicting soaring inflation and interest rates, year after year — were saying.

    Did they admit having been wrong? No, of course not. But their excuses shifted. Through 2011 and even through 2012, it was still mainly “just you wait!” — inflation was coming any day now, or maybe it was already here but sinister statisticians were faking the numbers. In 2013, however, it became “I never said that!” — declarations that they only said that inflation was a risk, not that it would necessarily happen, so the failure of inflation to materialize was no big deal."

    That's basically what you're saying-'it's no big deal.' Again, if Cochrane said something unfaslfiable in 2010 then there's no point in even listening to him.

  11. Anonymous12:16 AM

    I thought the first virtue of a wannabe academic is to mark your beliefs to a self-serving revisionist history of what happened.