Aggregate supply and relative supply -- and demand.

Micro and macro are different. We have to be careful about drawing macro conclusions from what happens at the micro level. We can't just extrapolate. This is in response to Casey Mulligan's post on evidence that supply matters. He draws macro conclusions about aggregate supply from micro evidence about relative supply.

Draw an Aggregate Demand curve with the Price level on the vertical axis, and real output on the horizontal. Suppose either that the AD curve is vertical, or that the aggregate price level is fixed (or just very slow to adjust). Or both, if you like. And suppose there's excess supply of output.

The short-side rule says that actual sales are determined by whichever is less: quantity demanded; or quantity supplied. In this case the short side of the market is the demand side. Aggregate output is determined by aggregate demand. Aggregate supply does not affect aggregate output.

That says nothing about the determination of relative output of various sub-sectors of the economy. A change in relative demand may change relative prices and relative output and employment across different sectors of the economy. A change in relative supply may also change relative prices and relative output and employment across different sectors of the economy.

Imagine there's a fixed amount of tradeable permits to produce pollution. That creates a perfectly inelastic demand for pollution. Shifts in the supply curve have no effect on the total quantity of pollution. But because the permits are tradeable, shifts in relative supply and demand will change the relative amounts of pollution produced in different sectors. If people want to buy, or sell, more Eastern goods, we get more pollution in the East, and less in the West, but the total amount stays the same. We cannot extrapolate from micro evidence to macro conclusions. What is true for each of the parts is not true of the whole.

The keynesian/monetarist vision of demand deficiency is just like that. Except we are talking about GDP, not pollution. And the tradeable permits are called "money".

Casey Mulligan has micro evidence that increases in supply in particular sectors cause increases in output and employment in that sector. He uses that micro evidence to draw macro conclusions about the effects of increases in aggregate supply. Those conclusions don't follow.

Of course, we could say that same thing about people who draw conclusions about macro fiscal multipliers from estimates of micro fiscal multipliers. They are mistaking shifts in relative demand for shifts in aggregate demand.

Comments

[Frances Woolley]

Nick, could you check that link to Casey Mulligan's post, please? I couldn't get it to work.

[Nick Rowe]

Thanks Frances. Fixed.

[Frances Woolley]

"Imagine there's a fixed amount of tradeable permits to produce pollution. That creates a perfectly inelastic demand for pollution."

Wouldn't that be a perfectly inelastic supply of pollution? I think the the demand for pollution coming from firms who want to create pollution in order to produce goods and services. Which would vary, depending upon the cost of buying a permit.

But your point - that changes in relative prices/sectoral output don't imply that there are changes in economy-wide output levels - still holds. E.g. a technological change like introduction of hybrid vehicles could reduce the demand for permits to produce carbon dioxide, and increase the demand for permits to produce whatever toxic stuff is associated with production of batteries.

[Nick Rowe]

Frances: I always get muddled. There's two curves that cross. So we call one a "demand" curve and the other a "supply" curve. It's like the demand and supply of crime. Are the criminals the demanders or the suppliers? Depends what we put on the vertical axis: the rewards to crime or the penalty for crime. Because we want the demand curve to slope down and the supply curve to slope up. But yes, it doesn't really matter except for trying to keep our heads straight with the analogy to regular markets.

[Andy Harless]

"Suppose either that the AD curve is vertical, or that the aggregate price level is fixed (or just very slow to adjust)."

I can suppose that, but an argument based on a supposition doesn't get us very far. Why would we expect these conditions to be true? One reason we might expect them to be true is that individual prices are slow to adjust. In its simplest version, this explanation would seem to contradict the observation that supply affects output at the micro level.

Of course the actual explanation is more complicated. You have to talk about how price stickiness in one market affects the behavior of suppliers in other markets and so on. And once you do the hard modeling, you may come up with a model in which supply matters a lot more at the micro level than it does at the macro level. But that's not intuitively obvious. I don't think you can deal with it effectively by making arguments that take for granted that there are such things as "aggregate supply" and "aggregate demand." Once one acknowledges that these are meaningful concepts, one has already stacked the deck in favor of the Keynesians.

The problem with the pollution analogy, I think, is that you have explicitly constructed it so that the price of (real) pollution is fixed. Pollution permits bear a fixed relation to the quantity of real pollution that they allow, because they're defined in real terms. Money, on the other hand, does not bear any fixed relation to the quantity of real GDP. If you want to argue that it is reasonable to assume a fixed relation between real and nominal GDP, then you have to justify that relation, or else you allow Casey Mulligan to assume a particular justification, and he can reasonably claim that, given his assumed justification, supply should not matter at the micro level.

[Nick Rowe]

Andy: good points.

In the simplest Keynesian model (say ISLM with exogenously fixed expected inflation, no debt-deflation or Pigou effects etc.) the AD curve is vertical in a liquidity trap. Relative demand curves still slope down as a function of relative prices, of course.

Assume the speed of price adjustment in any particular market is a function of excess supply in that market, given expected inflation. Fairly standard assumption. Then an increase in supply in one particular market would cause relative prices to fall in that market, and a movement along the demand curve in that market.

And remember where this is coming from: Casey Mulligan is saying these examples contradict the Keynesian theory. So, to check that claim, we do have to assume Keynesian theory is true, and then see if these examples contradict the predictions of that theory. Sure, it doesn't mean the theory is true. Just not proven false.

Yep, money allows a certain quantity of nominal GDP, given V. Pollution permits allow a certain quantity of real pollution. You need sticky prices to make the analogy work. But that's OK.

[Arin Dube]

Hi Nick,

Thanks for re-iterating this point. I made an analogous argument about why employment increases of certain groups cannot help us understand whether there is quantity rationing in the labor market in my guest post here: http://economistsview.typepad.com/economistsview/2011/07/got-jobs.html

Arin Dube

[Nick Rowe]

Arin: thanks. I just re-read your post. We are on the same page. BTW, does this show the big difference between Chicago economics department vs Chicago Biz School? Some people lump all of "Chicago" together.

[Too Much Fed]

Nick's post said: "Yep, money allows a certain quantity of nominal GDP, given V."

I've seen other economists say central banks should target NGDP and forget about real GDP (RGDP). Do you agree with that?

[Nick Rowe]

TMF: in the 1960's, Keynesian economists used to want fiscal and/or monetary policy to target real GDP and/or unemployment. We learned, both empirically and theoretically, that that was a big mistake. It can't be done, in the long run. No economist wants to target real GDP now. (OK, I expect there are exceptions, somewhere).

Target NGDP? Dunno. Maybe.

[Andy Harless]

It's not clear what one would mean by having a central bank "forget about real GDP." I would argue that inflation targeting -- at least as it's commonly practiced today -- is impossible if you forget about RGDP, because it involves making an estimate of potential RGDP. My view is that central banks should wash their hands of real variables entirely by conducting policy in a way that ignores the supply side. Basically, that means targeting NGDP or something similar. In terms of the equation of exchange, the central bank's job should be to adjust M so as to offset changes in V. Why make its job harder (and more subject to error) by requiring it also to offset changes in Y?

[Ralph Musgrave]

I agree with Nick: Mulligan is talking nonsense.

[Nick Rowe]

Andy: I generally approach this from the classic "target, instrument, indicator" distinction. For example, with the Bank of Canada:

Target is 2 year ahead forecast of CPI inflation.

Instrument is the overnight rate.

Indicator is everything it thinks might be relevant in making a conditional forecast of 2 year ahead CPI inflation, including current headline and core inflation, Y, u, the exchange rate, etc.

If there were no lags in the monetary policy framework, this 3-way distinction might not make sense. For example, under the classic gold standard with 100% reserves, where all the bank does is buy and sell unlimited amounts of gold at a fixed price, the price of gold is target, instrument, and indicator.

[Too Much Fed]

I was hoping someone would get where my question was heading. I'll try it this way.

If productivity growth is +2.5% or higher and RGDP is +2% or lower, what most likely happens to employment?

[Sergei]

Nick: the tradeable permits are called "money"

Why is that money? Does it imply that all "money" however you define it gets spent? However money, however you define it, as a stock. But spending is a flow. Tradable permits should be called income and not money. Money alone is irrelevant. Given that noone can define V and money * V is defined as income then just drop money and focus on income. At least you will not staring into the black-box called V. Oder?

[Nick Rowe]

Sergei: You lost me on "Oder". The river Oder? German for "or"? Maybe a typo?

"GDP permits = money" is a metaphor. Like all metaphors, it doesn't work exactly. But it works pretty well, for something that came off the top of my head a couple of days back.

The metaphor works because, whether we are talking about GDP permits or money, you can't sell stuff unless you get some in exchange. You can't buy stuff unless you have some and give it to the seller.

(And the metaphor fails because money is also used to buy stuff, like used furniture or bonds, which is not part of GDP, so you wouldn't need a GDP permit to buy.)

Yes, there's a stock/flow distinction. I discuss this in my previous post, to which this post links.

But we can't just start with income. Income is what we seek to explain. And in a monetary exchange economy, and where there's an excess supply of output, income depends on how quickly people are willing to circulate the stock of money, and how big a flow you get from that stock. If desired V=0, so nobody wants to spend their money, income is zero.

In a barter economy it would be very different. Even though Y=C+I+G and S-I=G-T would still be true in a barter economy, MV=PY would obviously make no sense at all. But we don't live in a barter economy.

We sometimes do need to stare into the black box called V.

[Sergei]

Nick: Income is what we seek to explain

I disagree. We seek to explain spending and not income. If there is no income, then there is no spending. However, even if there is income, there is no guarantee of spending. And again V does not matter because it is not V which defines how much spending is done from income.

[Nick Rowe]

Sergei: You are begging the question when you say that V does not define how much spending is done from income. Other things equal, don't you think a person who is holding less money than they desire will want to spend less than they otherwise would, for a given level of income? Because that's a case where desired V is less than actual V. You are going to have fun with my next post!