Scott Sumner  

AD: Eleven stages of enlightenment

Note on Borjas: Non-Binding Co... A(nother) Colbertian in Brusse...

Over at TheMoneyIllusion I've been having a discussion of aggregate demand. What does the term actually mean? In the comment section I see lots of average people giving common sense explanations, and also experts like Nick Rowe making high-level arguments. It might help if I walked people through the various levels of enlightenment:

Level 1: If people decide to spend less, then aggregate demand will fall. By "spend less" the speaker usually means save a larger fraction of their income.

Level 2: But saving equals investment, so if society saves more it will invest more. You are "spending more" on investment projects.

Level 3: But planned savings may exceed planned investment.

Level 4: In that case interest rates will fall, equalizing actual saving and investment.

Level 5: But perhaps it's income that falls, and that is what equalizes actual saving and actual investment.

Level 6: But why should income fall? How does an attempt to save more reduce M*V?

Level 7: If you try to save more and as a result interest rates fall, then velocity will tend to decline, as there is a lower opportunity cost of holding onto cash.

Level 8: Yes, velocity would fall, but I asked why M*V would fall. If the central bank uses the Taylor Rule, or if it targets inflation or NGDP, it will adjust the money supply to prevent M*V from falling.

Level 9: That may be true if interest rates are positive, but at the zero bound they cannot offset the fall in V.

Level 10: Sure they can. There is nothing special about the zero bound in interest rates. All that matters is the zero bound in eligible assets left to buy. And no central bank has ever come close to that zero bound. No central bank has ever said they were out of ammunition.

Level 11: But central banks are reluctant to do unconventional monetary policies at the zero bound, and hence if people try to save more, then M will not fully offset V.

I could go on and on, but I decided to let my opponents have the last word. After all, this "enlightenment" stuff is kind of arrogant. Now let's consider an extra 100 million immigrants suddenly flooding into the US. Does this raise AS or AD?

Level 1: Clearly AD rises, there are lots more shoppers!

Level 2: No, that's an AS shock, there are lots more workers.

Level 3: It depends on the monetary policy. Under the gold standard M is unaffected. So it's a positive supply shock and the wave of immigration causes deflation. Remember 1865-96?

Level 4: No, the wave of immigrants raises the return to capital. Interest rates rise and velocity increases. Both AS and AD rise.

Level 5: But we are probably not at the zero bound if 100 million immigrants pour in, and we now have fiat money, so it entirely depends on whether the central bank is targeting inflation or NGDP.

etc., etc.

The textbook AD curve generally assumes a constant money supply. Thus it might be viewed as a gold standard model. That doesn't mean that it can't be applied to fiat regimes, but you need to keep in mind that "supply shocks" could cause central bank reactions that lead to a simultaneous shift in the AD curve.

It's also important to be clear with one's language. If the central bank is targeting NGDP and 100 million people flood into the US then it is clearly a supply shock, AD doesn't change. But the aggregate quantity of goods purchased (demanded) will soar much higher, with the population growth. The stores will be overflowing with shoppers. Lots more aggregate quantity demanded, but no increase in aggregate demand. When you have a mental image of lots of shoppers, don't think you are visualizing AD. You are visualizing equilibrium quantity, which can reflect AD or AS shocks. The same is true of individual markets. The big surge in shoppers purchasing PCs in the 1990s was mostly a supply story, as evidenced by the falling prices.

You may think I'm being picky about language, but it really is important to keep these distinctions clear.

The other takeaway is that common sense notions about "spending" are a very misleading way of thinking about AD. After all, AD isn't just consumption, it's also I + G + NX. Your initial common sense view may be right in the end, as we saw in the 11 stages of enlightenment, but it will be right for the same reason that a broken clock gives the correct time twice a day.

Tyler Cowen says:

I increasingly think it is a mistake to draw too sharp a distinction between aggregate demand and aggregate supply, at least beyond the very first round of an economic shock.
I sympathize, as the shocks are often "entangled." I'd like to dispense with all discussion of AS and AD, and replace it with nominal shocks and real shocks. A nominal shock is an unexpected change in NGDP. A real shock changes the price/output split for any given level of NGDP. As Tyler suggests, one type of shock is often entangled with the other. But it's still important to keep them clear as a theoretical matter, so that we can think clearly about how monetary policy should respond (or not respond) to various types of situations.

Comments and Sharing

COMMENTS (14 to date)
Patrick R. Sullivan writes:

[Comment removed for only tangential relevance.--Econlib Ed.]

Nick Rowe writes:

This is nice and clear Scott. I (almost) agree with everything you said.

For the immigration case, Level 3: under the gold standard, M would presumably increase. For example, British immigrants swap their pounds for gold, and their gold for dollars. Even if they arrive penniless the rise in interest rates and increased capital inflows and net exports should result in gold inflows and an expansion in M in the US.

Picky: first set of levels, levels 4 and 5: you meant to say that it is the *planned* saving and *planned* investment that are equalised by falls in interest rates or falls in income. We don't need anything to change to keep actual saving and investment equal, since they are the same thing, by definition (including government and foreign saving and investment, of course). (Keynes made the same mistake, but it wasn't a typo in his case.)

Yancey Ward writes:

How does one assign causation in a nominal shock? Does it matter to you?

Scott Sumner writes:

Nick, Good point about the gold standard and gold inflows. But of course we could imagine an alternative scenario like a global baby boom that actually would hold M fixed, and get the qualitative result I was looking for. Or, of course, an NGDP targeting central bank.

I see your point about planned investment, althought I'd say it's partly semantic. One could say that it is the fall in interest rates that allows actual S and I to equilibrate despite the INITIAL plans being different. It wasn't a typo on my part, but I did have exactly the same interpretation as you suggested. But I see your point, when the interest rate falls the UPDATED planned S and I are also equilibrated.

Yancey, In the modern world the central bank controls nominal shocks. In the gold standard world it is more complicated.

ThomasH writes:

I don't know where to fit it into your 12 step program :) but I's say it's AD when people want to change the relation between income and savings by saving more (eg because they hear about a financial crisis.) It's AS when people want to change the relation of savings and income by earning more (immigrants come to get jobs) Whether they are able to do what the want to do in the aggregate depends on the monetary authorities, right?

Yancey Ward writes:

Scott, you seem to imply it is the nominal shock that can cause the real shock in some (most?) cases, rather than the other way around. Maybe the central bank can address the nominal side regardless of cause, but then why does the distinction even matter then?

Don Geddis writes:

@Yancey: Maybe concrete examples would help?

Nominal shock: the public suddenly wishes to hold more currency (velocity falls). If the central bank accommodates the additional currency desire by expanding the money supply, then spending (NGDP) remains stable, and there are no real effects. (If the central bank fails to do so, like in 2008, then the nominal shock will have negative real effects, lowering both output and employment.)

Real shock: Japanese tsunami, or 1973 oil embargo. Real output will be lower, as the world has lost some productive capability. This real shock will cause spending to fall as well. If the central bank does nothing, then the drop in NGDP will cause a secondary real effect of even more recession. If, instead, the central bank boosts the money supply to maintain NGDP, then the economy only suffers the original real damage from the original real shock, but not any additional secondary damage from the imposed (but then mitigated) nominal shock.

With a real shock, the central bank can prevent contagion into the rest of the economy. (But the directly affected sectors, will suffer a decline in output.) With a purely nominal shock, the central bank can essentially prevent any real effects at all.

P.S. these examples were negative shocks, but you could tell a similar story with positive shocks.

vikingvista writes:


Your view makes the most sense to me. The construction of S&D curves in microeconomics is easy to understand as particular markets and products are considered. But aggregating *all* markets and products (marshmallows, aircraft carriers, steel, accountants, etc.), thinking that somehow AS and AD can be distinguished, and that there is a particular meaningful equilibrium price and quantity, doesn't make sense at all, except perhaps as a shorthand for describing what you explained. S&D seems to me to be completely different conceptually from AS&AD, despite their symbolic similarity.

Garrett writes:

S&D seems to me to be completely different conceptually from AS&AD, despite their symbolic similarity.

They are. The use of the terms "supply" and "demand" in macro leads to a lot of confusion. This is why it makes more sense to think of shocks as "real" versus "nominal," as Prof. Sumner has been pushing for years.

Pacemaker writes:

Your eleven steps look like they'd fit quite comfortably in an intermediate macro textbook. Levels 4 and 5 are about the IS curve, and levels 6 and 7 are about the LM curve. Levels 9 and 10 comprise the newfangled section on the zero bound and unconventional monetary policy, although textbooks would add that it works by raising expected inflation, which lowers the real interest rate.

Brian Donohue writes:

Don Geddis, Very good comment- I get this.

Scott, do you agree with Don?

emerich writes:

All I can say is, what a great encapsulation of macro debate.

vikingvista writes:

Don Geddis,

Why would the public suddenly wish to hold more currency? Animal spirits? Real shock? Or perhaps it doesn't matter?

Does the mechanism of rapidly expanding the monetary base not itself produce a real shock through its effects on capital redistribution?

Scott Sumner writes:

Thomas, An attempt to save more depresses output under a gold standard, but I doubt it has much effect under a fiat money standard.

Yancey, See Don's comment.

VikingVista, See Garrett's comment. I agree that AS/AD is very different.

Low interest rates cause a higher demand for currency.

Pacemaker, I have a post over at MoneyIllusion today arguing the IS/LM model is worthless, and should be abandoned. Check it out.

Brian, Yes.

Emerich. Thanks.


Return to top