Home > foundations of econ, macro, teaching > AS-AD Day for Economics Theory

AS-AD Day for Economics Theory

Since I’m currently a little sick (and hence I don’t really want to do much real work, like trying to graduate), I think I might just take a brief hiatus from my blogging hiatus to push back a bit on all the AS-AD hate out there in the econo-blogosphere. To put it briefly, I don’t think that AS-AD is research quality, obviously, but I don’t understand why (almost) everyone seems to agree that it isn’t fit to be taught (presumably to first or second year undergrads; if we’re arguing about teaching it to advanced students, I have to ask, who’s teaching it to advanced students?).  Frankly, I think AS-AD is a great teaching tool and I’m rather fond of it.

The proximate cause of this post is this, via Peter Dorman, lengthy take-down of AS-AD from Fred Moseley.   I am sympathetic to some of the hate out there (like this from Krugman), but not this.   Not to be rude, but Moseley–who I have nothing against, personally or professionally–is just wrong. Let me go through the argument point by point.

Criticizing AS-AD as a Broader attack on Equilibrium Economics

1. It is logically inconsistent outside of equilibrium. The AS and AD curves do not refer to separate economic agents making separate decisions about S and D (as consumers and firms in microeconomics), but instead refers to different theories of the relation between output and the price level in the same economy. Outside of equilibrium, these two different theories predict two different levels of output for the same price level; but the same economy cannot produce two different levels of output at the same time.

This isn’t just wrong; it doesn’t even make sense–although in fairness to Moseley the problem is rather technical.   To understand why, I need to back for a moment and explain something about AS-AD; at least, how I understand AS-AD.

AS-AD is the only example in Economics that I can think of which can be understood as what physicist call an adiabatic approximation–what I like to call “equilibrium disequilibrium” (which I think is more descriptive).   Think of it this way: there are “equilibrating” processes and “disequilibrating” processes which evolve the system through time.  If the equilibrating process is “fast” compared to the disequilibrating process, then the system can be approximated as a series of equilibria.   If the system is in equilibrium {P(t),Y(t)} at time t, then at time t + dt, the system is in a new equilibrium, {P(t + dt),Y(t+ dt)}, and the change {dP,dY} depends only on the disequilibrating process (approximately).   In AS-AD, the equilibrating process are “plans and expectations” of individual agents and the disequilibrating processes are the goods and factor markets.   Since it is the goods/factors markets which are out of equilibrium, AS-AD is specifically an adiabatic approximation to the classical Spending Allocation model of Econ 101 in which these markets ARE in equilibrium.

So here’s the problem with what Moseley is trying to say: it doesn’t make a lick of sense to talk about disequilibrium dynamics of an (approximate) disequilibrium model.   He is of course correct in saying that the model would “predict” two distinct levels for output, but that is because output is already out of equilibrium!   Anyway, moving on.

2. Because of the logical inconsistency between AS and AD, the model cannot explain the adjustment process to equilibrium. For example, in the case of excess supply, the AD curves implies that output should increase in order to restore equilibrium and the neo-classical AS curve implies that output shoulddecrease in order to restore equilibrium. But the output of the same economy cannot both increase and decrease at the same time.  In the “sticky price” model, AS > AD has no meaning, because AD is a quantity and AS is a price.

Most of this I just covered, but there are some additional claims here.   The issue, I think, is whether AS and AD can exist separately.   If not, then the period-by-period equilibrium condition makes no sense, and so neither would the adiabatic approximation.  I think they can, and the issue here is really important to understanding the model.

So, what is AS?  Broadly it is the schedule of price adjustments to observed sales (and observed sales to price adjustments) for the firms and individuals in the economy.   For example, if the economy is populated by, say, my favorite (admittedly intractable) model of ignorant monopolists facing idiosyncratic demand curves against competitors selling highly substitutable output (fortunately, this model ought to look on average a bit like monopolisticly competitive firms subject to Calvo pricing).   That should give us a nice upward convex shape for AS in {P,Y} space.

And what is AD?  It’s the convolution of the loci of IS-LM equilibria (so it is the spending/income response to interest rates) with the Fed’s response to economic data (Fed reaction function/Taylor rule/whatever).   You can think of the IS/LM relationship as a constraint the Fed faces (there is, of course, no economic actor who decides where the IS curve or LM curve should be–they are both the results of the general equilibrium condition) and hence AD can be thought of as the output/inflation trade-off that the Fed will accept.

The key point I want to make is that each of these curves involve separate decision makers (businesses and consumers for AS and the Fed for AD) forming plans.   This is the “fast” process in our adiabatic approximation and so it is a perfectly valid, and each decision maker will wait to see what happens (goods sell/competitors raise prices for AS and inflation/output too high for AD) before adjusting those plans–the goods/factors markets are out of equilibrium.   This is exactly what I described as the adiabatic approximation above.   Next point:

3. The AS-AD model is empirically unrealistic. The model predicts that AS > AD causes prices to fall, but prices have not fallen since the 1930s (more than a percent or two once or twice). And it is a very good thing that prices no longer fall! Because significant deflation would be a disaster for such a heavily indebted economy as the US. And yet the AS-AD models in the textbooks still present deflation as an unproblematic solution to excess supply. Ben Bernanke, in his real-world job as Chairman of the Fed, is doing everything he can possibly think of in order to avoid deflation. And yet his intermediate macro textbook (co-authored) still presents the AS-AD model and deflation as a solution to excess supply.

This seems to me to be a critique of some exceptionally naive versions of AS-AD: so I’ll have to chalk this up as straw man punching and leave it at that.   Yes, if your particular version of AS-AD implies deflation every time growth falls below trend, then its probably wrong.   It probably doesn’t have production or realistic investment either–because AS-AD doesn’t really attempt to model any of these things.

With the AS-AD models the rest of us are using, inflation falls, not prices when there is sufficient slack in the economy.   Even that is not necessary: an inflation targeting CB will be successful in keeping inflation fixed right where it is–although, in the spirit of my adiabatic approximation rant, it’s only right that I point out that inflation targeting will cause inflation to meander a bit around target; tightly around target if the CB is viewed to be competent (so that expectations of excess inflation are low) and more wildly if not.

Why all the Hate

Moseley writes

We should not be teaching such a logically contradictory and empirically unrealistic to our students. It encourages sloppy thinking and memorization, rather than rigorous and critical and creative thinking.

I don’t see it.   As I’ve just been arguing, the logical contradictions are really just his discomfort with what I think is a perfectly reasonable approximation.   The model is plenty consistent.   If you are confusing your students so much then, perhaps, the problem is with you.   I would argue that there is no simpler way to teach inflation/output dynamics to freshmen (which I think is what Krugman was getting at).

Now, I can see why would might argue that AS-AD is not ideal, but it is still useful to know, even for researchers.   If for no other reason than as a quick gut check for fancier models.

Frankly, I’m a bit mystified by all the vitriol flung it’s way in this latest blog spat.   If you can explain to me what harm it is actually doing to students, please, leave a reply in comments because I truly don’t understand.   As I see it, the alternative to teaching AS-AD is to stop at IS-LM, which doesn’t include inflation or dynamics (your students–the better ones–I’m sure, would notice).   You could, of course, try to teach more modern models, say DSGE… except I’m not actually sure that in terms of “logical consistency” that DSGE actually beats AS-AD (I actually would have some critiques of my own to add, but that’s a subject for another time).   And don’t tell me you’re going to teach your students some heterodox theory; if they come away from your class speaking a different language than mainstream economists, you’ve done them a real disservice.

  1. No comments yet.
  1. July 14, 2014 at 2:25 pm

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: