Monday, March 2, 2009

We just aren't that good (as if that needed saying).

From Worthwhile Canadian Initiative:

Might fiscal policy fail to increase aggregate demand?

Yes, it might fail.

Economists ought not be confident enough in our knowledge of how the economy works to say they are certain that a temporary increase in government spending will definitely increase aggregate demand. We just aren't that good (as if that needed saying).

Here is one reason why it might fail. I think it is the least implausible reason why fiscal policy might fail.

A temporary increase in government spending, financed by borrowing, will increase the future level of the government debt, and increase expected future taxes.

Taxes are typically not lump-sum. Most taxes depend on income; the more income you earn, the more tax you pay. Much of our income comes from investment, in physical and human capital. Higher expected future tax rates will reduce the expected future return to current investment, and will reduce current demand for investment.

(This effect on investment is in addition to any effect that future taxes might have on permanent disposable income and current consumption, which by itself should not fully offset the increased government spending.)

The magnitude of this effect is an empirical question. It depends on the elasticity of investment with respect to future returns, and on the exact nature of the future tax increases that people expect. If investment demand were elastic enough, and the expected future taxes distorting enough, it would be possible for a temporary increase in government spending to cause investment to fall by more than government spending increased, so that aggregate demand would fall. An increase in government spending would cause the IS curve to shift left.

I do not know the answer to that empirical question. Nor does anyone else know the answer to that question with certainty. We can only rely on past experience, hope that we have interpreted past experience correctly, and hope that the lessons of the past apply to the present.

This might or might not be what William Poole was talking about. It is more likely that he was talking about something like this than talking about a vertical LM curve. He did mention higher future taxes and tax incentives for investment; he did not mention interest rates.

I draw three lessons:

1. We need to be careful how we interpret people; especially those with whom we disagree. We might learn more if we apply the Principle of Charity.

2. Our preferred policies might fail for reasons we might not have thought of. Critics might think of something we haven't thought of. "Who could have known?" has been heard too often recently.

3. It would be prudent to try to design fiscal policies to minimise the chances that they might fail for all reasons, just in case the critics are partly right. For example, government spending on investments that increase future income would be especially desirable in the light of this critique, because they would be less likely to require future increases in tax rates.



Doesn't Buiter's proposal work without borrowing?

Leigh and Nick,

Is Ricardian Equivalence a tautology, akin to "Robbing Peter to pay Paul"? If you take, it must be from someone, etc. After all, it sounds as if no one has empirical evidence for this view one way or the other. I can see that it would have some theoretical use, since it focuses on the correlation of a few terms, but it seems dubious empirically. In fact, I'm not sure that it can be shown empirically at all, since it assumes a certain view of human behavior in order to be valid. If that view of human behavior is invalid, then, at best, this is a tautology, or simply, a model construct, useful heuristically, if not empirically. I go on like this when anyone mentions Moore and Wittgenstein. Please excuse me.

Posted by: Don the libertarian Democrat |


"OK. Let's see if I can answer all this at once.

In very simple Keynesian models, a $100 increase in Government spending will cause aggregate demand to rise by more than $100. For example, if an extra $1 of income causes private demand to rise by %0.50 (a marginal propensity to spend of 0.5), then a $100 increase in G will cause AD to rise by $200. Hence the name "multiplier", because the ultimate effects are a "multiple" of the original increase. In this example the multiplier is 2.

But language changed over time, and we began to use the word "multiplier" as a sorthand for "the derivative of AD with respect to G". So we could talk about a multiplier of 2, as in the above example, but we could also talk about a multiplier of 1 (if a $100 increase in G caused AD to rise by the same $100), or even (oxymoronically, but that never stopped economists) a multiplier of 0.5 (if a $100 increase in G caused private spending to fall, so the ultimate effect on AD was a rise of only $50).

And sometimes we define multiplier as the effect not on AD, but on real income, Y. (Whether an increase in AD causes an equal increase in Y depends on the slope of the Aggregate Supply curve).

