Wednesday, July 16, 2014

Aggregate demand is aggregate information

Scott Sumner asks [1] what aggregate demand (AD) is:
So what do [economists] think AD is? ... What is held constant along a given AD curve?  Presumably a given AD curve is supposed to be holding constant things like monetary and fiscal policy, animal spirits, consumer sentiment, etc.
The information transfer model has a simple answer: AD is a source of information that is received by the aggregate supply (AS), with the price level (i.e. the "price" in the AD/AS model) detecting the information transfer.

For a single good we have something like "I would like a pound of bacon for 6 dollars" (demand information) is received by a physical pound of bacon (supply) being sold ... with the price paid "detecting" the information transfer.

Of course, you have people who would buy bacon at 2 dollars or at 10 dollars. This information is "lost" at the supply; in the former case because the sale doesn't happen and in the latter case because the extra 4 dollars is not collected.

This information from the demand includes all kinds of expectations and theories as well. "I think the price of bacon will rise tomorrow to 8 dollars" is also communicated by a purchase of bacon at $6 today -- but again with some information loss. "Bacon is yummy at any price" or "I am making bacon wrapped shrimp for dinner tonight" also comes through, but again with some information loss.

In this way, information received by the supply (IS) is generally less than information received by the demand (ID), or IS < ID. However, IS ~ ID is a good approximation for a functioning market [2]. And only in the case of IS ~ ID do you get supply and demand curves (otherwise you get supply and demand "regions" bounded by the supply and demand curves).

This picture means that IAD is the aggregate of all kinds of economic information, from everyone's future expected path of monetary policy, exchange rates or the size of the tech sector to the current price of commodities and your paycheck. AS is the set of all the things that get made and/or sold because of that information. Typically, IAD ~ IAS is a good approximation in a functioning market, and that gives you AD and AS curves which represent the effect on the price level of a given set of information being sent (AD) or received (AS). (These curves then intersect at some value where AD = AS.)

If you gathered up all the information people put forward at one point in time (think of it as one grand "specification" for an economy ... like an engineering spec) and then varied the amount of stuff actually produced by the economy, that would trace out the "AD curve". The price level falls as you produce more stuff to that paricular specification. In that sense, you can say the entire curve is "the" AD ... at one point in time. This answers Sumner's question about what is held constant along an AD curve. [3]

Now maybe this is just my opinion, but I think this paints an extraordinarily clear picture of what AD is.

[1] This post is an adaptation of my comment on Sumner's blog.

[2] I made the argument once before that maybe ID ~ IS is a condition required to have a sensible theory of macroeconomics.

[3] For another take, from the perspective of thermodynamics, see here. That specification is the "demand bath" analogous to a heat bath in isothermal processes.


  1. Jason, I think you have a typo here:

    "AD is a source of information that is received by the aggregate supply (AD)"

    Last bit there should be "(AS)"

    Also, you introduce "IAD" and "IAS" but you don't spell out exactly how those differ from "AD" and "AS" respectively. I guess it's fairly clear ("I" stand for information), but it caught my eye since it was different than your comment on Sumner's blog. There you concluded that AD ~ AS was a good approximation, but here you have IAD ~ IAS. I assume you're happier with the latter, but why the change exactly? Is it still OK to claim AD ~ AS is a good approximation too?

    Also, Sumner follows up with this comment:

    "Jason, How is that model more useful than the standard model? What problem does it solve that the standard model cannot solve?"

    Nice going, BTW. I was hoping you'd chip in a definition for him!

    1. I responded but the comment was put under moderation (probably because it was long and had a bunch of links).

      Fixed the typo and added a sentence about AD = AS ... IAD ~ IAS (information) means you can derive supply and demand curves that intersect at some value of AD = AS (dollars).

    2. I'm glad you responded. I thought perhaps you were occupied w/ other things and I was going to jump into modern jackass territory today and give him a brief write up on my impression of one of the benefits of your model, before he moved on to other topics.... but I think I wisely did not hit "submit." :D

      I suspect the "goes into moderation" limit on the number of links on Sumner's site is about 4. Sadowski would regularly hit the limit and then start putting links in square brackets (e.g. []) to get around that. The square brackets (or perhaps just the adjacent characters) seem to prevent the text from getting upgraded to links, but of course the interested reader can still cut and paste. I've used that trick on occasion too.

    3. Ah, your comment is out of moderation. Nice. I was happy that you wrote this:

      "This model of the price level also reduces to the quantity theory of money when inflation is high, and explains the deviation from the quantity theory at low inflation."

      Because that's essentially what I wrote (but did not submit). However, I also played with adding a couple of other ideas, which I erased:

      1. Jason's model is not intended to be a replacement for other macro models, but rather a complement to them. It has strengths and weaknesses. E.g., It is probably not the best choice for analyzing a host of important policy options (e.g. tax policy changes, IT vs NGDPLT, etc).

      2. How can macro benefit from Jason's model (i.e. what are its strengths)? His macro model (BTW, he also has micro models) are "very macro" in the sense that they may not capture ordinary business cycles, or tell you much about the macro effects of a host of specific policies, but instead they capture large trends, or perhaps the long term effects of classes of (CB) policies[1], for instance...

      [this is where I was going to mention the bit about the information transfer index, and how its calculated value for a particular economy can indicate if the QTM should apply or not, etc]

      [1] what I intended by this comment about "classes of policies" (though I wasn't going to get into it), is the idea of "feedback" (market based) CB policies leading to an endogenous solution to your P equation, vs "blind" policies leading to the exogenous solution.

      So what's your critique of the above? Like I say, had I made any comment, I decided to restrict it to just the example of how kappa can tell if QTM should apply or not, but I'm curious how far off I was on the other stuff.

    4. BTW, if Sumner responds to your post I predict he will complain about your use of the phrase "unit of account." Not that there's anything wrong with it, but

      1. It appears like you use it in a non-standard way

      2. Sumner usually complains when someone brings up UoA in any context (I have the impression that he doesn't think it's a useful concept, or perhaps that it's an uninteresting concept).

    5. Regarding "unit of account", I'm really just not sure what to call it, so I kind of resorted to more traditional economic names for the functions of money. It's really related to how much information conveyed by a unit of account.

      Both of your descriptions above seem fine. The ITM does say that tax policy is largely irrelevant to first order, although government spending does have the rather straightforward impact you'd expect (if only because NGDP = C + I + G + (X - M) and G is there to first order). It can be "offset" if kappa is small, though.

      Regarding the business cycle, I'd say the ITM finally gives you the baseline trend around which you can define what a business cycle really is. But maybe that is too pompous.

    6. Well my prediction failed: he didn't mention UoA, but he did ask this:

      "Jason, Does the model forecast P better than alternative approaches, like TIPS spreads?"

    7. I answered. The model gives both interest rates and inflation, so it should be able to predict TIPS spreads! I will do a post on that as soon as I find some time.

    8. I hope you don't mind, but I gave him a link to your P* comparison too: that qualifies as an "alternative approach" to compare P forecasts against I think.

    9. Thanks.

      I wrote a new post showing the ITM is 5 times better than the TIPS spread at predicting inflation :)

  2. O/T: I see that Noah is revisiting the topic of "Austrians" today. If you really want to bump up your site visits and start selling ads, maybe you should think about doing the same?... :D

    But be sure to keep it very abstract, so that nobody can figure out what your position is... but add lots of exclamation points, and double and triple up on the question marks and *asterisks*, so that it seem like you're very passionate about your position (whatever it is). :D


Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.