Does QE cause deflation?

Oh my gosh. I am really excited. For years, I've been waiting for a chance to disagree with Brad DeLong about something econ-related, and the day has finally come!! It's enough to make me break my blogging hiatus a few days early.

Remember back in 2011 when Narayana Kocherlakota theorized that low interest rates cause deflation? Well, on Wednesday, Steve Williamson made a similar claim, writing that in a liquidity trap, QE will cause long-term deflation. Williamson based his post on this paper.

In a testy response, Nick Rowe called Williamson's post "horribly wrong," lamenting: "What the hell has gone wrong with some of the best and brightest in economics?" Brad DeLong then jumped in, accusing Williamson of mistaking an unstable equilibrium for a stable one. Paul Krugman echoed that accusation.

But David Andolfatto, in this excellent post, showed that DeLong and Krugman's criticisms are misplaced. In a typical New Keynesian model - the kind that now mostly dominates business-cycle theory, and the kind preferred by Rowe and Krugman - it's true that Williamson would be picking an unstable equilibrium. But Williamson is not using a New Keynesian model! Williamson's model is actually quite different. And as Andolfatto points out, there are macro models out there that are very similar to New Keynesian models, but have one small twist that makes the "QE-causes-deflation" equilibrium the stable one!

So DeLong and Krugman have gone too far. They are arguing from their preferred model, and that's fine. But to say - as Brad does - that Williamson doesn't deserve a "union card as an economist" is wrong. And to say - as Krugman does - that Williamson has made a simple "misconception" by forgetting about stability is wrong. Williamson is simply using a different model than the standard model, and DeLong and Krugman have not yet examined that model carefully.

What's more, Andolfatto points out that Williamson's critics should be more empirical and less theoretical in their arguments:

The fact that Japan has spent decades [with deflation despite zero nominal interest rates and periodic bursts of QE] suggests that [such an equilibrium] may in fact be stable... 

It seems to me that the critics should have instead attacked his results and interpretations with empirical facts (or am I too old-fashioned in this regard?). After all, Williamson at least motivated his post with some data (the diagram at the top of this post). And he makes what is potentially a testable prediction (notice the if-then structure of the statement):

In general, if we think that inflation is being driven by the liquidity premium on government debt at the zero lower bound, then if the Fed keeps the interest rate on reserves where it is for an extended period of time, we should expect less inflation rather than more.

Yes. Although macro data is not generally very conclusive, it's important to look at it anyway. Williamson's model, if I'm not mistaken, predicts that QE will cause a short burst of slightly higher inflation, followed by a long period of lower inflation disinflation in the long-term. That seems consistent with what happened after Japan's QE during the Koizumi years. It also seems consistent with what the U.S. is experiencing now. (As for Japan's new Abenomics QE, it is too early to tell.)

Of course, Krugman, DeLong et al. have their own explanation for Japan' s ongoing deflation (and the U.S.' ongoing inflation) - the "secular stagnation" hypothesis. And that's fine. To determine whether QE's failure to cause inflation is caused by a Williamson model, secular stagnation, or some third phenomenon is far beyond the scope of this blog post.

My point is to say that neither theory nor data tell us that Williamson's model is obviously wrong.

(Well, OK, actually, that's not quite true. As Nick Rowe points out, Williamson's model implies that QE will cause a never-ending deflationary spiral. Obviously that's not realistic. But that kind of infinite spiral exists in most New Keynesian models too, if a central bank keeps interest rates too high. Generally, none of these models makes any sense in extreme cases.)

Now, on plausibility grounds, Williamson's model is suspect - he assumes that Congress follows a certain policy rule, and in fact his results depend on Congress acting in that one simple predictable way. That's almost certainly not very realistic. So don't read this post as me endorsing the Williamson model! Then again, New Keynesian models - and indeed, all of modern macro models - contain huge lists of equally unrealistic (and empirically falsifiable) assumptions about the behavior of economic agents. So don't get me started down that road...

But anyway, I completely disagree with DeLong's and Krugman's criticisms of Williamson. They shouldn't have pounced on the stability thing. And as for Nick Rowe, as nearly as I can tell, he's just angry that anyone in the world could have anything other than a monetarist/New Keynesian view of how the economy works. And instead of getting mad, I think he should be engaging with alternative models. (Update: Rowe has a response.)

OK, so now let me move on (as Andolfatto also does) to giving Steve Williamson a hard time. 

In his new 2013 paper, Williamson says that:

1. QE causes low interest rates and deflation,

2. That's a good thing, because deflation is good, and

3. Fiscal stimulus (i.e. having the Treasury issue more debt) would be effective in getting us out of a depression.

Fine. Good. These things flow right out of Williamson's model. Paul Krugman would actually seem to have a lot to agree with in #3! 

BUT, a year and a half ago, Williamson wrote this:

I think some serious inflation is coming, maybe sooner than later. The Fed thinks it can control this with reverse repos and term deposits at the Fed. No way. When will the inflation happen? In line with this post, look out for increases in house prices. The higher house prices will support more credit, both at the consumer level, and in higher-level financial arrangements. The "bubble" will grow, and support the creation of more private liquid assets, which will in turn substitute for publicly-issued liquid assets, causing the price level to rise.

His prediction was based on this 2012 paper.

These two predictions are exactly opposite!!!

(Oh, and to round out the set of predictions, here Williamson argued that because of Wallace Neutrality, QE can't affect inflation one way or the other!)

Maybe Williamson changed his mind about how the world works between 2012 and 2013. Great! Fine! He should write a post about what made him change his mind. 

BUT, what Steve and I usually argue about is the general state of macro - he says macro is in fine shape, I say it hasn't discovered much. I think this reversal supports my thesis. If a top-flight macroeconomist, who knows the whole literature backwards and forwards, can so easily change his workhorse model in one year, and reverse all of his main predictions and policy prescriptions, then good for him, but it means that macroeconomics isn't producing a lot of reliable results.

Sure, I know that Williamson (2012) and Williamson (2013) have different sets of assumptions, and that's why their conclusions are so opposite. But how does Williamson expect us to tell which set of assumptions corresponds to the real world, and which is just fantasy-fun-land? The papers, of course, offer no guidance, which is utterly normal for macroeconomics. A million thought experiments, and no way to tell which one to use. Is this science, or is this math-assisted daydreaming?

Anyway, to sum up, the only person I really agree with in this whole debate is Andolfatto. Which is totally unrelated to him buying me a beer at the St. Louis Fed conference earlier this year. 


Updates:

Brad DeLong responds to me in the comments. Steve Williamson joins in. I'm relieved to find I still disagree with Brad!

Steve Williamson responds to DeLong/Krugman/Rowe on his blog.

Steve Williamson responds to me on his blog. Key quote:

Back in days of yore, my concern was that we could indeed get higher inflation. How? I had thought that the Fed had the ability to control inflation, but when push came to shove, they wouldn't do it. Once people caught on to that idea, we could get on a high-inflation path that was self-sustaining. Of course, since I said that, I've continued to work on these problems, and stuff has been happening. In particular, we're not seeing that high-inflation path. 

So a year of low inflation made Steve go looking for an alternative model to explain the fact that QE wasn't causing inflation. And the feature of the model that Steve changed was his assumption about the behavior of policymakers. That clears things up a bit. And Krugman will be happy to know that Williamson updated his priors. But I think my criticism stands. If one year of data and one change in assumptions can utterly reverse both the prediction ("QE causes inflation" --> "QE causes deflation") and the policy recommendation ("less QE" --> "more QE, and maybe more fiscal stimulus"), I still think this is a teachable moment about the limitations of modern macro. If we have to change our working model of the macroeconomy to fit the most recent macroeconomic event, that demonstrates how silly the whole endeavor is, right?

Scott Sumner chimes in, agreeing with Nick Rowe. So does David Beckworth, who does some quick-and-dirty data analysis. Beckworth and Williamson argue in the comment section.

Tyler Cowen chimes in, agreeing with me.

Krugman has a response. He says that the basic problem with Williamson's model is that there's no explanation for why firms have an incentive to change their prices in response to QE - and hence, no micro-level explanation for why QE causes deflation. That's a legit question, of course. I think that in Williamson's model, QE just lowers aggregate demand by depriving banks of usable collateral (long-maturity govt. bonds), so that they need to hold more cash. That's like a liquidity preference shock, which causes cash hoarding and lowers AD, causing firms to lower their prices. I'm not 100% sure, and it doesn't seem particularly realistic, but that's how I think the model works. It turns out that in Williamson's model, things that increase output decrease inflation. So there is sort of a reverse Phillips Curve, at least when the interest rate is at the ZLB.

Scott Sumner gets increasingly annoyed at Williamson.

More discussion with Brad in the comments.

Brad DeLong responds on his blog. Key excerpt:

[One story of how QE causes deflation] is: (i) people think “there is going to be a deflation; I need to hold more currency”; (ii) people sell their assets to the Federal Reserve, and so the Federal Reserve injects more currency into the economy; (iii) people notice that they have more currency, and think “it would not be rational for me to hold this much currency unless there were a deflation going on”; (iv) people think “if there is a deflation going on, I need to cut my prices in order not to lose my customers”; (v) people start cutting their prices, and the deflation begins. 

This is also a fine story. But it is not a story of the Federal Reserve undertaking Quantitative Easing. It is a story of a self-fulfilling deflationary-expectations panic.

This is true, but I'm not sure that this is what is going on in Williamson's model; in his model, I think the QE (and the weird fiscal policy response that it induces) actually starves banks of collateral, which reduces aggregate demand and causes deflation that way. It all hinges on the way Williamson models the behavior of banks. I'd definitely like to see Williamson explain this part better, but I have kind of run out of time to pay attention to this issue at the moment...

Steve Williamson responds to Krugman. This response is pretty interesting, as you can really see the differences in philosophies of how macro itself should be done.

Karl Smith comes out of blogging retirement to enter the debate on the DeLong side.

Williamson writes a second, longer response to Krugman, trying to explain his intuition. He says that his results come from financial markets, not from goods markets. What's kind of interesting here is that Krugman/DeLong are basically saying that Williamson's model isn't microfounded enough, or at least the microfoundations of firm behavior aren't made explicit enough. Usually, Williamson would tend to be the guy saying "But how does it work?" while Krugman would be the guy saying "IS-LM works, just use it." 

And here is the second part of Williamson's explanation of his intuition.

Bob Murphy chimes in, on the side of DeLong and Rowe. It's kind of neat how this debate has totally shuffled the usual "opinion coalitions".

Nick Rowe has one more response to Williamson, and still thinks Williamson is nuts. Key line: "[Williamson's model] explains why Zimbabwe had hyperdeflation." *ZING*

Yichuan Wang, Scott Sumner, and Tyler Cowen have more. The econ-blogger field continues to be mostly anti-Williamson. 

Iza Kaminska jumps into the debate, endorsing Williamson's ideas and providing her own intuition for why QE is deflationary. 

Ryan Avent enters the debate, taking a rather nuanced position. He presents some quick-and-dirty evidence in favor of Williamson's assertion. 

Tony Yates says who cares if QE is deflationary or not, the point is that it didn't do what we wanted it to do.

Nick Rowe continues to say not-very-nice-things about the "QE causes deflation" supporters.

Tyler Cowen has a great roundup of evidence that QE is, in fact, inflationary. The evidence is not conclusive - evidence in macro almost never is - but it definitely puts Steve Williamson's story on the back foot.

Steve Williamson attempts to explain his "QE causes deflation" idea with a Socratic dialogue of sorts.

John Cochrane is interested in the question. He writes down a simple model with a Fisher equation and a Taylor rule, and finds that "higher interest rates cause inflation" (which is similar to "QE causes deflation") is a stable equilibrium in that model, which has multiple equilibria. Williamson says Cochrane "gets it." Nick Rowe and Scott Sumner respond to Cochrane.