Strict Ricardian Equivalence says that an increase in G financed by bonds is equivalent to an increase in G financed by current taxes. A logical corollary is that a cut in taxes financed by bonds will have zero effect.

Ricardian Equivalence thus says that the tax cut multipler is zero. And that the (bond-financed) government expenditure multiplier is equal to the "balanced budget" (tax-financed) government expenditure multipliers. In very simple Keynesian models, the balanced budget multiplier is 1. So if we add Ricardian Equivalence to a very simple Keynesian model we get a bond-financed government expenditure multiplier of 1 as well. In more complicated keynesian models (add imports, or an effect of income on interest rates) and the multiplier gets smaller still, but still positive for an increase in government expenditure, even under Ricardian Equivalence.

Ricardian Equivalence is not a tautology. It is almost certainly false (or at least, not exactly true). We can think of good theoretical reasons why it will not be exactly true. It is very hard to test Ricardian Equivalence in isolation. We can only test it in combination with other hypotheses.

I am not up to date with empirical tests of Ricardian Equivalence. I can remember one test by Greg Mankiw, many years ago, where he tested the combined hypothesis of permanent income theory+rational expectations against the current income theory. (Ricardian Equivalence assumes permanent income+rational expectations, while the simplest Kenyesian model assumes the current income theory of consumption.) He found that the facts seemed to be roughly halfway between the two theories. That seemed plausible to me, and even though it was not a direct test of Ricardian Equivalence, I tend to think of Ricardian Equivalence as being about half true, unless someone convinces me otherwise.

I think of it this way: "Ricardian effect" will tend to reduce the size of multipliers, but only full Ricardian Equivalence can reduce a multiplier to zero, and then only the tax-cut multiplier, not the government spending multiplier.

The theory I sketched above, of a negative multiplier, is very different from Ricardian Equivalence (though Ricardian Equivalence would make it easier for my effects to get a negative multiplier). Ricardian Equivalence is about the effect of future levels of taxation on permanent disposable income and hence on current consumption. I am talking about the effect of future marginal tax rates (not the same as tax revenue) on current investment. Ricardian Equivalence is about wealth effects on consumption. I am talking about incentive effects on investment.

Leigh: if you take a standard simple Keynesian model, you will never get the result that increases in G are self-financing. Or rather, you would only get it with a mpc>1, which makes the equilibrium unstable, if it exists.

But it is possible to take a fairly standard ISLM, plus liquidity trap, plus Phillips Curve, plus adaptive expectations, plus a Taylor-rule type monetary policy, and get a temporary increase in G to be self-financing. I sketched it in a post a month or two back. (Damn, but I can't remember the post title). The trick is that the above model has two equilibria, each locally stable. And you can use a temporary increase in G to jump you from the low equilibrium to the high equilibrium. But nobody paid any attention to my radically exciting post, boo hoo!

May respond to other points later.

"Doesn't Buiter's proposal work without borrowing?"

You mean helicopter money? Yes, it works without borrowing. Just print money and give it to people as a transfer, or tax cut.

OK, the government "borrows" the money from the central bank, and gives it bonds in return. But since the government owns the central bank, it's a wash.

The tricky thing is: what if you print a lot of money now, to get the ball rolling, but once it does start rolling you need to reduce the quantity of money, because you realise you've overdone it, and it's starting to cause hyperinflation? The the Central bank needs to buy it back, by selling the bonds it got from the government. And now the government's debt really does go up.

That's what Buiter of the blog, as opposed to Buiter of that paper, was so concerned about.

I know I'm alone in this, but I think that's a more solvable problem than Debt-Deflation, which is a kind of economic vertigo. But if Buiter doesn't even recommend it, I guess it won't be going anywhere.

Since I'm for:
1) Buiter's QE
2) The Swedish Plan
3) Narrow Banking
4) A sales tax cut as stimulus
I'm not doing very well in this crisis.

No comments: