Friday, December 18, 2009

Here I go Again Criticizing Paul Krugman

Again I agree with Krugman and am quibbling. In this case I quibbling about theory due to Krugman not terminology as in the post below. Krugman argues that Bernanke's answer to Brad DeLong's question is insane. I agree with Krugman, but he makes appealing but invalid arguments.

I can't summarize Krugman's argument as well as he does but the part to which I object is

Future economic historians will, I believe, see this as fundamentally absurd — as absurd as the inflation fears that paralyzed the Bank of England in the early 1930s even as the world went into a deflationary spiral. Yes, there may someday be a 1970s-type episode in which the Fed needs to fight inflation, not encourage it — but it’s a long way off. Furthermore, why on earth would we imagine that the Bernanke Fed, by showing itself willing to inflict gratuitous pain in 2010, would make it easier for whoever is running the Fed in, say, 2020 to control inflation then, let alone that the tradeoff of real pain now versus hypothetical pain much later, if it even exists, is worth making?

I agree with the policy proposal, but I note a logical inconsistency in you (Krugman's) argument. A natural reaction to Brad's question is to ask : how can the Fed cause higher inflation right now when we are in a liquidity trap? The answer, due to uhm Krugman, is that the Fed can't cause higher inflation now, but will be able to cause higher inflation in the future when the economy is out of the liquidity trap.

Krugman's proposal was for the Bank of Japan to commit to a higher inflation target for the fairly distant future when Japan was out of the liquidity trap -- that would be imply, as noted by Krugman, that the unemployment rate is what the monetary authority wants it to be plus or minus epsilon.

So to get a lower long term real interest rate now, the Fed would have to commit to higher inflation at some time when it can target inflation which means at some time when the normal rules hold and long expected inflation does not affect real variables.

The logic of the argument requires that the high inflation target be costly. More to the point, it requires that the Fed can now commit to a future policy different from that which they would choose in the future if they weren't precommitted.

If the Bernanke Fed can't influence beliefs about the someone else Fed, and those are the beliefs that will matter at all times when we would like lower inflation, then the Krugman/DeLong proposal won't work. All the arguments in this post about now vs in the future don't work, because the proposal is to do something different in the future, since the Fed can't cause inflation now except via expectations about monetary policy when the US is no longer in a liquidity trap.

I'd say the valid argument is simpler. The costs of moderate inflation (say up to 10%) are miniscule compared to the costs of 1% more unemployment. Worrying about whether actual inflation will by higher than 3% if target inflation is 3% is like shouting fire fire when you see a gas stove.

One final comment. I think I understand what Bernanke is doing. My hypothsis follows. His aim is to be reconfirmed. He knows reconfirmation won't be blocked by liberal Democrats, but might be delayed by Republicans who are blocking everything. So he wants to convince Republicans that they want to reconfirm them. Therefore he is saying crazy things so that he sounds like crazy Republicans. Also the sane Republicans other than Snowe and maybe collins in the Senate (if any) are evil. They have decided that the worst things are the better things are. Any Republican Senators who understand economics have decided to pretend that they don't aimiing for high unemployment in November 2010.

The terrible thing is that I think Bernanke's calculations are correct. His testimoney is insane nonsense and it is the testimony best designed to get him reconfirmed quickly.


Andy Harless said...

I'm not sure that either you or Prof. Krugman have successfully countered Chairman Bernanke's argument. You're merely constructing different straw men. I disagree rather strongly with Chairman Bernanke too, but it is really an argument about parameter values, about which reasonable people can disagree.

Here's how I would formalize Chairman Bernanke's argument. For any given policy regime, there is a subjective distribution of possible inflation outcomes. That distribution can be characterized by its mean and its variance (and perhaps higher moments too, but they aren't part of this argument). Outside of a liquidity trap, the mean of the expected inflation rate makes no difference in terms of welfare. However, the variance does make a difference: higher variances reduce welfare. Over a cross-section of possible policy regimes, the variance is positively correlated with the mean. Absent the possibility or actuality of a liquidity trap, therefore, we would choose the lowest feasible mean, so as to get the lowest feasible variance. Given that a liquidity trap is possible (and may happen to actually exist right now) we make a tradeoff between the advantage of a high mean (in reducing the risk and duration of a liquidity trap) and the advantage of a low variance. The Fed believes that the mean of the optimal choice is 2%.

(Actually, I'm oversimplifying the argument by omitting the dynamic component. But I think my straw man is a useful one.)

If you accept my interpretation, the decision turns on two questions. First, how high is the correlation between the mean and variance of inflation across policy regimes? Second how bad are the (probability-weighted) welfare consequences of a high variance as compared to those of a low mean?

With respect to the first question, I would argue that, under today's circumstances, the correlation is very low and, indeed, not even positive. The relative risk of seriously overshooting a 3% inflation target is less than the relative risk of seriously undershooting a 2% target. The former depends only the Fed's commitment, whereas the latter may not even be within the Fed's power to avoid. But if you follow the empirical evidence (much like the empirical evidence that buttressed the AAA ratings of various derivative securities now in default), it does not support my view.

I also believe that the bad welfare consequences of a high variance have been exaggerated. There is no question that a low variance is to be preferred ceteris paribus. But, with so many sources of risk, I don't think uncertainty about the inflation rate is the main thing to worry about. Except to the extent that the uncertainty includes extreme possibilities -- hyperinflation or serious deflation. I would argue that the risk of the former is minimal under any vaguely reasonable policy regime, whereas the risk of the latter is non-negligible under today's regime. (I will concede that, since we are more concerned with ex ante expectations rather than outcomes, it is relevant -- even if my view is correct -- that not everyone shares it.)

So it is a judgment call. I don't claim to have better judgment, in general, than Chairman Bernanke, but in this case I think he is seriously off the mark. I think, though, perhaps, rather than making a cynical play for reconfirmation, he is merely figuring (unconsciously) that it is better to be conventionally wrong than unconventionally right. Central bankers are boxed into the low inflation orthodoxy. It's hard to think outside the box when you live inside it.

Andy Harless said...

Another comment after re-reading: I'm not sure your criticism of Prof. Krugman is valid. At least there's a way he can wiggle out. The Bernanke Fed has to be able to affect mean expectations about future policy in order for the "Krugman/DeLong proposal" to work. (Shouldn't Greg Mankiw get some credit, too: he actually wrote an op-ed about the subject AFAIK before Krugman or DeLong had mentioned it for the US.) But if you believe that the correlation between the mean and the variance is very low, even conditioning on a relatively high mean, then the Bernanke Fed need not be able to affect the variance. "To control inflation" sounds more like an expression about the variance than about the mean. And Chairman Bernanke refers to "confidence in the central bank’s willingness to resist further upward shifts in inflation" (italics mine). By my reading, this is a statement about the variance. My paraphrase: "If we set the target at some point that would induce the optimal expectations for the mean of the inflation rate (which need not equal the target itself), we will also increase expectations for the variance. ["Further upward shifts" are possible but not certain.] Then to paraphrase Prof. Krugman's response: "Oh, come on. Over a long horizon, the variance is going to be high no matter what. The effect of the higher target on the variance (even conditional variance for the distribution truncated on the left) is a drop in the bucket."

Bruce Wilder said...

If I were a Bayesian, using past experience to form odds, I would think the hypothesis that Bernanke is a very conservative Republican, determined to aid and preserve the plutocracy, at any cost to the rest of society has to be the odds-on favorite.

I think I would file the whole Bernanke nomination and confirmation by Democrats as yet another case of doing A, while wishing B.

If you don't the economy to be run in the interest of the plutocracy, instead of ordinary, possibly unemployed workers, then don't appoint Bernanke to the Fed.

And, if he sounds crazy and vicious, if he recommends "reforming" Social Security as an alternative to bank robbery, allow for the possibility that he is what he seems. He has a very consistent biography so far, and his testimony is not all that anomalous with it.

Kaleberg said...

Like Greenspan, Bernanke is a Republican party hack, and will do whatever he believes will most benefit the party. (If you assume this, you can predict behavior on the Fed with greater than 98% precision.) All the other economic argument stuff is just obfuscation, and no one, except a few economists who are extremely credulous, are expected to take it seriously.

Why else is a $2T tax cut considered good while a $1T health care plan considered bad? Why are deficits only a problem when the President is a Democrat, but good when the President is a Republican? Why else is unemployment not considered an economic problem? Why else would the head of the Federal Reserve suggest that the U.S. government default on its debt to the Social Security Trust Fund, but to no other party?

I could go on, but it is time that economists started ditching a few of Ptolomey's epicycles and studying the actual behavior of the planets.