Time-shifting planned demand vs. permanently increasing actual demand

Most policies aimed at increasing demand merely time-shift planned demand. Does that mean they don't really work? Does it mean we get recovery now at the expense of an even bigger recession later? Do we just have to cross our fingers and hope that something else eventually comes along to increase demand in future?

A cut in interest rates causes each individual to time-shift his planned demand for consumption goods. He plans to consume more now, and less later. The present value of his planned total demand is still constrained to equal the present value of his income. Does that mean that cutting interest rates can only be a temporary cure for deficient demand? Does that mean that if we cut interest rates today, to cure deficient demand today, we merely "borrow" demand from next year, and will have to cut interest rates even further next year, to "borrow" still more demand from the year after next, and so on? Does cutting interest rates to boost demand put us on a never-ending treadmill where we have to keep on cutting interest rates again and again?

Or, suppose we bring investment into the picture. If we cut interest rates today, and that increases planned investment today, that means firms will have a higher capital stock next year, creating an even bigger supply of goods next year, meaning we will have a bigger problem of excess supply next year. So doesn't cutting interest rates to cure an excess supply of goods today just postpone the problem until next year, putting us again on that treadmill of needing to cut interest rates again and again?

Or, suppose we bring fiscal policy into the picture. Governments have budget constraints too. If governments spend more today, it means they must spend less in future. Or tax more in future, which reduces households' future income and consumption spending. Even if we totally ignore Ricardian Equivalence, isn't expansionary fiscal policy just borrowing demand from the future? Isn't it too just putting us on a treadmill of needing to keep on running a bigger deficit and increasing debt forever?

OK, students of macroeconomics, why not? Can you answer that question, without peeking?

Here's my answer:

Suppose we initially start out with deficient demand today, and expected to continue on into the indefinite future, with income and planned expenditure staying the same. We want to increase demand permanently. So we cut interest rates, and shift planned demand from the future into the present. Will it work?

Yes. A policy that time-shifts planned demand can permanently increase actual demand. But it only works if the increase in planned current demand causes current income to rise, and this in turn causes people to revise upwards their beliefs about permanent income. It's that upward revision of people's beliefs about permanent income that permanently cures the recession. Because if people revise upwards their beliefs about permanent income, they won't merely time-shift their planned consumption; they will increase their planned current consumption and their planned future consumption.

Until we hit the supply-side constraints, aggregate demand is a widow's cruse. The more we spend, the more income we have to spend. It doesn't work at the individual level. But it can work at the aggregate level. One person's spending is another person's income.

Could it fail to work? Yes. If people saw the increased income today, and reasoned that it came from time-shifting demand from next year, and so revised their expectations of income for next year downwards instead of upwards, so their beliefs about permanent income don't change, then they actually will carry out their downward sloping paths for planned consumption, expecting a downward-sloping path of income.

If there's a surprise cut in interest rates, what happens next depends totally on how people revise their expectations about future income, both in response to the cut itself, and in response to any increase in current income created by an increase in current planned expenditure.

But what if the agents in the model have rational expectations? Would it be rational for people to revise upwards their expectations about future income?

Suppose initially that all the agents in the model are economically illiterate. They don't understand the model, and don't know what the parameters are. Suppose they expect that future income will equal current income. In effect, they think that income follows a random walk. If the central bank cuts the real rate of interest, those agents will be surprised when income turns out to be higher than they expected. They get more optimistic about their future incomes, and spend more, and get surprised again. They keep on getting surprised on the upside until the central bank decides to raise interest rates because demand is now high enough.

Now suppose that only 99.9% of the agents in the model are economically illiterate. The remaining 0.1% are omniscient macroeconomists, who know the model, all its parameters, and know they are a tiny minority. They know that when the central bank cuts the real interest rate, their incomes will rise and continue to rise for a number of periods. So they increase their planned current and future consumption by a large amount immediately, because they know their permanent incomes are much higher. But they are such a tiny minority their own consumption makes very little difference to what actually happens.

