« Helicopter money is normal, or else we're doomed! | Main | Language games and expectations of "doing nothing" »


Feed You can follow this conversation by subscribing to the comment feed for this post.

"I can hear your objection now...."

No, my objection is: announce whatever you like, if P* doesn't change it means the representative agent doesn't believe you. "Anything you do — monetary or fiscal — affects current consumption to the extent, and only to the extent, that it moves the expected future price level. Full stop, end of story." And as Krugman says, that means you can change G without changing C, which is why the G-multiplier is 1.

You might like to look at Simon Wren-Lewis's response.

Simon's response does not change the fact that fiscal policy cannot generate adequate growth, because by its very nature it remains in a beta position to money creation. The Fed continues to short growth because it is not yet convinced that time value matters. By restoring time value and making it a direct source of wealth in its own right, the inadequacy of helicopter money can also be overcome.

I like everything about this post. Why does the debate often turn into an anemic, semantic minefield about the difference between "monetary" and "fiscal" policy? Because so many people want to look for their keys under the streetlight. Krugman fixes P* and guts the expectations dance straight off the planet. Then he acts as if only financial frictions could impede his story, because even then, in his disclaimer, he refuses to unfix P*. And it is not crazy that he thinks this will be persuasive. Because so many other people want to fix P*in their models. Free P*!

Kevin: take the limit of Paul's model as the period gets shorter and shorter, and it becomes a continuous time model. Now let the central bank choose dM(t)/dt. That's equivalent in a discrete time model to choosing M(t+1)-M(t) in period t. So in period 1, the central bank can choose M in period 2. Done.

It's all in the framing.

"A central bank's monetary policy is a minor subset of those expectations."

A good thing to understand, but the case for economics becomes a rather trivial observation once that larger application is understood.

There is also a contradiction in the time analysis. Unless you define the current period as being infinitely short, it is always possible to change things in the current period by implementing in a residual sub-period. This was my immediate reaction reading the Krugman piece, and I think it makes the whole idea a little hokey in terms of any real world constraint.

Becky: I don't know what you mean by saying that fiscal policy is "in a beta position to money creation." Away from the ZLB, I'd have no problem with that claim, or at least I could square it with models that make sense to me. But the point that Krugman and Wren-Lewis are making is that at the ZLB, dG translates directly into dY because dC/dG=0.

Nick: if we see a change in P(t+1), then we know that the representative agent believes what the central bank says about M(t+1). If P* fails to move the central bank has failed. Now I don't doubt that a sufficiently flamboyant central banker could always shift P* but Mario Draghi, not so much.

Maybe you're making a point about continuous time which is too subtle for me. I tend to beware of economists who come bearing infinitesimals. But I think you're just assuming something that Krugman disputes, i.e. that central bankers can and will change expected prices. Is this just Scott Sumner's "can't means won't" doctrine?

I think I'm agreeing with your post. This is especially excellent:

"Time is not discrete. "Now" is a very very short period of time...The value of any financial asset, like money, is always and everywhere about expectations of the future. All financial assets are just promises, written on bits of paper. They are commitments about the future, and nothing more, and those commitments create expectations, and those expectations are what determine the demand for those financial assets...to talk about a world where there is a financial shock and the central bank is unable to make commitments and unable to create expectations and so unable to respond correctly to that shock is sheer nonsense.In a world without commitments there would be no money, no financial assets, no markets, no trade, no property rights; and the life of man would be solitary, poor, nasty, brutish, and short. What we call "society" is nothing more than a set of expectations about how people will act contingent on different things happening. The history of civilisation is the history of creating the right sort of expectations. A central bank's monetary policy is a minor subset of those expectations. What we call "monetary policy" is a set of contingent expectations. Economics is the study of society. Economics is the study of expectations."


"All financial assets are just promises, written on bits of paper. They are commitments about the future, and nothing more, and those commitments create expectations, and those expectations are what determine the demand for those financial assets"

I'm stealing this and putting it on my money & banking syllabus. Students will receive extra credit if they get a tattoo of it.


Please take the time to read Brian Romanchuk's comments on this topic:


Best regards,

Leo Kolivakis

Publisher of Pension Pulse blog.

A word to the wise, Nick: Never blog before coffee.

It seems to me you have to deal with time consistency. You can't just change p* to suit your purposes today, and then something different tomorrow. For now the policy rule is Taylor-like, and maybe sometime in the future it'll be NGDPLT, and maybe later something else. But that means we can never credibly commit to any rule very far into the future since future humans have minds and opinions of their own. So if a liquidity trap is expected to last a long time then it strikes me that there may not be anything the monetary authority can do.

K: we dealt with the time-consistency problem when we escaped the Hobbesian State of Nature. Or rather, the very fact that we are not in the Hobbesian State of Nature proves that we are continuously solving the time-consistency problem.

The very fact that stupid little bits of paper are valued proves that we have escaped the time consistency problem with money. Money is a commitment.


If you change your 2% inflation commitment to a 5% NGDP commitment then your inflation commitment wasn't credible. Therefore your NGDP target isn't credible either. If you want to be able to make credible commitments they have to be temporary. Otherwise you lose your credibility when, as right now, you want to make a different commitment.

An even bigger problem is that a change in commitment is an indication of fickleness. So a new "commitment" of a fickle monetary authority will be judged against the seriousness and longevity of the previous commitment which, in the current case, has been good for a long time. The market will suspect that the fickle monetary authority may be just as likely to return to the previous target in five years as they are, after all, fickle. Ie the sheer act of changing your "commitment" makes it not credible.

Start thinking about "expectations" as having a distribution , just like incomes and wealth ( and changing over time ) , and you might get more traction. As things stand now , you're simply preaching to the 1% choir.

Kevin: Paul assumes the central bank at time t chooses M(t), and can't change M(t+1) without making some sort of commitment.

Let me reframe: assume the central bank at time t chooses deltaM(t), and can't change delatM(t+1) without making some sort of commitment.

In other words, Paul assumes people expect changes in M last one period, unless the central bank makes a commitment otherwise. Let me assume people expect changes in M are permanent, unless the central bank makes a commitment otherwise.

Both assumptions are equally arbitrary.

dlr: "Krugman fixes P* and guts the expectations dance straight off the planet."

It's worse than that. By fixing expected P*, and expected M*, he begs the expectational question. Why should people expect those are fixed? Why should people expect that changes in M are temporary? Why not permanent? Why not any of the 1001 things they could expect?

Assume that each period, the central bank either "does nothing" or "does something".

If it does nothing in period t, then M(t)=M(t-1).

If it does something in period t, then M(t)=/=M(t-1) and it chooses M(t).

Therefore, I have "proved" that all changes in M(t) are expected to be permanent changes, unless the central bank can make credible commitments to do something in future periods.

So if it finds itself in a liquidity trap at time t, the central bank simply increases M(t), and people will expect that increase to be permanent (unless the central bank makes a credible commitment otherwise), so the economy escapes the liquidity trap.

If this is so simple why is the world in the state that it is?

Jrootham: because one (arbitrary) set of expectations gets reified into the natural order of things, and so "doing nothing" gets defined in one particular question-begging way, and people can't "see" the world any other way.

One particular "superstructure" gets built right into the very notation of Paul's model, so we can't see things any other way. If Paul merely changed notation, so that each period the central bank chooses m(t) rather than M(t), where m(t) is defined as his M(t)-M(t-1), then if the central bank increased m(t) in period t, and was expected to "do nothing" in all subsequent periods, people would expect a *permanent* increase in M for all subsequent periods, and an increase in P for all subsequent periods, and his model economy would escape the liquidity trap and would immediately return to full employment.

"Then the central bank does something stupid. At the beginning of period 1, it announces that the period 2 money supply will be cut by (say) 10%, and will stay at that lower level forever. And suppose people believe that announcement."

Does anyone write swaps on Velocity? I would really love to get long some V if this happened.

..and your definition of what is a financial asset, is truly deserving of compliments...

Nick, if all changes in M(t) are expected to be permanent changes then there's no steady-state for the price-level. It's just a random walk. Krugman's long-run price-level, P*, is undefined in that case.

The policy rule in his model is M(t)=M* for all t > 0 (where t=0 denotes the current period where all the action is). If the central bank can convince us that M* has truly changed then P* will change also, so escaping the ZLB is no problem.

Now, you can modify his model by assuming, say, a constant money-growth rule. If M grows at a decent rate then we're not likely to hit the ZLB in any case. But if the rule is too conservative then I think Krugman's conclusion will hold there too: we can only escape the liquidity trap by (credibly) replacing that conservative rule with a more inflationary one. (That's if we rule out fiscal policy which works in that case also.)

AFAICT modifications of that sort merely give us a messier model without changing the moral of the story. For example, we may find that although the long-run inflation rate is determinate, the price level isn't.


"Now is a short length of time." I thought central banks had a schedule at which they announced changes in policy. Isn't now = current three months.

Kevin: "But if the rule is too conservative then I think Krugman's conclusion will hold there too: we can only escape the liquidity trap by (credibly) replacing that conservative rule with a more inflationary one."

True. But what that means is that the central bank CAN do something in the current period to escape a current period liquidity trap. It can increase the current money growth rate. If we define the central bank's current action as choosing the money growth rate, then an increase in the current money growth rate, holding expected future money growth rates constant at their original level (because the central bank cannot credibly commit to its future actions), means its current action causes an expected *permanent* increase in the level of the money supply.

Or, we could define the central bank's current action as choosing the second derivative (wrt time) of M(t). Or choosing the current price of gold. Or, whatever.

It's all equally arbitrary. And civilisation advances when we choose a better over a worse one of those equally arbitrary expectations. But if you are stuck in the old ways of thinking, one set of expectations will look "natural", and the others will look "unnatural".

Derivs, and PSummers, and dlr, and JKH: Thanks!

I don't think anyone writes swaps on velocity.

Chris: The very fact that the Bank of Canada commits to a 6.5 week target for the overnight rate (barring emergencies) shows that it is already making commitments about the future. 6.5 weeks > 1 second.

K: sometimes we change the laws, or even the constitution. But they are still laws. There was always the risk the gold standard would be abandoned, but that didn't mean there was no gold standard. Yes, there is a tension here. But it's not all-or-nothing.

Except that the escape from the liquidity trap, in Paul's model and every other time consistent model, is determined by the asymptotic boundary conditions of the model. Paul merely sets up the final behaviour as does the NK model via the natural rate and the Taylor rule. In the real world, the asymptote is the one thing we have no idea about and which every agent may have their own opinion about. You can't imply a unique representative agent from disparate agents, some of which believe in asymtotic stability, others who believe in asymptotic fiscal dominance, and others yet who believe in asymptotic runaway deflation.

In all our models the asymtotic rule is known and determines everything that matters. Changing it temporarily (no matter for how long) doesn't change anything. And I don't know of any model in which it is even possible to transition from one asymptotic behaviour to another, or for that matter, change the rule. What we really need is some way to model the fundamental asymptotic uncertainty, which isn't going to happen, I think, in any representative agent model. Uncertainty is fundamental as are heterogenous opinions and learning.

K, yes, the models aren't really up to it. And that's because of more than heterogeneity. We could even have a perfectly representative agent who was ratex compliant but nonetheless employed heuristics and learning type dynamic reaction functions that we are unable to convincingly model while also explaining why the world doesn't appear to ping pong from one crazy equilibria to another. That is no excuse, though. There is little to say that such a model, with unhinged "final behavior" as you say but nonetheless offsetting features constraints to implosion/explosion, is not the better model of the world. P* is the can opener.

Great post about rational expectations. But in the real world, bounded rationality means the fifty women don't all kill their husbands on the same day.




While Paul Krugman may be frustrated about that people with real decision power (first exception Abe now?) do not take him and his endless simplifications and strawmen seriously ,

I wondered, whether this is the right place to warm up my old question:

“I am still waiting for one single positive response to my old question: Please tell me one specific Krugman PAPER, and in a very few sentences, what YOU specifically find good about it.” And is willing to defend his judgement versus questions from me : - )

After Simon van Norden didnt deliver what he promised.


How about Nick Rowe takes up the challenge with respect to Krugmans BPEA 1998 Liquidity trap paper here

I like the following response to krugman at the NYT best:

"Ron T
is a trusted commenter Mpls 2 days ago

The problem here is that assumptions are divorced from reality. It is like assuming a flat Earth and weightless airplanes. "


> What precisely does the central bank "doing nothing" mean?

From a macroeconomics point of view, let me ask rhetorically what the Central Bank does in the first place?

We conceive of the CB using monetary policy to set an Aggregate Demand curve, but we cannot observe the entire curve; we only see at any particular moment the intersection of the AD/AS curves.

In fact, the idea of an AD curve in the first place presupposes some sort of monetary policy rule, even if that rule is "we use shiny pieces of metal and we can't find any more." Discussing points off of the current, short-run equilibrium requires us to use implicit or explicit notions of monetary policy.

So, being less nihilistic, what does it mean for the Central Bank to "do nothing?" It means that the Central Bank does nothing when Aggregate Demand moves along the expected curve. It does something when AD changes away from the expected curve.

This argument is then in turn circular, because we don't know with certainty what AD curve the central bank is targeting in the first place. The Eurozone seems to be targeting more of an "AD region" that permits anywhere between 0% and 2% inflation. For a CB that blindly followed a Taylor Rule, then following that rule itself amounts to "doing nothing," even if interest rates and money supply growth are highly variable. Changing the rule, such as by changing the assumed natural real rate of interest, would be "doing something."

Krugman here seems to be conflating "doing something" with "doing nothing." His root argument is that the AD curve is not downward-sloping enough (that AD is not increasing quickly enough with a fall in the price level relative to trend); the CB actions that move along that trend do not suffice to raise demand. But by my language above, that's precisely "doing nothing" -- the CB will not be able to fix a liquidity trap by doing nothing.

If your headline is correct, I have an impertinent question. Why hasn't economics developed (or adopted from other social sciences) a theory of expectations?

Min: we have. Adaptive expectations/error learning, which (I think) was borrowed from other social sciences (psych?), then rational expectations, then various models of learning, including Bayesian learning.

Majro: you are right, I think. We always need to have some sort of "monetary policy rule" in mind when we draw an AD curve. Different rules give different AD curves, with different slopes, that are subject to different shocks. E.g., gold discoveries used to shift the AD curve, but don't (much) now.

I'd make two modifications to the headline-

ECONOMICS is always and everywhere about expectations AND PREFERENCES.

Though it's true that expectations are too often ignored, especially in monetary theory.

Thanks, Nick. :)

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad