In defence of the AS-AD framework

I disagree with Peter Dorman (HT Mark Thoma). I have used the AS-AD framework many times in my blog posts (and I don't remember anyone ever snickering at me for doing so) because I find it useful. Paul Krugman offers a partial defence of AS-AD, with the following caveat:

"So there is a place for AS-AD, although it’s an awkward one, and the transition to IS curve plus Taylor rule plus Phillips curve, which is the model you really want to use for America right now, is a moment that fills me with dread every time we take it on in a new edition."

I understand that dread. I know where he's coming from. But I have decided we need to face it down. Because it's a false dread.

I prefer to call it the AS-AD framework, rather than the AS-AD model. That's because it's a useful way of looking at many models, sometimes very different.

Peter says that the AD curve assumes the money supply is fixed. It doesn't. Or rather, it doesn't have to. You don't need to make that assumption if you draw an AD curve. And any textbook that draws an AD curve for an open economy with fixed exchange rates certainly does not assume the money supply is fixed.

You get one AD curve if you assume the central bank targets the money supply. You get a different AD curve if you assume it targets the exchange rate. You get a third AD curve if you assume it targets NGDP (it's a rectangular hyperbola). A fourth AD curve if you assume it targets the price level (it's horizontal). A fifth AD curve if it targets (or tries to target) real GDP (it's vertical). A sixth AD curve if it targets (tries to target) a rate of interest (it's vertical again). And so on. (And if you draw a vertical AD curve and a vertical LRAS curve you can see immediately that something is wrong, and why I said "tries to target".)

For a modern central bank like the Bank of Canada, which targets an interest rate in the very short run, what it thinks is potential output in the medium run, and 2% inflation in the long run, you might want to draw a vertical short run and medium run AD curve, and a horizontal long run AD curve (though you might put the inflation rate rather than the price level on the vertical axis).

But when we teach economics we are not (or should not be) just teaching about the here and now. It's not just about America (or Canada) right now. My students come from all over the world, where monetary policies can be very different. And we want to teach a framework that can help them understand the past, and different possible futures, as well as the present. Canada has had fixed exchange rates in the past. It might switch to targeting NGDP in the future. I can show my students both those policies with a (different) AD curve. I cannot show them either of those policies with an IS curve plus Taylor Rule. (And the Bank of Canada does not currently follow a Taylor Rule anyway).

That's my defence of the AD curve. Now for the AS curve.

The biggest trouble with the AS curve is its name. Because only in a very limited class of models is it a supply curve. Because, strictly speaking, a "supply" curve tells you how much output firms want to sell, at any given price. If you've got sticky prices, and the economy is in recession, the economy might be on its SRAS curve, but firms will want to sell more output than they are actually selling. Even the LRAS curve isn't really a supply curve, if you have monopolistically competitive firms, because monopolistically competitive firms don't strictly have supply curves.

We should probably just call it "the other curve". But the "AS" name has stuck. Never mind. (And the AD curve isn't strictly a demand curve either, for that matter. Because, as the Old Keynesians taught us, but younger generations of New Keynesians might have forgotten, the level of output demanded depends on income, which depends on the amount of output actually sold, which depends on the amount of output demanded. So the AD curve is really a semi-equilibrium condition, showing points at which output equals output demanded, and unlike micro demand curves it does not tell us how much output will be demanded if we are off that curve).

If you believe (as most of us do) that prices are flexible in the long run but sticky in the short run then you get a LRAS which is normally vertical and a SRAS which isn't vertical. But the exact shape of that SRAS depends on what particular prices are sticky, and on many other features of the model.

Let me give one example. Suppose that in the short run all output prices are perfectly fixed. That gives you a horizontal SRAS curve. But does the SRAS curve extend to the right of the LRAS curve? That depends. If you have perfectly competitive firms that refuse to sell additional output if it's not profitable for them to do so, even if the demand is there, then the SRAS curve stops dead when it hits the LRAS curve. If you start on the LRAS curve and AD shifts right but prices are fixed, firms simply ration customers. But if you have monopolistically competitive firms the SRAS curve will continue to the right past the LRAS curve before it too eventually stops dead.

That's a big difference. You can show that difference with AS curves. You can't show that difference with a Phillips Curve.

Now you might say that the trouble with the SRAS curve is that the price level is on the vertical axis, rather than the inflation rate, which is what should be on the vertical axis if you have a Phillips Curve. Because prices may be sticky, but they are not stuck. OK, you have a point. But then the inflation rate isn't stuck either. And nor is the rate of change of the inflation rate stuck.

I call the insistence that we put inflation on the vertical axis the "fetish of the first derivative". There's nothing special about the first derivative. If you prefer Pdot to P, why stop there? You are off down a slippery slope to putting Pdoubledot on the vertical axis, then Ptripledot, and so on.

Unless you believe that the price level is a jump variable, and most of us don't (except maybe for a small subset of perfectly flexible prices), you ought to start out with the price level on the vertical axis. (And empirically, despite what the Calvo Phillips Curve says, inflation doesn't seem to be a jump variable either.) Yes, the SRAS curve will start shifting over time if the AD curve shifts. But so will the Short Run Phillips Curve. And so will the Short Run Phillipsdot curve. Forget the fetish of the first derivative. If 2% inflation and 3% real growth is normal in your time and place, just draw the AS-AD curves on a bit of paper, and tell your students to imagine that bit of paper is moving East-North-East at 5% per year. That's what I do.

So Paul: forget the dread. AS-AD is AOK.

(And any of you young DSGE-infused whippersnappers who even think of snickering might want to make very sure they actually understand AS-AD first!)

Comments

[greg]

It seems to me it comes down to what you can show. For instance a country with an AS curve. That AS curve does not change if its trade is balanced. However, if that country runs a trade deficit, its AS curve shifts to the right, (and if it runs a trade surplus its AS curve shifts to the left,) and the equilibrium point moves down (or up,) the downward sloping AD curve, changing the price level faced by domestic firms. However, domestic firms produce on the original AS curve, so their revenue is reduced. And since their demand is equal to their revenue, assuming no borrowing, the AD curve shifts downward. If we assume we are in a region where the AS curve is no longer vertical, but upward sloping, then the quantity produced by domestic firms also decreases, and we enter a contractionary, and deflationary, spiral. (Similarly, a country running a trade surplus enters an expansionary and inflationary spiral.)

I haven't been too precise with the definitions of AS and AD, except to keep them constant. Any validity to this argument?

[Nick Rowe]

greg: we put domestic production (GDP) on the horizontal axis. An increased demand for our net exports may (or may not, depending on the model and on the central bank's monetary policy) shift the AD curve to the right.

[Bill Woolsey]

Nick:

One characteristic of market monetarists is that many of us teach undergraduate macro.

It appears another characteristic is that we take the AD-AS model (or framework) seriously.

Does a negative slo[pe] for the AD curve require a constant quantity of money? Not exactly. Constant nominal income does the same. As long as nominal income doesn't change in proportion to the price level there is going to be a negative slope.

Of course, with interest rate targeting, there is exactly that tendency.

Interesting that Market Monetarists are all skeptics/critics of interest rate targeting. That is, critics of what central banks like to do.

What is the key problem with AS-AD? No interest rates on the diagram.

As explained by Dorman, to tell the story that central banks want told, you need interest rates. Sure enough, an exogenous interest rate and IS gives you real expenditures on output. Okun's law gives employment and I suppose unemployment. The output gap version of the philliip's curve gives inflation.

Inflation and unemployment are the political lightning rods that central banks must limit and avoid.

If the goal is to stabilize interest rates subject to the constraint that unemployment and inflation not become unacceptably high, then IS and the Phillips curve makes sense.

If interest rates are not important directly, and only inflation and output gaps are problems, AS and AD is just fine.

You have already shown how AS-AD can be used to show the indeterminacy of interest rate targeting. (And what you do is cause macroeconomic disequilibrium to reverse deviations of the price level from target at best. That doesn't sound very good.

Further, the horizontal aggregate demand curve at the target price level makes it painfully obvious the disastrous consequence of the rule for short run supply shocks.

And, by the way, if you do the price level analysis (rather than use growth rates,) then we can see how it is quite possible to have a 3% growth rate of real output with real output remaining well below potential. Gee, maybe the levels matter.

[edited to fix typo above, because I didn't want that typo to detract from a very good comment. NR]

[Frances Woolley]

Nick: "But when we teach economics we are not (or should not be) just teaching about the here and now."

Yes! absolutely!

Interesting comments about the AS curve. As you say, the name Aggregate Supply Curve is often inappropriate. But names matter - yesterday I saw an ad for an on-line casino called "BetFair". It's guaranteed to offer anything but fair odds, but the name conveys somehow both fun (funfair) and equity (fairness).

How do you think the name "aggregate supply" curve affects the way that people perceive and interpret the curve? How does the name matter?

As a micro person, I'd answer the question by saying that calling something a supply curve makes it sound that it can only be moved by fundamental technological changes, and so shifts focus to discussions about productivity or technology, but you might have a different perspective.

[Alex Bollinger]

"Fetish of the first derivative" is the phrase of the day. Interesting post.

[Steve Roth]

"You get one AD curve if..."

What curve do you get if it targets both inflation and (in an admittedly desultory manner) unemployment?

[Nick Rowe]

Bill: excellent comment. I have nothing to add.

Frances: Yes, I think the name does matter. The neat thing about calling something a "Phillips Curve" is that that name doesn't mean anything, so it's open to different interpretations of what causes it. The best name would be: "That curve that is like a Phillips Curve, only in levels not rates of change".

But the LRAS curve gets moved by fundamentals like technology, resources, preferences, market structure, etc. We could think of the LRAS curve as showing all the "real" forces, the AD curve as showing all the "monetary" forces, and the SRAS curve showing how the two interact when prices are slow to adjust.

Alex: thanks! I'm not sure if "fetish of the first derivative" is original with me. I remember someone (Brad DeLong?) saying that the only difference between us and David Hume is that we have switched to talking about the first derivative.

