Gauti Eggertsson and Miles Kimball: Quantitative Easing vs. Forward Guidance

In my October 5, 2012 post Ryan Avent on the Fed’s Plans to Keep Rates Low Even After Recovery is Underway I wrote

I had an extended email discussion with an economist in the Federal Reserve System (whom I will not name) who argued passionately that, given our lack of knowledge about what will work and what won’t, the Fed should be using both large-scale asset purchases and promises that it will keep stimulating the economy even after it has reached the natural level of output—planning to push the economy above the natural level of output for a while.

Now that Gauti is at Brown rather than inside the Federal Reserve System and the Fed has raised rates, the identity of this economist can be revealed: Gauti Eggertsson. Gauti was delighted by the idea of publishing our correspondence when I suggested it the other day. 

I think you will find this discussion interesting. For me, it is very close to the last moment before my efforts to advocate the elimination of the zero lower bound crowded out efforts to urge the merits of quantitative easing (though I did defend when Martin Feldstein criticized it the following June). For Gauti, these thoughts are a reflection of a large body of academic research he has done on these topics, as you can see from Gauti’s CV.  

You can tell we are both thinking things through, because both of us often follow up an email with another one giving additional thoughts. 

Gauti: Hi Miles,

So I just don’t get your objections to the Feds recent use of forward looking language and commitment. The Fed is after all legally mandated to trade off inflation and output (dual objective). It seems you base this objection is based upon that “Wallace Neutrality” does not hold and hence The Fed should focus on the balance sheet. This objection does not make sense to me.

So I think many people may be sympathetic to the notion that full neutrality may not hold – including myself – and I say this even as being one of the author of a proposition that extends Wallace original neutrality proposition to show that it even still holds in a DSGE model with New Keynesian frictions provided that we are at the zero bound (see first proposition in Eggertsson and Woodford  Brookings Papers on Economic Activity (2003)

In fact I think this proposition describes the neutrality people have in mind a bit better than Wallaces original one, as he was claiming it applied to all open market purchases even at positive interest rates).

The point is just that we just don’t know how much this proposition does not hold, i.e. all empirical evidence suggest that there is a great deal of uncertainty about by how much “buying stuff” increases demand.

The point in my original paper with Mike is that one important dimensions which the neutrality proposition DOES  fails, is that you can affect demand by changing expectation about future interest rate, even in the New Keynesian model (but we assumed a fixed interest rate policy rule in our proposition). Moreover, this channel is extremely strong according to that model which we illustrate in a simple model.

It seems to me that given the dire situation we are in we should be doing everything we can to stimulate demand. Sure, that includes massive asset purchases (we are now at about 3 trillion and counting!). What I don’t understand is why you want to throw away one very effective tool, namely manipulating expectation about future interest rates and inflation, which many of our models suggest is very effective. That seems to me deeply counterproductive.

In other words you seem to be acting as if there are zero costs to balance sheet actions and infinite costs to using forward guidance and policy commitment about interest rates. I don’t know in what model that sort of policy making would be optimal. I think under any kind of reasonable type of model uncertainty it would make sense to be acting on all margins. If asset market purchases are not working, hopefully forward guidance is. If forward guidance is not, hopefully asset market purchases will.  Etc.

Best – Gauti

PS. I like reading your blog, its great fun to read. Hope you  keep it up! 

Miles:  Thanks for reading my blog! 

I should clarify. Given the gaps in our knowledge, I don’t see any problem with forward guidance of the sort the Market Monetarists are recommending: here is our nominal GDP target (which can be adjusted upward if the growth rate of technology is higher than we expected or downward if not) and we will keep interest rates low until we are fully on track to hit it. This is more or less the same thing as price level targeting if adjusted for technology shocks as I said. I am worried about forward guidance that is promising to be more stimulative than that, which it was my understanding that some of the models recommend. Is that wrong?  

The focus on lengths of time (as opposed to states of the economy) bothers me, since I think we should be doing enough QE that we get there fast enough that the dates they have set are after the economy has already been fully recovered for a while.  

Actually, my understanding of the forward guidance as they are doing it now is that they are not making much of a precommitment. Is the Fed actually viewing it as a precommitment? If they want to go further in precommitting, I hope they would make commitments relative to the state of the economy and not for a length of time.  

Monetary policy only has a lag of 9 to 12 months. If we do it right (many many more trillions of asset purchases) the economy should be fully recovered by Fall 2013. So it doesn’t make any sense to commit beyond then if one focuses on time. Focusing on the state of the economy the commitments could be quite helpful as you say because they would be a backup in case the QE doesn’t do the trick.   

Am I missing something? 

Miles: By the way, an important part of my argument that it should only take 9 to 12 months is that we should be just as willing to overshoot as to undershoot. Instrument uncertainty in relation to QE should make us a little conservative, but not too much.  

Another part of my consideration in saying what I did is that I am pretty sure that I am a small enough fish at this point that nothing I say will reduce the confidence people have in the Fed’s forward guidance, which the Fed is  going to do anyway. By the next recession after this one, I think we can have departures from Wallace neutrality much better figured out and either get *more* serious about forward guidance if QE is weak and costly or emphasize it less if QE is powerful and not too costly at the right dosages. So I am trying to provide more options for the Fed in the future.  

Miles: One place I am coming from is that I think interest rate smoothing is a terrible idea. When we need to move the economy in some direction, the Fed funds rate should be moved very very fast in my view. I know you have a paper justifying interest rate smoothing to some degree, but it just seems to me that if I have a continuous-time model and turn up the power on the microscope the optimal fed funds rate has to look like a random walk at a high-enough magnification. I somehow doubt that that gets to be a bad approximation at an every-six-week frequency. 

Am I missing something?

Gauti:  Provided the target is high enough, the nominal GDP target maps relatively well into the optimal commitment, as least in the simple New Keynesian model, which is why people like Mike Woodford have been proposing it. What many have against it, is that actually it may be too stimulative relative to what that the current stance is as it they worry it may imply a commitment to lots of inflation, in case real growth does not pick up. The bottom-line of all these models, however, is that it is optimal to commit to some amount of future inflation and/or output boom beyond what a discretionary policy maker might like to do at that future date. Whether this is done via commitment to some price level, nominal GDP target, etc, is more a question of “communication”.

The calendar date has in my mind never been particularly elegant. The best thing I can say about it is that perhaps it created the perception that the Fed would be slow to raise rate, slower than people otherwise might have thought, and that putting that date down raised to cost of pre-emptive tightening. The problem is that those sort of announcements may just feed into pessimism, which I take to be your concern, as you give the impression you are keeping rates low for long, not to stimulate the economy, but because you think things will be crappy for a very long time.

As for firm commitment, I think this part of the last FOMC statement was at least helpful in that respect

“To support continued progress toward maximum employment and price stability, the Committee expects that a highly accommodative stance of monetary policy will remain appropriate for a considerable time after the economic recovery strengthens. ”

Regardless of intentions of the Fed, I think the market is interpreting this as a commitment to allow for inflation to overshoot it long-run target some as the economy starts recovering, since the Fed will be trading off inflation and output then in accordance with its dual mandate, as it is trading off inflation and output today (intra and intertemporally). You can actually see this pretty clearly in tips markets. That’s a good thing in my mind. It should reduce the real interest rate and make some of those firms sitting on piles of cash wonder if investing it is not better than sitting on the money .

With respect to more asset purchases, there are two issues as I see it. One, their effect is uncertain, which is of course no reason not to do them (in fact that may instead be a reason to do even more to make sure they are doing anything). The other is that in practice it is actually not all that trivial to design which assets to buy and communicating it to the public, since things like MBS are not in unlimited quantities. So one has to start buying bunch of privately issued paper that has risks. Moreover, I think one way in which these purchases have effect – perhaps even the only way they have an effect – is by signalling something about future policy stance, thus working through inflation expectations. If one takes that perspective, it seems much more straight forward to communicate directly what one is trying to achieve, in particular commit to some inflation. Ideally one should do both at the same time, and I think the last statement was a bit in that direction.

My main problem with how you were writing about this, was not so much that you were for asset purchases based upon you the “Wallace neutrality” failing. That’s all fine. I don’t know how much they will do, but by all means lets try. What troubles me, and perhaps I’m misreading your views, is that you seem to think that they are so effective that there is no reason to use any other tools that may very well work much better, such as committing to somewhat higher inflation and lower future nominal rates. Given the uncertainty of the effectiveness of all tools, I think a reasonable policy should be acting on all margins. There is also the issue that we should not pretend that balance sheet actions are not costly, there are certainly big risks there as well for tax payers (the Icelandic Central Bank, for example, lost approximately 30 percent of GDP due to having taken bad collateral in the crisis in Iceland).

Gauti: I think the instrument uncertainty can actually cut both ways (I gather you may here have in mind Brainard’s 67 paper?). In this case, I actually think it might point towards the direction of doing more. 

