Saturday, June 1, 2013

Why You Don’t See the Aggregate Supply—Aggregate Demand Model in the Econ Blogosphere

Introductory textbooks are supposed to give you simplified versions of the models that professionals use in their own work.  The blogosphere is a realm where people from a range of backgrounds discuss current issues often using simplified concepts so everyone can be on the same page.

But while the dominant framework used in introductory macro textbooks is aggregate supply—aggregate demand (AS-AD), it is almost never mentioned in the econ blogs.  My guess is that anyone who tried to make an argument about current macropolicy using an AS-AD diagram would just invite snickers.  This is not true on the micro side, where it’s perfectly normal to make an argument with a standard issue, partial equilibrium supply and demand diagram.  What’s going on here?

I’ve been writing the part of my textbook where I describe what happened in macro during the period from the mid 70s to the mid 00s, and part of the story is the rise of textbook AS-AD.  Here’s the line I take:

The dominant macro model, now crystallized in DSGE, is much too complex for intro students.  It is based on intertemporal optimization and general equilibrium theory.  There is no possible way to explain it to students in their first exposure to economics.  But the mainstream has rejected the old income-expenditure models that graced intro texts in the 1970s and were, in skeleton form, the basis for the forecasting models used back in those days.  So what to do?

The solution has been to use AS-AD as a placeholder.  It allows instructors to talk about both prices and quantities in a rough market context.  By putting Y on one axis and P on another, you can locate any macroeconomic outcome in the upper-right quadrant.  It gets students “thinking like economists”.

Unfortunately the model is unsound.  If you dig into it you find contradictions that can’t be papered over.  One example is that the AS curve depends on the idea that input prices for firms systematically lag output prices, but do you really want to argue the theoretical and empirical case for this?  Or try the AD assumption that, even as the price level and real output in the economy go up or down, the money supply remains fixed.

That’s why AS-AD is simply a placeholder.  It has no intrinsic value as an economic model.  No one uses it for policy purposes.  It can’t be found in the econ blogs.  It’s not a stripped down version of DSGE.  Its only role is to occupy student brain cells until the real work of macroeconomic instruction can begin in a more advanced course.

If I’m wrong I’d like to know before I cut off all lines of retreat.

12 comments:

Edward Lambert said...

I am recently developing another model based on aggregate supply and Effective Demand. That is right, effective demand. It gives the same but much more information than the AS-AD model. Here is the latest post on the model.
http://effectivedemand.typepad.com/ed/2013/05/aggregate-supply-effective-demand-model-points-of-price-dynamics.html

Unknown said...

IS/LM still has a long and healthy life in its Mundell-Fleming international-trade incarnation, where the main issues of interest can be treated quite well without resort to the complexity of DSGE.

My impression is that the biggest problem with simple IS/LM is the sleight of hand between nominal and real interest rates; you can construct AS and AD to avoid the problems you describe (e.g., IS-MP to endogenize the money supply), but suppressing the nominal/real magic is harder. Given that this a primary issue in domestic policy, there's no way to resurrect IS-LM except for pedagogy.

Unless, again, you're working in trade macro and have explicit interest rate divergences anyway...

Peter Dorman said...

Gang, this isn't about what kind of macro model you'd like to see in the textbooks. That's worth discussing, but not here. What's on the table is how we should understand the intense attachment of intro macro textbooks (and instructors) to AS-AD in spite of the fact that they would never want to be seen using it in polite company.

Unknown said...

You've explained it already: it gets students "thinking like economists" whilst still remaining extremely simple, and there's no point trying to correct its most egregious flaws because the whole hydraulic approach is junk anyway.

We could all teach IS-MP instead but what for? Students who actually go on to EC300 classes will be discarding all of it. And in the meanwhile we lose six decades of pedagogical tradition since Paul A. Samuelson first invented the practice of starting with AS/AD and going into an LRAS/SRAS distinction nobody believes any more, etc.

Unknown said...

It's hardly the only pedagogical lie going on here, incidentally: over in micro we still teach undergraduates and high-schoolers all about Marshallian "consumer and producer surplus", in a way that all but collapses once we ask graduate freshmen to discard partial for general eq, and shaded graphs for convex sets and fixed-point theorems.

Again, (1) Paul Samuelson Did It (2) it's helluva a lot simpler.

Unknown said...

I seem to have lost a comment to the spam filter...

Nick Rowe said...

Testing first, because I keep getting trapped in spam filters, even on my own blog!!

Nick Rowe said...

Peter: I disagree.

1. I have done a number of posts using the AS-AD framework. When I take the trouble to use Paint, I even draw the picture. I did one just yesterday. Some people sometimes snicker at some of my posts, but I don't remember anyone snickering at me for using the AS-AD framework. But I'm not sure if I'm an outlier in the blogosphere. It might just be that some bloggers just don't bother to draw the picture.

2. The AD curve does not have to assume the money supply is held fixed. For example, if the central bank is targeting the rate of interest (or real GDP) in the short run, but targeting the price level in the long run, you get a vertical short run AD curve and a horizontal long run AD curve. And this is roughly what modern central banks do (except they target inflation rather than the price level, so maybe you should put the inflation rate rather than the price level on the vertical axis). See for example my old post: http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/03/why-the-lm-is-usually-vertical-and-the-ad-curve-usually-horizontal.html

Or if the central bank targets NGDP you get a rectangular hyperbola AD curve.

Maybe a better way to pose the question is a normative one: what shape ought the central bank choose for the AD curve? http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/11/whats-the-best-shape-for-the-ad-curve.html

3. You can draw an AS curve without assuming that output prices adjust faster than input prices. The key is that what we call an AS curve isn't really an aggregate *supply* curve, especially in the short run. Because a *supply* curve is supposed to tell us how much firms *want* to sell, and the "AS" curve in macro doesn't necessarily really mean that. Monopolistically competitive firms don't even have *supply* curves. We should just call it "the other curve"! But if you keep that in mind, there's nothing problematic about representing the standard DSGE model, which assumes monopolistically competitive price-setting firms, in a simple AS-AD picture like I do here: http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/01/macroeconomics-with-monopolistic-competition-in-pictures.html

Benoit Essiambre said...

It is a mistake to dismiss models with fewer variables and replace them with models with more variables when the simple models are already difficult to tune to real world data.

A model with more variables is only useful in science when the simple models are well understood and their parameters easily determined.

Models are exponentially more difficult to tune to data as you start tracking more variables. You quickly get problems of overfitting because of scarce data on particular combinations of variable values. Instead of telling you more precise things about the state of your system, it locks onto spurious patterns that happened by chance in your data. This means the model no longer allows you to make predictions about unknowns but instead only describe in too much details the past knowns.

Anybody contemplating tradoffs between simpler and more complex models should read Yudkowski's essay on the subject "A Technical Explanation of Technical Explanation":http://yudkowsky.net/rational/technical

rosserjb@jmu.edu said...

I know this is not Peter's question, and I do think that while we do not see it all that much in the econoblogosphere, some do use AS-AD, and certainly in private conversations, quite "respectable" economists will talk about AS versus AD effects when discussing macro.

One way to put them together is to use the Keynesian income-expenditure model to derive the position of the AD curve. Obviously the upward-sloping AS curve has some problems and is a bit of a handwave, but one can find it described as in the textbooks in one of the later chapters of Keynes's GT, 23, I think. It is not that awful and generally tells one more about policy than does most DSGEs.

rosserjb@jmu.edu said...

I know this is not Peter's question, and I do think that while we do not see it all that much in the econoblogosphere, some do use AS-AD, and certainly in private conversations, quite "respectable" economists will talk about AS versus AD effects when discussing macro.

One way to put them together is to use the Keynesian income-expenditure model to derive the position of the AD curve. Obviously the upward-sloping AS curve has some problems and is a bit of a handwave, but one can find it described as in the textbooks in one of the later chapters of Keynes's GT, 23, I think. It is not that awful and generally tells one more about policy than does most DSGEs.

rosserjb@jmu.edu said...

I'll add to here a comment from some I put on economistsview after Waldmann and Krugman got into the fray. To be able to make the leap from income-expenditure to AS-AD one needs to make the former be done in nominal terms. It is doing it in real terms that tends to make things into such a mess. However, there remains a loose end in determing the intercept of the AE, wherein remains the nominal-real tangle. Otherwise, it works through reasonably logically and usefully.