Now reduce the percentage of economic illiterates and increase the percentage of omniscient economists in the model. The omniscient economists are now numerous enough to affect the actual outcome. There's now a bigger jump in income immediately after the surprise announcement of the cut in interest rates. Which means the omniscient economists' expectations of permanent income rise by a bigger amount than when they were a tiny minority. And the economy gets to where the central bank wants it to be more quickly, and so raises interest rates again more quickly too.

Now take the limit of this model as the percentage of omniscient economists in the model approaches 100%. The central bank needs only cut interest rates for an infinitesimally short time, and the economy jumps to where the central bank wants it to go. The mere threat to cut interest rates, if the economy didn't jump to where the central bank wants it, would be enough to do the job.

I think that taking the limit like that is a sensible way to do stability analysis under rational expectations.

But my preference, especially where we are talking about novel circumstances where people will have difficult predicting the effects of policies, would be to stop about halfway. There's a distribution of information and knowledge across the population. Assuming that half know nothing and half know everything wouldn't be a bad first approximation. Then examine the process whereby the half that know nothing slowly learn the new equilibrium.

[What I've left out, of course, is true monetary policy. What happens if the central bank permanently increases the supply of money? What happens if the temporary increase in government spending is financed by a permanent expansion in the money supply, rather than by borrowing against future tax increases or spending cuts? No time-shifting there. A topic for another day.]

Comments

[Morganwarstler]

"It's that upward revision of people's beliefs about permanent income that permanently cures the recession."

This is true. That's why we need liquidation... fast and furious. Take the 7M foreclosed homes, and IMMEDIATELY auction them in $1 auctions to guys with dry capital:

1.this kills insolvent banks.
2. it creates upward revision in the minds of the people WHO HAVE MONEY TO SPEND. That's what's missing - they all know there are "deals of a lifetime" hanging out there. Stop letting the surviving banks buy up bad loans cheap of the banks being closed. INSTEAD, pump the assets directly into the the hands of millions of local rich guys.

Printing money doesn't increase demand, it just increases uncertainty.

We need to serve the DEMAND(S) of the money that's already out there - and those guys want to buy houses for pennies on the dollar.

[Scott Sumner]

Nick, Why would time shifted demand mean it "didn't work?" I thought that in natural rate models we assumed that demand management didn't affect the trend rate of real output, but rather the volatility of that output. If that's the goal, then you want to shift spending around. Indeed that's all you can do. There are periods we produce too much, and other periods when we produce too little. The goal is to smooth out production.

Keynes believed that we almost always produce too little, is his (non-natural rate) model the one you are assuming here?

Of course in nominal terms we can permanently increase spending quite easily, but I'm pretty sure that's not what you are getting at, as Zimbabwe proved how easy it is to do that.

I'm sure I am off base here---the only question is whether I am even on the right playing field.

[asp]

"A cut in interest rates causes each individual to time-shift his planned demand for consumption goods."

Nice theory, but many individuals, most businesses and some levels of government did not increase their consumption during the latest crisis.

[anon]

In its most basic form, the problem of managing "aggregate demand" arises due to firms with declining average costs driving a wedge between marginal cost and price.

A firm which truly set price equal to marginal cost would be indifferent to small variations in sales; but almost no business sells its products at price comparable to short-run marginal cost. Thus, cet. par. a firm which has a higher "demand" for its widgets will earn higher returns. An unexpected increase in aggregate demand likewise allows the firm to charge markups on a higher volume. (The firm's marginal costs will not become much higher until it employs its capital to the fullest extent and turns over its inventory as quickly as possible.)

However, in the long run, a business must not just decide on its scale of production, but set prices and plan investments in physical facilities and other kinds of capital. Firms with higher (resp. lower) expected demand will adjust their markups and the scale of their facilities to account for their revised expectations. (In the real world, there is no sharp distinction between "short run" and "long run"; but most capital investments are "sunk" and must be planned for long periods of time, and firms find it costly and inconvenient to constantly change prices in response to random variations in demand across sectors, businesses, etc.) Thus, I agree with Scott Sumner when he disputes that "a permanent increase in AD" is possible or desirable at all. "Aggregate demand" seems to be inherently a short-run concept.

[Roman Pearce]

"A cut in interest rates causes each individual to time-shift his planned demand for consumption goods."

Only if they are in debt. For savers the opposite is true. Please consider the following effect, it certainly applies my situation:

http://mpettis.com/2010/04/chinese-savings-and-the-wealth-effect/

For what it's worth, I'm 31 with a young family. Super low interest rates are a kick in the teeth. We built a nest egg to ensure financial security and stability, and that along with our income is being rapidly eroded. I don't own a house, those are now 10x my annual income and climbing at 2x per year. It's impossible to pay it off. I suppose the economists' solution is higher wages. Where? From what I can see, everyone who is/was in big debt is having a big party at my expense. I think super low interest rates are simply preventing the economy from restructuring after a rather large debt binge.

[rjs]

"A policy that time-shifts planned demand can permanently increase actual demand. But it only works if the increase in planned current demand causes current income to rise, and this in turn causes people to revise upwards their beliefs about permanent income."

looks like we're going in the wrong direction, then...about 35% of the jobs lost in the last two years paid below $15 an hour...& about 76% of the job growth this year has been in jobs that paid below $15 an hour..

http://www.nytimes.com/2010/09/01/us/01jobs.html
http://www.nytimes.com/imagepages/2010/08/31/us/20100901_JOBS_graphic.html?ref=us

[JKH]

This reads like some sort of Ricardian equivalence type proposition for economic growth in general – that you can only borrow incremental economic activity from the future, and that you can never see a boost in the current level of activity that doesn’t need to be reversed and “paid back” at some point in the future.

Economic growth in general is confidence driven, at least in part. Sensible people, consumers as well as business people, are aware of this.

If you consume X now and you borrow to consume X + Y, is your future planned consumption:

a) X + Y
b) X + 2Y
c) X
d) X – Y

I assume your base case is d), but I’m not sure.

I doubt that’s the standard belief of a consumer who borrows.

I think that in normal environments the typical borrowing consumer factors in at least some expectation of confidence-related income growth, sufficient to recover the amount and cost of borrowing, and that the answer is closer to c).

[anon]

"This reads like some sort of Ricardian equivalence type proposition for economic growth in general – that you can only borrow incremental economic activity from the future, and that you can never see a boost in the current level of activity that doesn’t need to be reversed and “paid back” at some point in the future."

It's just Say's law applied to consumption plus savings. The existence of savings today implies that these savings will be drawn down in the future; thus future consumption (hence future aggregate demand) will be higher. Economists often state that Say's Law is refuted, either due to the possibility of savings or due to the existence of AD externalities, but these critiques don't seem to apply to this version of Say's law.

[Nick Rowe]

Morgan: you don't really support your assertions. And this post is about the internal consistency of the New Keynesian macroeconomic model, and I don't think your assertions are true within that model. They may be true within some other model, but I don't know what that model is.

Scott: You are off base, but you are in the right playing field.

There are two Aggregate Demand problems. (Or, we can decompose the AD problem into two separate problems):

1. What ensures (or how can monetary and/or fiscal policy ensure), that undesirable fluctuations in AD don't happen (that AD only fluctuates to match fluctuations in LRAS, if there are any)?

2. What ensures (or how can monetary and/or fiscal policy ensure), that the average level of AD over time matches the average level of LRAS. What's to stop (or how can monetary or fiscal policy prevent) a permanent deficiency of AD?

Yes, standard NK macro assumes 1 is the only problem. It ignores the Old Keynesian (and pre-Keynesian, like Major Douglas, and that guy that Lenin ripped off on Imperialism, Hobshawn, whatever) fears about 2.

But it's exactly 2 I'm tackling here. What is the mechanism through which a NK model can either assume 2 away, or else specify how good monetary policy can prevent 2?

In other words, what I'm asking is: "What, if anything, would make the NK model a genuine Natural Rate model?" As far as I can see, NK's don't address this question; they sort of assume it away. Think of me (for this post) as an Old Keynesian looking with puzzlement at a New Keynesian model.

Now you don't see 2 as a problem. That's because you have M/P as an important variable in your model. Either P falls to increase M/P, or the central bank increases M to increase M/P, but either way you can always increase demand for goods. But M/P is not in the NK model. It's a model of monetary exchange, without money. It assumes away a problem (a permanent deficiency of demand) that it has no right to assume away.

[Nick Rowe]

asp: and a NK would explain this by saying that the central bank didn't drop the real rate fast enough when the natural rate fell. But I want to focus on the problem being that expected future income fell quickly too, because people thought that the recession might last (indefinitely). And (nearly all) NK's ignore this problem. They shouldn't, and can't.

anon: in LR equilibrium in a perfectly competitive model, all firms have P=MC. NK models assume monopolistic competition, so in LR equilibrium, P is greater than MR, and MR=MC. This means that the LR equilibrium level of output ("full employment" output) is lower in a NK model than under perfect competition. And it's lower than optimal. In one sense, AD is always too low in an NK model. But there's nothing that AD policy can do to fix this problem permanently in a NK model. It's a recognised problem, and I ignore it here. The problem I'm focussing on is whether AD could still be permanently below the (admittedly low) level of LR equilibrium output in a NK model, and whether and how monetary policy could get the economy to the monopolistically competitive LR equilibrium.

By the way, the underlying problem with monopolistic competition is not that firms' AC curves slope down. The problem is that firms' demand curves slope down, so MR is less that Price. It is true that with free entry and exit of firms, the LR equilibrium in micro means that downward-sloping demand curves mean the AC curve must also slope down locally (because the demand curve is tangent to the AC curve). But "Long Run" in macro means something very different from "Long Run" in micro. In micro, "LR" means "no firm has any incentive to exit or enter". In macro, it means "no firm has any incentive to change relative price or output". It's a very different definition of LR. I'm talking macro here. Hold the number of firms exogenously fixed, if you like.

[Nick Rowe]

Roman: a cut in interest rates causes a positive wealth effect for debtors, and a negative wealth effect for creditors. The canonical NK model assumes a representative agent for simplicity, who is therefore neither a creditor nor a debtor. This is equivalent to assuming the two wealth effects cancel out in aggregate. Now maybe they don't cancel (it would be a fluke if they did, but it could go either way in principle). But I want to explore the internal consistency of the NK model, and these wealth effects aren't in that model.

rjs: hourly wage rates aren't the same as income. It's GDP (or GNP, or NNP) you need to look at. That's (aggregate) income from the production and sale of newly-produced goods and services. That's what fell. That's what mattered.

JKH: There is some sort of analogy here to Ricardian Equivalence, but you can't push it too far. In Ricardian Equivalence, a cut in taxes today doesn't even shift demand to the present, because people anticipate the future higher taxes, and so recognise their permanent disposable income is unchanged, and so don't change their consumption plans at all.

Yes, when the real interest rate is cut, people change their plans according to your d. But when that change in current plans causes them to be surprised upwards about their current income, and they revise their expectations of future income upwards too, the aggregate effect ends up more like your b (except it might be a much bigger number than 2).

anon: yes, this is all related to Say's Law. But I'm going to stay off Say's Law here, because I've already done posts on it, and it's too complicated for me to explore that relation here.

[Scott Sumner]

Nick, Isn't Bennett McCallum someone who sort of shares your belief in the importance of money, and also buys into NK models? I recall reading that in his mind there is no contradiction, as M is in the model somewhere, even if only implicitly. But it was years ago when I read McCallum, so I may be wrong. Nevertheless, I wonder if you've looked at his views on these issues.

Here's my other question about your response. It seems like in a sense you are criticizing the NK model for having no nominal anchor. Is that right? If there were a nominal anchor (like M) then wage and price flexibility (which NK models have) would be enough to ensure a natural rate equilibrium in the long run.

[anon]

"Now you don't see 2 as a problem. That's because you have M/P as an important variable in your model. Either P falls to increase M/P, or the central bank increases M to increase M/P, but either way you can always increase demand for goods. But M/P is not in the NK model."

Even NK models without money have something comparable to the "general price level". Indeed, in some NK models (Dixit-Stiglitz), the macroeconomic externality _is_ the effect of price variations in the monopolistically-competitive sector on consumers' income constraints, e.g. _totidem verbis_ on the general price level.

[David Pearson]

The Fed engaged mortgage finance as the primary monetary transmission mechanism in 2003. During the resulting housing boom, consumers raised their expectations of future income as a result of Morgtage Equity Withdrawal and other capital gains. As a result, they borrowed more to finance current income and AD rose. When presented with evidence--in the form of flat or declining home prices--that their future income expectations were incorrect, they realized that current consumption was too high. To support the debt remaining from their consumption binge, they would have to spend less to service debt and rebuild savings. The result was a steep fall in AD.

So we have an example where Fed policy more or less directly targeted mortgage finance; the mortgage boom raised expectations of future income; this resulted in an increase in AD; households saw that incomes would not follow the expected trajectory; their realization that the marginal debt taken on was unserviceable then led to de-levering and a fall in AD.

So I would say the evidence supports the thesis that if the Fed impacts future income expectations, this would not necessarily lead to a permanent increase in AD. Unless you are willing to call the four year increase in AD from 2003-2007 "permanent".

[Nick Rowe]

Scott: "Here's my other question about your response. It seems like in a sense you are criticizing the NK model for having no nominal anchor. Is that right? If there were a nominal anchor (like M) then wage and price flexibility (which NK models have) would be enough to ensure a natural rate equilibrium in the long run."

Yes, and yes. Most macro models (the ISLM for example) have a downward-sloping AD curve in {P,Y} space. So, given some price flexibility in the long run, they usually (unless something goes wrong) have some sort of tendency to get back to the natural rate eventually. That downward-sloping AD curve creates both a nominal anchor, and eventually brings you to the natural rate.

The sort of NK models I am talking about (I should have been more explicit, because I'm talking about the later Neo-Wicksellian version, where the central bank sets a nominal interest rate) have a vertical AD curve, for a given real rate of interest. A lower P does nothing to increase AD. Actually, that's not quite right. The AD curve is very very thick. It covers the whole of {P,Y} space. If you set the real interest rate at the natural rate, then any combination of {P.Y} is a point on the AD curve. Because the Euler equation tells you that the real interest rate must be some function of the *ratio* of current to expected future demand.

anon: interesting comment. If you have a model with a downward-sloping AD curve, and monopolistically competitive firms (as earlier NK models did, before they replaced the old-keynesian IS curve with the modern Euler Equation version), then you can indeed talk about the "macroeconomic externality" of monop comp in this way. And that's how it was talked about.

But once you drop M and P from the model, you need some other way to define and think about the macroeconomic externality. And you can. Start in LR equilibrium, where each firm has MR=MC. If one firm tried to lower its relative price and expand sales, that would have a second order of small effect on reducing profits for that firm, but would have a first order effect increasing the real profits and real wages at all other firms. And if all firms tried to do this, there would be downward pressure on inflation, so the central bank could increase AD, and real income would rise, and everyone would be better off.

But actually, I don't think this is qualitatively any different from the microeconomic externality of monopolistic competition. An individual firm that cut its price a small amount would cause a second order loss in profits, but a first order increase in consumer surplus.

But your general point still stands. These Neo-Wicksellian New Keynesian models are monetary models without money. The general price level and money, are there and not there, at the same time. It's implicit.

Are you the same "anon" that's been posting the past few days? You know stuff.

David: Did the Fed really target mortgage interest rates in particular, or all interest rates in general?

You present everything from the borrower's perspective. Remember the lenders. The demand of both borrowers and lenders makes up aggregate demand. And if people had overly-optimistic expectations about future income, this would make AD bigger than it would otherwise be, and mean the Fed would have to set the real interest rate higher than it otherwise would have done.

Was the fall in incomes after 2008 due to some exogenous fall in Aggregate Supply? Or was it caused by a fall in AD, due to a de facto tightening of monetary policy? I would say the latter. It wasn't that people had overly optimistic expectations of future income (though some may have had); it was that the Fed let AD fall after 2008. (Scott has been influential in my thinking here).

[anon]

"interesting comment. ... But actually, I don't think this is qualitatively any different from the microeconomic externality of monopolistic competition."

The externality _is_ "microeconomic" in the sense that (AFAICT from Dixit-Stiglitz and similar models) it falls out of standard general equilibrium analysis plus monopolistic competition (The additional assumption of price rigidity is apparently needed in order to analyze the short-run response to shocks in AD).

The usual microeconomic analysis of monopolistic competition is a partial-equilibrium model of a single product market. This analysis disregards income effects and the consequences of such effects on other monopolistically competitive markets--thus, I'm not sure that it can qualitatively account for the macroeconomic concerns you discuss above.

[Nick Rowe]

anon: start in LR equilibrium, where each firm has MR=MC. Now suppose we increase output and cut prices at half the firms, holding output and prices constant at the other half (and increase AD to match the increase in aggregate output. The half the firms that hold their outputs constant now see MR rise, in real terms. That's the income effect. But they also see MC rise, in real terms, since real wages get pushed up when half the firms expand output and employment. Depending on whether MC or MR increases more, the other half the firms would either want to lower or increase output. In other words, they could be "strategic complements" or "strategic substitutes". The income effect creates strategic complementarity. But the scarcity of resources and rise in real wages creates strategic substitutability. So the macroeconomic externality could be either bigger or smaller than the microeconomic externality (which ignores those positive or negative feedback effects).

Which is why I've never put too much weight on the "macroeconomic externality" concept.

[anon]

"But they also see MC rise, in real terms, since real wages get pushed up when half the firms expand output and employment."

ISTM that this would be a pecuniary externality--a pure transfer from firms to workers and resource owners--which is perhaps qualitatively different from the real effect of firms expanding output. I'm not sure, but it seems that "strategic substitutability" would be somewhat unlikely.

[Nick Rowe]

anon: "strategic substitutibility" is quite possible. Suppose labour is the only variable input, and suppose that the aggregate labour supply curve is very inelastic, so W/P rises a lot when half the firms increase output and employment. In the limit, with a vertical labour supply curve, the long run equilibrium output is identical to a perfectly competitive economy. One firm expanding output means other firms must contract output.

Profit-maximisation implies MRxMPL=W. Since MR=P(1-1/e), where e is elasticity, we can re-write this as: P(1-1/e)xMPL=W

In macroeconomic equilibrium, with symmetric firms, all firms will set the same price, and we can divide both sides by P to get (1-1/e)xMPL=W/P.
If e is a constant, if MPL is a function of L, and W/P is also a function of L, through the labour supply curve, we can solve this equation for L, and thus for Y(L). The only difference between monopolistic and perfect competition is that e is infinite under perfect.

[anon]

Nick Rowe: Isn't this a model with a small or nonexistent Keynesian output gap, though? Why would this be a reason to disregard the "macroeconomic externality" concept?

On a separate note, in a model where labor is entirely inelastic, the wage is effectively an economic rent which doesn't enter in the cost of production: as long as the labor market clears and the MPL is positive, every laborer is employed in his highest and best use. This probably explains why your model has the same output as the perfectly competitive economy.

[Nick Rowe]

anon: good point. It does however illustrate that the macroeconomic externality could be either smaller (as in this case) or bigger than the microeconomic externality of imperfect competition.

Separate note: agreed. Except that labour could be supplied perfectly elastically to the individual firm, but perfectly inelastically to the economy as a whole. So if different firms had different elasticities of demand, we could still get the wrong mix of employment, even if aggregate employment were at the right level.

[Nick Rowe]

except, on your first point, I wouldn't call the gap between monop comp LR equilibrium and perf comp LR equilibrium an "output gap". I would define "output gap" as the gap between monop comp LR equilibrium and actual output.