Fiscal policy, ideology, and framing

TiC

Take a standard New Keynesian macro model. Assume it sometimes gets stuck in a ZLB liquidity trap, where monetary policy can't work.

There is a very simple solution: use fiscal policy.

Government spending should be cut whenever the economy is not in a liquidity trap. That is the clear policy implication of New Keynesian macroeconomic models.

I am very pleased to see that Canada's Conservative government is currently following that New Keynesian policy recommendation, and is cutting government spending now that the Canadian economy is no longer at the ZLB. All New Keynesians will applaud this sensible policy. Any opposition to this use of fiscal policy must be ideological.

/TiC

(See, it's all in the framing. Cutting government spending when the economy is not in a liquidity trap (which sounds righty) is equivalent to increasing government spending when the economy is in a liquidity trap (which sounds lefty). Given the standard consumption-Euler equation in simple New Keynesian models, the level of government spending has no effect on the natural rate of interest. (More generally, the effect could go either way, depending on preferences.) It's the rate of change of government spending that affects the natural rate of interest. By reducing the (expected) growth rate of government spending, the natural rate of interest is increased, which helps the economy escape the ZLB. Reducing the (expected) growth rate of government spending sounds righty, but it isn't. A lefty could argue that the best way to reduce the (expected) growth rate of government spending is to increase government spending today to an unsustainably high level, so that people will expect future cuts.)

It's important to ask the right question about fiscal policy. Especially in New Keynesian (and other) models where expectations matter.

New Keynesian models are different from Old Keynesian models. Fiscal policy in Old Keynesian models was about the level of government spending. Fiscal policy in New Keynesian models is about the (expected) rate of change of government spending.

I decided to write this post after reading Mark Thoma's post on Pontus Rendahl. I think there's another way to look at the relation between fiscal policy and ideology. We need to be able to see it both ways.

Comments

[Phil Koop]

"Cutting government spending when the economy is not in a liquidity trap (which sounds righty) is equivalent to increasing government spending when the economy is in a liquidity trap (which sounds lefty)."

No, it's not. Cutting government spending the moment the economy leaves the liquidity trap would be the corollary of increasing spending when it enters it - provided, of course, spending was actually increased in the first place. Even (New) Keynsians believe the economy will eventually exit a liquidity trap, however cack-handed the government may be.

"We need to be able to see it both ways."

What is this "both" of which you speak? Surely there are more than two ways to view the relation between fiscal policy and ideology. But to be brutally honest, I would be satisfied if you could come up with one. This is not one of your stronger posts.

[Stephen Gordon]

TiC = Tongue in Cheek?

[Nick Rowe]

Phil: In NK models, increasing G when the economy enters the liquidity trap is not what works. Cutting the expected growth rate of G when the economy enters the liquidity trap is what works. The former works only to the extent that it creates the latter.

[Nick Rowe]

Stephen: yep!

[rsj]

Given the standard consumption-Euler equation in simple New Keynesian models, the level of government spending has no effect on the natural rate of interest.

But those are knife-edge conditions without generic income uncertainty. As soon as you slightly deform the utility functions and add additive income uncertainty, the level of income plays a dominant role in the solution of the optimization problem.

Really you can only find exact solutions to the euler consumption problem or to the investment problem without consumption when your utility function is very special --e.g. log utility for the investment problem so that multiplicative shocks pass through the logarithm nicely, and quadratic utility, so that additive shocks pass through the marginal utility function nicely.

In a more generic situation, in which the symmetric of the marginal utility function is not the same as the symmetry of the income shocks, you cannot find closed form solutions at all, but the numerical solutions show that the level of income plays a dominant role. Just add a slight touch of realism so that you perturb the utility function somewhat, and things start looking more Keynesian.

[Nick Rowe]

rsj: Let's see. In the standard formulation, the natural rate of interest depends only on the ratio of present to future consumption, so it would be independent of the (permanent) G/Y ratio. If we depart from that simplifying assumption, it could go either way. A (permanent) increase in G/Y could either increase or decrease the natural rate, and so be either expansionary or contractionary (for a given r), depending on preferences. It might look more Old Keynesian, or it might look more anti-Old Keynesian.

[Frances Woolley]

Stephen - when in doubt, I look in urbandictionary.com. It does, however, list a couple of other possibilities...

[rsj]

I think a better rule of thumb is to use Friedman's proposed discount rate for permanent income, which was about 30%. So that means that effectively households have a time horizon of about 5 years. It is not that they have that small of a horizon, but when you add additive income risk, the effect of the optimization problem is to reduce that horizon very significantly.

So if the horizon is only 1 year -- no one is forward looking -- then government can lead the economy by the nose in the old keynesian manner. If the horizon is infinite, then the government can't do anything.

What I am arguing is that additive risk drastically reduces the effective horizon even for infinitely-lived forward looking agents, so that the economy looks more Keynesian.

[rsj]

And I should add that this is *very* sensitive to the level of individual wealth, or at the macro-level, the wealth distribution.

If one person has most of the wealth, and everyone else is living hand to mouth, then their horizon is very short, and the rich guys horizon is close to the perfect foresight (riskless) version, so the economy as a whole may have an effective time horizon of just one year.

If all the wealth is spread equally, then the economy as a whole has a longer horizon. Another example of how simplifying the assumptions to assume RA will give you the wrong intuition.

Keynesian policies are much more effective when society is more unequal, and become less effective as society becomes more equal. In some sense, you get diminishing returns with income policies, but as soon as you roll the policies back and wait a bit for wealth to concentrate again, then Keynesian policies become more effective again.

[Ralph Musgrave]

Nick uses the word “expected” four times in his post. Is there actually any evidence that households think about or pay much attention to expected changes in government spending or expected growth? Or, just to be cynical, is this simply a complicating factor that can be added to models, which helps keep academic economists employed?

[W. Peden]

Ralph Musgrave,

I doubt that most people think much about "government spending", other than economists and politicians, but we all plan partly on the basis of the future particular items of government spending on US. To take a topical example, people protest about tuition fee changes BEFORE they take place; it's as if people have expectations of the future.

As for expected growth: again, usually not "GDP" as a big aggregate, but do we really need massive empirical research to tell us that people plan on the basis of expected job opportunities and incomes?

Introspection isn't the methodological cornucopia that some Austrians apparentely think it is, but it does avoid some schoolboy errors like modelling agents who lack forward-looking expectations.

[Nick Rowe]

Ralph: I was going to talk about econometric tests of the permanent income hypothesis, but W. Peden's example beat anything I could have said.

Let me just say that in standard NK models, if people expected that government spending would be held constant in the future at whatever level it is at now, then changes in G would have no effect, even in a liquidity trap.

[Deus-DJ]

"I doubt that most people think much about "government spending", other than economists and politicians, but we all plan partly on the basis of the future particular items of government spending on US"

Ok I get it, so expectations matter more than income effects...right....and you can excuse this sort of thought as being ideological by blaming it on the model and not on your beliefs...RIGHT....oh but nobody can blame the model because old models didn't take into account expectations, and this one does, and hence it's superior....RIGHHHHHHHHT......

[Kathleen]

From "Real Government Spending in a Liquidity Trap" by Gauti Eggertsson, Princeton University;

"This paper explores the effects of real government spending in a New Keynesian model. ... Finally we analyze optimal monetary and fiscal policy under commitment in a liquidity trap. The optimal commitment solution involves counter cyclical fiscal policy. The government increases spending beyond its target rate in the trap and reduces it below the target rate it once out. The intuition is simple. The government can increase the natural rate of interest in a liquidity trap by one of two ways: either increase spending in the trap (relative to future spending) or commit to reducing it once out (relative to current spending). A government that can commit will both increase spending beyond the target rate in the trap and commit to reducing it below the target rate once out."

Is this an accurate depiction of the New Keynesian position?

[Nick Rowe]

DDJ: "Ok I get it, so expectations matter more than income effects..."

False dichotomy. The effect of expected future income is an income effect.

Does your spending this minute depend only on your income this minute?

Kathleen: that's fairly accurate. But it's not exactly right. If we wanted to be more precise we would think in continuous time, and restate it as: G should be falling (relative to trend) during periods when the economy would otherwise be in a liquidity trap, and rising (relative to trend) at other times. (in discrete time, if liquidity traps last for one period, it comes to the same thing.)

Oh dear, I really shouldn't be getting picky with people like Gauti Eggertsson on subjects like this, but Oh well. There is a genuine difference here between Old and New Keynesian positions, that people are missing, because what Gauti says sounds very much like the Old Keynesian policy recommendation. In Old Keynesian models, AD depends on the level of G. In NK models, it depends on the expected rate of change of G.

[W. Peden]

Deus-DJ,

I'd add to Nick Rowe's point that Ralph Musgrave didn't say "income effects matter more than expectations", but rather asked "Is there actually any evidence that households think about or pay much attention to expected changes in government spending or expected growth?"

Though Nick Rowe,

Isn't the income-expenditure relation only held to hold in a time period at a macroeconomic level? As a microeconomic fact, it is staggeringly obvious that income doesn't determine expenditure.

[Deus-DJ]

Nick, only if you're NOT assuming Ricardian Equivalence, which no doubt neoclassical theory does. Heterodox economics on the other hand understands that government deficits adds to profit via the Kalecki equation (or, hell, C+I+G+NX=Y....). If however we assume ricardian equivalence but then understand the Kalecki Equation, then the effects of expectations will only have impact in the very short term. This is of course what I mean by the income effect.

[Ralph Musgrave]

W.Peden, Taking your tuition fee example, obviously an individual household facing tuition fees in the future is likely to adjust current spending accordingly. But when it comes to what you call “big aggregates” I think you are saying that households do not make predictions or have expectations (other than expecting things to be about the same next year as this year).

The latter was the point I was trying to make. That is, economists often claim that a significant proportion of households DO TRY to make predictions about the “big” items: inflation, interest rates, levels of government spending, etc. I suspect that about 95% of households make no such predictions and have no “expectations” – other than the banal expectation that things will be about the same next year as this year.

[W. Peden]

Ralph Musgrave,

Let's say I'm a student in 2010, when a government comes to power in the UK with a five-year programme of cuts in government spending. Do I make the assumption that tuition fees are going to go up and arrange my affairs accordingly? In such a case, I might not be thinking in terms of "government spending" as such, but simply know that "deficit reduction" implies "higher fees".

Another example: imagine someone who is considering a career in the civil service. Does he take announcements of "deficit reduction" seriously and reconsider his career options, knowing that jobs, pay and pensions are all going to be hard to obtain?

Then take the example of those working for the government. Don't they work out the implications of the announcements of "deficit reduction" or "austerity" for their pay/job security and form expectations on these bases?

As for interest rates: when I talk money with people, I hear about little else. When will interest rates rise? Will they stay behind inflation? Basic financial questions facing any household that has any savings. My saving behaviour would not stay the same if the UK government suddenly announced that it was going to target inflation at 0% or 6%; currently, I have plenty of (mostly implicit) beliefs about what's going to happen to nominal and real interest rates and I behave accordingly.

Ditto inflation, though people systematically overestimate inflation. Any trade union boss would lose her job in a flash if she negotiated without using forward-looking inflation expecations.

So in the case of inflation and interest rates, expectations are usually based more on explicit predictions, whereas in the case of something a bit more rarefied like "government spending" one forms implicit predictions.

Again, while not wanting to concede everything to the Austrians and introspection, a bit of basic introspection and the reasonable assumption that most people are probably not mindless automatons reveals that people (a) form expectations of the future that are not simply backward-looking & based on assumption of uniformity of economic conditions and (b) take these expectations very seriously when making decisions.

[Nick Rowe]

DDJ: "Nick, only if you're NOT assuming Ricardian Equivalence, which no doubt neoclassical theory does."

Another wild assertion.

1. To believe that expected future income matters is not to believe REP.

2. Neoclassical economics dates from 1871. Robert Barro (re)introduced REP in 1974. It was controversial within neoclassical economics, and still is. Any neoclassical macroeconomist can rattle off half a dozen reasons why REP is unlikely to be exactly true, and teaches those reasons. Neoclassical macroeconomists disagree about whether or not it is a reasonable approximation under some circumstances. Many neoclassical models violate REP. Many neoclassical macroeconomists believe that the government debt is a burden, which automatically means they do not believe REP.

3. All neoclassical models assume (usually explicitly, but at a minimum implicitly), that Y=C+I+G+NX (except, in very rare cases, where the modeller has screwed up the addition). Like all neoclassical macroeconomists, I teach this to my first year students, for what it's worth.

4. Neoclassical economists, however, do murder kittens.

[Nick Rowe]

W Peden: "Isn't the income-expenditure relation only held to hold in a time period at a macroeconomic level? As a microeconomic fact, it is staggeringly obvious that income doesn't determine expenditure."

Yep. At a minimum you would need past and expected future income to help understand expenditure at the micro level. Rates of return, credit constraints, # of kids, etc. Plus, of course, preferences. The simple hydraulic "flow of income determines flow of expenditure" model clearly fails badly at the X-section micro level. (A lot depends on how we define "expenditure" also: is it consumption, or consumption+investment, or what?)

[Deus-DJ]

Nick,

heterodox economists step on snails to deal with their lifelong, unwinnable battle vs. orthodoxy (which I must admit, as much as you're the most open neoclassical to hearing heterodox critiques(too bad the heavy hitters don't care for posting online though!), you're almost the proudest neoclassical).

as for Ricardian equivalence, ok, but remember this is a post about ideology. IMHO ricardian equivalence is assumed no matter what your models say, and no matter what you say personally. Why would you otherwise make a distinction between expectations and when it actually occurs if its going to have the same effect (putting aside the question of whether expectations would have as much an effect)? sorry i just don't understand this part, thanks.

[reason]

Nick,
I don't actually disagree - but change the model a bit. Make automatic stabilizers bigger - so big that they have to be offset by monetary policy in normal circumstances. Then you don't need active fiscal policy at all.

[Nick Rowe]

DDJ: The heterodox can never win. As soon as they start to win, they become the new orthodoxy. Milton Friedman used to be heterodox, then he became orthodox, and now he's become a bit heterodox again.

reason: I find it hard to imagine how automatic stabilisers that powerful would work at the micro level. Marginal tax rates in excess of 100% would do it, but....In principle it might be possible, by indexing tax rates to macro variables, for example, but I just can't figure out what it would look like in practice. Fiscal policies have all sorts of other objectives that have nothing to do with getting AD right. Those other objectives might conflict with fiscal policy being used for getting AD right.

[W. Peden]

Nick Rowe,

"Fiscal policies have all sorts of other objectives that have nothing to do with getting AD right. Those other objectives might conflict with fiscal policy being used for getting AD right."

The history of the UK Treasury since about the 1920s seems to have been a matter of civil servants trying to convince politicians of this fact, as well as the similar point that fiscal policy has objectives other than maximising growth or reducing the natural rate of unemployment.

[Nick Rowe]

W Peden: Yep. And fiscal policy, no matter how powerful the government is, cannot be in two places at once.

[reason]

Nick,
if you add in automatic expenditure boosting and some positive feedback effects you could get there - or very close.

[reason]

And Nick - if you think that - then what do you think of those people who are claiming fiscal policy will be ineffective BECAUSE it will be offset by monetary policy. Aren't you contradicting their position?