Tuesday, September 4, 2012

Woodford at Jackson Hole

Mike Woodford's Jackson Hole paper is making a big buzz, and for good reasons. Readers of this blog may be surprised to learn that I agree with about 99% of it. (Right up to the "and hence this is what we should do" part, basically!)

Any student of economics should read this paper. Mike lays out in clear if not always concise prose, and remarkably few equations, the central ideas of modern monetary economics, on all sides, along with important evidence.

Mike's central question is this: how can the Fed "stimulate," now that interest rates are effectively zero, and given that (as Mike reviews), "quantiative easing" seems extremely weak if not completely powerless? He comes up with two answers: (Hint: starting with the conclusions on p. 82 is a good way to read this paper!)

First, the Fed can make promises to keep interest rates low in the future, past the time when normally the Fed would start to raise rates. He hopes that such promises would lower long-term interest rates, through the usual expectations hypothesis mechanism that long rates are expected future short rates. He is sympathetic to "nominal GDP targeting" as a way to commit to those promises.

Second, drop money from helicopters, i.e. "coordinated monetary-fiscal policy." Basically, the Treasury borrows money, writes checks to voters ("helicpoters"), and the Fed buys the debt. I certainly agree the latter policy can create inflation (I wrote as much in "Understanding Policy"), though both Mike and I  emphasize that policy needs some expectations and commitments asterisks too.

Why monetary stimulus?

One reason I disagree so little with the analysis of this paper is because of the part that Mike left out (rightly, it's already 98 pages): Mike didn't explain why he thinks more monetary "stimulus" is a good thing right now.

Treasury rates are at 50 year lows. The 10 year Treasury rate is 1.5%. At 2% inflation, that's a negative 0.5% real rate. Yes, the economy is in the toilet, but surely too-high Treasury interest rates are not the crucial economic problem right now.

So the case for "stimulus" must be that some other, unstated lack of "demand" is the problem, and that all "demand" is the same so that monetary "stimulus" will cure that problem. I disagree on that one.

Mike's enthusiasm for deliberate inflation is even more puzzling to me.  Mike uses the word "stimulus," never differentiating between real and nominal stimulus. Surely, we don't want to cook up some inflation just for its own sake -- we want to cook up some inflation because we think it will goose output. But why? Why especially will increasing expected inflation help? Because that is the aim of all the policies under discussion here -- promising to keep rates low even once inflation rises, adopting "nominal GDP targets," helicopter drops, or similar policies such as raising the inflation target.

I don't put much faith in Phillips curves to start with  -- the idea that deliberate inflation raises output. I put less faith in the idea floating around Jackson hole that a little inflation will set us permanently back on the trend line, not just be a little sugar rush and then back to sclerosis.

But it's a rare Phillips curve in which raising expected inflation is a good thing.  It just gives you more inflation, with if anything less output and employment.

So, in my view, the problem isn't overly tight monetary policy. The economy's problems lie elsewhere. Monetary policy is basically impotent. And it's hard to see that deliberate monetary "stimulus" via expected inflation will help the real economy. We should be telling the Fed to stop pretending to be so all important. You've done what you can. Thanks. You'll do best now by sitting on your hands and letting others cure the real problems.  But that kind of advice doesn't get you (me!) invited to Jackson Hole! The Fed wants to "do more."

So, let's leave alone the question whether a bit of deliberate inflation is a good thing -- I think not, but that's where we disagree -- and analyze Mike's proposals for  how the Fed can create some inflation. Here I mostly agree, with a few asterisks.

Open mouth operations 

So, interest rates are stuck at zero. Can the Fed do anything about it? Many economists have advocated promising that rates will stay at zero further in the future. I've been a bit sceptical of this advice, for example in" Understanding Policy"
I read this move as sign of desperation. Teddy Roosevelt said to speak softly but carry a big stick. These steps are speaking loudly because you have no stick. What will the Fed do if it announces a higher target but inflation does not change? [Announce a larger one still?] We are here in the first place because the Fed is out of actions it can take. Talking is the ‘‘WIN’’ (Whip Inflation Now) strategy that failed in the 1970s. 
More generally, I'm skeptical of the idea that wise governance consists of "managing expectations" by government official's promises.

Mike starts with a review of the literature that studies whether announcements -- "open mouth operations" have had effects in the past. Here's a good example.

These are "Intraday OIS rates in Canada on April 21, 2009. The dotted vertical line indicates the time of release of the Bank of Canada’s announcement of its “conditional commitment” to maintain its policy rate target at 25 basis points through the end of the second quarter of 2010."

On many occasions Fed announcements, coupled with no actions, do move markets. Monika Piazzesi and I once looked at high-frequency data and came to the same conclusion.

But these what do we make of this fact? They certainly do not mean that the Fed can talk down rates at its pleasure. Mike briefly acknowledges one possibility: Markets do not interpret these announcements as changes in policy, or "intentions" but instead simply inform the markets of the Fed's deteriorating economic forecasts.   If the Fed gets news, or forms an opinion, that the economy will be weak, then future interest rates will be lower even if the bank follows the same old Taylor rule.  We can see this reaction even if the central bank has no influence at all over market interest rates (as in Gene Fama's latest) but has a decent forecasting shop. A coming recession means that interest rates will fall no matter what the Fed does about it, so long term rates fall now. Mike has a long section on open mouth operations that don't work, or go the wrong way, and pages of advice for central bankers on how to move markets the way they want.

Mike makes an excellent point though. Overnight rates last overnight. If the Fed has any influence at all on long-term rates, it is entirely through expectations. Talk may not matter, but expectations are everything.

Promises, Promises

Assuming that the Fed does have total control over short term rates, the answer to my Teddy Roosevelt quip is this: Yes, the Fed is powerless to do anything now. But the time will come that the economy recovers or inflation breaks out, and the Fed will want to raise rates. Those 1.5% 10 year rates reflect expectations over some paths in which short rates rise. If the Fed can credibly promise not to raise short rates, even in circumstances in which it would normally be expected to do so, then by expectations hypothesis logic today's 1.5% ten-year rate will decline, as will the implied 10 year real rate (we're assuming the Fed can hold short rates at zero in the future despite the outbreak of significant inflation.)

The deep, intractable problem with this idea is commitment. This occupies the bulk of Mike's analysis, but I don't think he, or others advocating these policies, successfully solves it.

Every day I promise that tomorrow I'm not going to have dessert. Every tomorrow I change my mind. Because I can. Tomorrow, if inflation breaks out, the Fed will want to raise rates sharply.

How can the Fed promise today to do something it will very much regret tomorrow, and get people to believe that promise?  More deeply, how does the Fed commit to allowing "just a bit" of inflation in the future, and not starting down the path of the 1970s again?

Here (p. 42, 44) Mike comes out in favor of a nominal GDP targets. In his view, they're not as good as the optimal policies he and Gauti Eggertsson have calculated, but clarity and communication are important, and Mike can see that nobody but he and Gauti understands the optimal policy.

Nothing communicates like a graph. Here's Mike's, which will help me to explain the view:

The graph is nominal GDP and the trend through 2007 extrapolated. (Nominal GDP is price times quantity, so goes up with either inflation or larger real output.)

Now, let's be clear what a nominal GDP target is and is and is not. Many people (and a few persistent commenters on this blog!) urge nominal GDP targeting by looking at a graph like this and saying "see, if the Fed had kept nominal GDP on trend, we wouldn't have had  such a huge recession. Sure, part of it might have been more inflation, but surely part of a steady nominal GDP would have been less recession." This is NOT what Mike is talking about.

Mike recognizes, as I do, that the Fed can do nothing more to raise nominal GDP today. Rates are at zero. The Fed has did what it could. The trend line was not achievable.

The point of a nominal GDP target to Mike is this: When and if inflation breaks out (which raises nominal GDP) or (let's hope) real GDP starts growing again, the Fed, following the usual Taylor rule linking interest rates to GDP growth or inflation, would normally raise rates. If the Fed instead changes to a nominal GDP target, then the Fed will not raise rates, until the cumulative inflation or real growth brings us back to the dashed line. Then, and only then, will the Fed raise rates.

And, it will (supposedly) use all its hard-won anti-inflationary toughness to keep nominal GDP (inflation at that point) from growing faster than the trend line. In fact, it will become super-tough. In the past, with an inflation target, the Fed swallowed inflation shocks. With a nominal GDP target, the future Fed will supposedly commit to a slow deflation after a 1% surprise inflation shock, to bring the level of nominal GDP back, just as now it is committing to a substantial inflation to bring up the nominal GDP level.

In sum, this nominal GDP target discussion is not about what the Fed does now, or what it should have done in 2008. It is not about whether over the long run a nominal GDP target is better or worse than a Taylor rule (roughly, its first difference), which is a good topic for another day. It is a proposal to manage expectations about what the Fed will do in the future, and its hope is to lower long-term rates now.

Sounds good? Not so fast. Odysseus had himself tied to the mast so he could not change his mind. The Fed is changing rules now, in response to extreme conditions. What stops the Fed from "changing rules" again, the minute inflation does break out? True precommitment means setting things up so you can't change your mind, or at least so there are substantial costs to changing your mind. When Woodford 2016 comes back to Jackson Hole saying, "to fight this galloping inflation we need to change to the Gold standard rule" what stops that?

"Rules" without costs are no better than promises. I don't just promise each day not to have dessert. I change each day to the "no dessert" rule. Each night, I change back to the "no dessert, starting tomorrow" rule.

Furthermore, people might be less worried about the tough anti-inflation Fed than the new we-want inflation Fed. The second promise of the nominal GDP target is to contain expectations that once inflation breaks out it explodes. One inflation breaks out, and the Fed isn't responding, will people really say "oh, that's the new nominal GDP target Fed, they'll get really tough once we get to the 2007 nominal GDP trend?" Or will people think "oh-oh, we've got the 1970s Fed on our hands again"!

Suppose it's 2016, inflation has brought nominal GDP to trend, but real growth is still stagnant, unemployment is still high, the eurozone mess is worse, and candidate Hilary Clinton's poll numbers are tanking. Will Mike--and maybe more importantly, Christina Romer, Paul Krugman, Brad Delong, and the rest of the dovish punditry  recently converted to nominal GDP targeting -- really stand up and say, "we're on the nominal GDP target. We have to keep our promises. Raise rates and open the bar early."? More importantly still, do people now believe that will happen?

(There is also a larger question here, why do we that people will believe fine-tuned promises from the Fed about some brand new, never-tried rule, about how it will behave 5 years from now. To the public, how are the Fed's promises different from annual rosy scenario budget forecasts out of every Administration? How many average Jay-Walking voters even know who Ben Bernanke is or what nominal GDP is?)

I think the lesson of all precommitment economics is, that if you want people actually to believe the commitment, it must have substantial costs to change. Making the target a legal restriction might do. But the Fed adamantly doesn't want any restrictions on its power.

If you cannot limit your power ex post, you cannot commit to anything ex ante. If you cannot commit ex ante to do things you will not want to do ex post, your promises are empty. Even if they are "rules" not "actions."

Mike beautifully sums up what we're looking for on p. 82,
Central bankers confronting the problem of the interest-rate lower bound have tended to be especially attracted to proposals that offer the prospect of additional monetary stimulus while (i) not requiring the central bank to commit itself with regard to future policy decisions...
That criterion dooms a nominal GDP target or any other promise that is not "forward-looking" or "discretionary."  

Especially the Fed. Institutions work from historical perspective, and the Fed regards itself as fresh from the great success of "unconventional" policy experimentation in the great crash of 2008. What, tie ourselves to some rule that might keep us from saving the world again with our innovative discretionary policy? Not a chance.

(And even a legal restriction, writing nominal GDP targets into the law, is no guarantee. The ECB has a legal restriction against buying sovereign debt. Ha Ha Ha.)

The Fed was an alcoholic in the 1970s. It went on a 12 step program, reformed in the 1980s, and not it's a teetotaler on inflation. It wants to promise to go back to being a social drinker -- just three drinks until my nominal alcohol target is fulfilled for the night. And it doesn't want to let its spouse pour the drinks.

Quantitative easing

Mike moves on to quantitative easing. Here, the Fed buys short term treasuries, long term treasuries or other securities, issuing money in the process. Does this "stimulate?"

Mike starts (p. 49) by masterfully destroying the theoretical idea that QE should work. Yes, monetarists think the quantity of "money" matters, even at zero interest rates. They believe that because they think velocity is stable. The historical experience behind that conclusion does not have long periods of zero rates. When interest rates hit zero,
the demand for reserves should become infinitely elastic, so that variations in the precise quantity of excess reserves (as opposed to other short-term, essentially riskless assets) that banks must hold will have no consequences for equilibrium determination. ...once that lower bound is reached, further expansion of the supply of reserves should not have any consequences for aggregate expenditure or the general level of prices (or for that matter, for broad monetary aggregates).
Mike goes on to skewer long term bond purchaes -- they are the same as ineffective QE plus a rearrangement of the maturity structure of debt, which at least should not involve the Treasury doing the opposite.

Starting on p. 60, he points out that no asset market purchases should have any effect. If the Fed buys mortgages or long term bonds, yes, the private sector seems to hold less risk. But the Fed is ultimately holding risk that is guaranteed by the Treasury and hence by your taxes -- The  Modigliani-Miller theorem of Fed impotence. The starting place should be that purchases have no effects.

Of course there are frictions, liquidity effects, and so on. But with this theorem, all monetary theory must be about really understanding the frictions. (I did say this is a great review of monetary theory! Students, pay attention to these sections) For example, the monetarist position that only the issuance of money matters, but what assets the Fed buys do not matter, comes from recognizing one and only one friction, the necessity of money for making transactions. Mike reviews all the currently hypothesized frictions underlying asset purchases. Go read.

Though Mike goes for frictions a lot more than I do, we end up at the same place: a logical conundrum. If the Fed can affect, say long-term treasuries because that market is segmented, cut off from, say, mortgage markets, practically ipso facto changing long term treasuries won't spill over into markets you care about such as mortgages
Second, the existence of market segmentation makes it possible for central-bank purchases to affect the price of an asset, but at the same time limits the generality of the effects of a change in that particular asset price on the rest of the economy. In order for the policy to be judged effective, it is necessary that influencing that particular asset price can be expected to achieve an important aim. In the case of the CPFF, this presumably was the case — only the financing costs of a particular narrow class of borrowers were affected, rather than financial conditions more generally, but the program achieved a specific goal that motivated its creation. One cannot, however, point to such a program as evidence that purchasing any kind of assets eases financial conditions generally. Instead, to the extent that market segmentation is relied upon as the basis for a policy’s effectiveness, one should expect the effects to be relatively local, and the composition of the asset purchases needs to be tailored to the desired effect.
Well, if it makes no theoretical sense, maybe it works anyway? Mike's graph here
is better art than the graphs I made in a QE oped here. QE is supposed to lower interest rates. You have to tie yourself in knots to get this graph to say that interest rates are lower in the grey periods when the Fed is buying lots of stuff.

The Fed and its defenders do: they point to the declines in rates just before QE episodes as evidence for QE's power, then point to the rise in rates as verifying that the economy got better.  Mike explodes this view deliciously (p.71). The view that only the announcement-day decline measures the effects of QE relies on efficient markets. And if markets are efficient, then QE doesn't work, because it relies on segmented markets.

Mike concludes with an interesting observation: the only way that it makes sense for QE to have any effect is not directly, but because it signals to markets just how desperate the Fed thinks the situation is, and therefore communicates that interest rates will be zero for a long time.

But that makes no sense (p.84 of the conclusion is quietly devastating on this widespread view.) QE has the same commitment problem. The only hope for it to work is for people to think the money will stay out there once interest rates rise above zero. But the Fed has loudly told us how easy it will be to soak up all this money the minute it needs to do it, which is reassuring for inflation. But the point was to stoke inflation!

Helicopter drops

So, in conclusion (p. 82 -- hey, at least the blog post is shorter than the paper!) Suppose the Fed wants some inflation, what should it do? The only thing that can create some inflation, if the Fed wants to do that is helicopter drops, which are really fiscal policy: (p. 87):
the most obvious recipe for success is one that requires coordination between the monetary and fiscal authorities. The most obvious source of a boost to current aggregate demand that would not depend solely on expectational channels is fiscal stimulus—whether through an increase in government purchases, tax incentives for current expenditure such as an investment tax credit, or subsidies for lending like the FLS.. At the same time, commitment to a nominal GDP target path by the central bank would increase the bang for the buck from fiscal stimulus, by assuring people that premature interest-rate increases in response to rising economic activity and prices would not crowd out other types of spending than those directly affected by fiscal policy. And the existence of the central bank’s declared nominal GDP target path should also limit the degree of alarm that might arise about risks of unbridled inflation when special fiscal stimulus measures are introduced.
The Treasury borrows and, with Congress, spends the money. The Fed buys the debt and issues money. That's how we do helicopters today.

Even helicopter drops aren't easy however.  If people think that the government will raise taxes tomorrow to pay back the debt, and the Fed will unwind the purchase, even helicopter drops don't cause inflation. There really is no escape from "expectations." Helicopters -- or boondoggle stimulus projects -- are thus a communication mechanism for the government to say, "no, we are not raising taxes to soak up this debt. We really are leaving the money outstanding so it will inflate. You'd better spend it fast." And that's just what Mike wants, more "spending." (See "Understanding Policy" for more).

But Mike is being inconsistent here. He told us how impossible it is to commit to a nominal GDP target. And he told us how irrelevant the maturity structure of goverment debt is. Not raising taxes is really a fiscal commitment not a monetary one. Why is Mike back to a costlessly chageable promise to target nominal GDP? I think he recognizes that the commitment not to undo the helicopter drop is crucial to his proposal, and so he has to rescue that somehow.

So, in the end, I find Mike and I in strong agreement on mechanics. IF the Fed wants to inflate, a helicopter drop is the only way to do it. Even that is about expectations. And it's essentially fiscal policy. And, of course, we have now arrived at a point that completely contradicts the intial search: A policy of announcements, open mouth operations, that the Fed can follow alone.

The question, which Mike does not address, is this: Why in the world would such a deliberate inflation -- which in this case is a deliberately-induced flight from US government debt, exactly what Europe is so desperately trying to avoid -- be a good idea right now?

The rest of p.82 is chilling really. It is a lovely statement of the Fed's problem:
Central bankers confronting the problem of the interest-rate lower bound have tended to be especially attracted to proposals that offer the prospect of additional monetary stimulus while (i) not requiring the central bank to commit itself with regard to future policy decisions, and (ii) purporting to alter general financial conditions in a way that should affect all parts of the economy relatively uniformly, so that the central bank can avoid involving itself in decisions about the allocation of credit. Unfortunately, the belief that methods exist that can be effective while satisfying these two desiderata seems to depend to a great extent on wishful thinking 
We saw how (i) dooms open mouth operations, and conversely dooms the chance the Fed can affect the economy by announcing any new rules and targets.

Yet the Fed wants to be powerful. That leaves (ii). "Allocation of credit" means lending to particular favored markets and institutions.  The Fed understands the huge danger of going here. Lending to cronies is how central banks operate in all the basket cases of the world. But, if the Fed is unwilling to say "Inflation 2%. Banks steady. Interest rates zero. We've done our job," and wants to stay powerful, direct lending (which is really fiscal policy) or direct intervention in the policies of the TBTF banks under its control is going to be increasingly attractive.