[Nick Rowe]

Steve: it can't strictly target both. Because that would mean 2 AD curves (one vertical and one horizontal). But it can target a mixture or weighted average of the two. If it's an average, you get something roughly like a rectangular hyperbola (as with NGDP) except that inflation should be on the vertical axis. If there's some sort of threshhold target ("We target 5% unemployment, unless inflation is over 3%") it's going to be sort of L-shaped.

[John Chilton]

Bernanke and Frank, Principles of Macro, puts inflation on the vertical. Thoughts? Presumably it was a pedagogical choice with some motivation. "Anchored expectations (of inflation?)"?

[Steve Roth]

@Nick: Thanks!

[Scott Sumner]

Nick, Excellent post. My only comment is that one can draw the AD curve as a rectangular hyperbola even if the central bank is not targeting NGDP. For example, suppose the central bank "accommodates" price shocks by stabilizing NGDP, but responds to financial crises by letting NGDP fluctuate.

[greg]

Nick, why wouldn't an excess of imports over exports shift the AS curve to the right? (And similarly a leftward shift if exports are greater than imports?) I got the idea that GDP is on the horizontal axis, but why just GDP? Why wouldn't you chart all sources of production, including foreign, that a nation's economy consumes. Wouldn't that be closer to the reality? Isn't that really the Aggregate Supply?

[Wonks Anonymous]

This is the Friedman quote on Hume:
We have advanced beyond Hume in two respects only: first, we now have a more secure grasp on the quantitative magnitudes involved: second, we have gone one derivative beyond Hume.

Found here:
http://www.humesociety.org/hs/issues/v15n1/mcgee/mcgee-v15n1.pdf

[Nick Rowe]

greg: Let Y be Canadian production, C consumption, I investment, G government spending, X exports, M imports.

There are 3 ways we could do the accounting:

1. Y+M=C+I+G+X (imports add to supply, and exports add to demand).

2. Y+M-X=C+I+G (imports add to supply, exports subtract from supply, the right hand side is call "domestic absorption").

3. Y=C+I+G+X-M (exports add to demand, imports subtract from demand).

All 3 ways are equivalent, and give the same answer (unless you screw something up). None is wrong. But 3 is the conventional way of doing it, and it's normally easier if everyone speaks the same language. Plus, if we are especially interested in Y, because Y is also income, then it's easier to see what happens to Y if we put Y on the axis. Plus, this means that the sort of things that shift AS (things like resources and productivity) are very different from the sort of things that shift AD, so it's easier to use the framework.

[Nick Rowe]

John: well, if you want to talk about inflation targeting, and if you want to assume the central bank follows a Taylor Rule (which gives the AD curve a downward slope in that space if the central bank falsely assumes the natural rate never changes) then that is probably the easier way to do it.

Scott: thanks! But I think I disagree with your comment. In principle, we could draw the AD curve any shape we want, but have the central bank shift AD in different ways in response to different shocks. But I don't think it's useful to do that. The whole point of drawing a curve is to find the right curve so that it doesn't shift much, or only shifts in a limited number of cases. For example, if the bank targets NGDP, then it's best to draw a rectangular hyperbola AD curve, because then the AD curve will only shift if: the bank changes its target; the bank screws it up.

Wonks: (I just rescued your comment from the spam filter.)

So it was Friedman, and not Brad DeLong. Shows how bad my memory is. I really should have remembered that. Thanks.

[Bill Woolsey]

Nick:

I sure don't do this in class, but I like your idea that the shape of the AD curve depends on the monetary regime.

I draw negatively sloped AD curves and shift them around.

By the way, with a gold standard, AD is negatively sloped, right?

[Nick Rowe]

Bill: I think so yes. For a small open economy, under fixed exchange rates, the slope of the AD is determined by the elasticity of net exports wrt the real exchange rate (plus a few other things, depending on the model). I think the AD for the world economy on the gold standard would also slope down, but I would have to think a bit to figure out what determines the slope. Something to do with the elasticity of demand and supply for gold.

[greg]

OK, Nick, so I've been looking at it from No. 2: Y+M-X=C+I+G. I'm trying to get this point of view. So set M>X, and simplify by ignoring I and G. So C>Y. But I don't get from this equation that the price level goes down along the AD curve, which, other things being equal, seems to me what should happen, since the aggregate supply increases. (AS curve shifts to right.) And if the price level goes down...so does the income of domestic producers, since they are still producing at the original AS curve. (Real domestic production does not change, with vertical AS curve.)

[Nick Rowe]

greg: start with M=X. Now increase M holding C, I, G, and X constant. In approach 3, AD shifts left, so P goes down.

Real domestic production and real domestic income (measured in terms of domestically-produced goods) stay the same, if the AS curve is vertical. But domestic income falls in dollar terms, since P falls.

[greg]

Nick: Yeah, I get that. So it would be deflationary. And if the AS curve curved to the left at a lower price level, a la Keynes, running a trade deficit would be contractionary, too.

Now start with M=X. Now increase X holding C, I, G, and M constant. In approach 3, AD shifts right, so P goes up.