Gauti: So the interest rate smoothing aspect of my work as it relates to this crisis is simply that even if you can’t cut the rate below zero, you can have an effect by making a commitment to keep rates low for some time in the future in a way that was different from your historical reaction function. I’m not quite sure what you mean with respect to the continuos time analog of the model, and that it would imply random walk. I’m not really seeing that. In the models I’m used to work with then usually the optimal thing to do is to make the nominal interest rate track the “natural rate of interest” as closely as possible, i.e. the real interest rate that would take place in absence of nominal frictions. To the extent that the natural rate is very volatile, then the nominal interest rate may also be quite volatile (I’ve a problem seeing it ever being a random walk, however, as it got to be bounded in some range given all the restrictions imposed by the model). 

Miles: You haven’t been misreading me. I have been saying that the asset purchases are so effective that we should not be committing to something later on that would not be optimal from the perspective of later. The concern is about having more than the optimal inflation then. I don’t have any problem with buying risky assets, in fact I think that is much safer for the Fed than buying long-term Treasuries because of predictable capital gains and losses on those if the purchases are having any effect at all. There is a good argument to made that in this particular crisis we might need some precommitment–especially given the fact that the Fed has been so limited by law in the assets it can buy it can’t really do the asset purchases the way they should be done, but since I think I will have very little effect on what happens during the current crisis, I am focused on the underlying principles of how things should be done in general. But the worst of all worlds is for the Fed ex post to think it made a stronger precommitment than the markets read it as having. Since the Fed is, in fact, making fairly weak promises, I don’t want the Fed to think it is making strong promises while the markets thought it was making only weak promises.  

What you are saying about instrument uncertainty pointing to doing more is important. Could you explain more? 

Any diffusion process looks like a random walk when you turn up the power on the microscope because the drift gets small much faster than the fluctuations. I am claiming that 6 weeks is a short enough time that whether the Fed increased the Fed funds rate last meeting *ought* to have very low predictive power for which way it jumps this meeting. What happens this meeting to the Fed funds rate should be *mostly* about information that arrived in the last 6 weeks, which is equally likely to favor raising the rate as lowering it. There would be a little drift, but not much. This is away from the ZLB.  

Miles: Two more things:

1. If asset purchases do affect expectations, then why not use them, since they won’t actually tie the Fed’s hands later on when we figure things out better? However, I want to avoid putting the Fed in a position of taking huge capital losses if and when it tightens if we can help it, so I don’t want to affect expectations through that channel. But if it affects expectations by showing everyone that the Fed isn’t excessively hawkish, that is great.  

2. Is there any paragraph or two in existence that gives a good intuitive explanation of the virtues of price-level targeting as opposed to inflation targeting that really gets to the heart of what is going on mathematically as well as being intuitive? If not, it would be great if you could try to write that paragraph or two. 

Gauti: What I find odd about your perspective, is the unwillingness to trade of some output by committing to some inflation, yet at the same time recommending very aggressive balance sheet action. It seems to me relatively clear that given the uncertainty about the effects here, we should do both, i.e. explicitly state we are willing to tolerate some inflation going forward and intervene in various asset market. My sense is that the latter is actually more costly than the former. (Was it very costly, for example, to bring down inflation from four percent to two in 1994-2004). I’m just not seeing the case against over-shooting inflation from the long term objective of the Fed, given what we know, and the dual objective that tells us to trade the two off.

What I was saying about  doing more if you have instrument uncertainty just comes out of Brainard 1967 AER paper. I think people often misread his paper as saying that under uncertainty you should do less that compared to certainty equivalenve. But as he shows, this would only be the case of the uncertainty about your policy is uncorrelated with the size of the shock you are reacting to. If the effectiveness of policy is correlated with the severity of the real shock you can easily get that you should act more the more uncertain you are about policy effectiveness (imagine for example that in the states of the world in which output is very far away from potential is also the state of the world in which policy has very small effect. Then you want to do more the more uncertain you are about effectiveness…..)

Gauti: On those two points:

1. Well you may argue that asset purchases affect expectations because they directly change the asset composition of the fed, eg buying long term debt makes it now more costly to raise short rates due to the resulting balance sheet losses. Anticipating this, people expect lower nominal the more the fed holds of long term debt….

2. Mike Woodford and I discuss the price level target in our BPEA in 2003. The basic idea is just that a correctly specified price level target increases inflation expectations and thus reduces real rates, increasing demand. A nice feature of a PLT is that if you miss it, i.e. there is deflation, then inflation expectations will increase even more, as now you have accumulated a bigger “price-level gap” which you need to fill. Thus there is an “automatic stabilizer” element to it.

Miles: Two more things:

1. If asset purchases do affect expectations, then why not use them, since they won’t actually tie the Fed’s hands later on when we figure things out better? However, I want to avoid putting the Fed in a position of taking huge capital losses if and when it tightens if we can help it, so I don’t want to affect expectations through that channel. But if it affects expectations by showing everyone that the Fed isn’t excessively hawkish, that is great.  

2. Is there any paragraph or two in existence that gives a good intuitive explanation of the virtues of price-level targeting as opposed to inflation targeting that really gets to the heart of what is going on mathematically as well as being intuitive? If not, it would be great if you could try to write that paragraph or two. 

Miles: What about this point that I made, though:

But the worst of all worlds is for the Fed ex post to think it made a stronger precommitment than the markets read it as having. Since the Fed is, in fact, making fairly weak promises, I don’t want the Fed to think it is making strong promises while the markets thought it was making only weak promises. 

In other words, I think there is a key issue of not interpreting a commitment as being any stronger than the markets think it is.

Miles: To answer your question about the overall perspective, since I think my voice is a small one, you should interpret the point I am making as a marginal adjustment point: I really do believe that *relatively speaking* people are overemphasizing forward guidance relative to making the asset purchases large enough. So I really do think there should be a shift toward larger asset purchases–large enough that the amount of forward guidance could be reduced somewhat. As you are saying, I don’t think I would take a strong stand in the current situation that forward guidance should be eliminated entirely given our uncertainty about how all the policies work. But it really does bother me how long into the future the forward guidance is extending. And I just don’t see why a full recover should take more than 12 months if we do enough asset purchases. If we made forward guidance conditional on how the economy is doing rather than timing, and did huge asset purchases that should make the forward guidance moot because the economy would recover anyway if I am right, of course there is no problem.   

Miles: And I definitely am willing to risk extra inflation by doing asset purchases that *might* be too large. But it bothers me to *plan* to do something that will definitely be too much in the future.  

Miles: As I think about it more. I think there is a big political cost of the precommitment to overstimulation. It would be remembered for a long time that the Fed overdid the cure and caused a lot of inflation that it did not bring down very fast because it had promised not to. So the whole idea of doing dramatic things to stimulate in a situation like now might be (inappropriately) discredited. People remember the end. (One of Danny Kahneman’s favorite results.)

By contrast, if we do big things in a way where we can reverse course very fast if needed, the end will look like it was being done well. That is true even if, say, there is a side effect of capital losses to the Fed, which is the side effect of large-scale asset purchases that I worry about. 

Miles: Also, I worry a lot that the doves lose out to the hawks if the doves don’t include *some* hawkish statements in what they say. Precommitting to have too much inflation makes it very hard to maintain one’s credibility as someone who will keep inflation down. 

Miles: I feel that the foreward guidance in 2001 or so, which led to low interest rates in 2003 had exactly this kind of discrediting political effect. If the Fed hadn’t felt bound by the foreward guidance it had given, wouldn’t it have raised interest rates faster in 2003? And then maybe there wouldn’t be the false canard about how the Fed caused the financial crisis by those low interest rates in 2003.

Miles: One more consideration I have is that I think we are still learning a lot about how to do monetary policy, fairly fast. Precommitment makes it harder to use the new things we learn in the meantime.  

Gauti: I think there is always the risk that what a central bank says is not credible, and then it may be more costly to fulfill the pledge if you don’t manage to convince people. It seems to me that this is an important consideration, but should not be overdone since you have ways to back up your words. IF the Fed made clear statement about being willing to tolarate some overshooting of its inflation target, my sense is that it would widely be viewed as credible and expectations would adjust. But I agree there should be a backup plan about what to do if its not credible, e.g. as measured by market expectations. I see buying stuff (balance sheet actions) as playing mainly a role there. So the Fed says its going to overshoot on inflation…. the market does not believe it ….. the Fed buys stuff until the market gets the message….. (and note here that wallace equivalence type violations work in tandem with the objective of shifting expectations).

Gauti: I agree that the emphasis on time commitment is a bit problematic. With respect to if people over emphasize forward guidance versus quantitative easing, my sense is that it is exactly the other way around. We have expanded the balance sheet by 3 trillions, and done qe1, qe2 and qe3. And it is only in qe3 that there is any meaningful forward guidance (the “commitment” about time duration had much more the sound of a pessimistic forecast than any meaningful commitment).

Gauti: Ah, but that basically says that you are uncomfortable with anything that is not time consistent. But most things we do are not timeconsistent (most punishments mandated by law, for example….) And by the way, so is a commitment to low inflation……..

Gauti: Cost – yes. But the alternative — very high unemployment and a balance sheet expansion that may not accompolish anything is also a very costly affair.

Gauti: I think this is real concern, and why a commitment to a fixed price level target, for example, is problematic. But it seems like a problem you can solve by making the commitment a bit broader taking those considerations into account….

Miles: I think just saying that it wants to get prices on a track that would be a 2% annual increase since 2008 (even though that means tolerating some catchup inflation) would be a big step and a good one.