Nothing earth-shattering here. This post is about teaching macroeconomics better. But there's just a chance it might help some of you understand macro a little bit better.
Whenever we teach the Keynesian Cross (income expenditure) model, and the distinction between National Income Accounting identities and equilibrium conditions, we always talk about inventory accumulation. Planned vs. unplanned inventory accumulation. Desired vs. actual inventories. All that stuff. We should stop doing this. It only confuses the students, and may confuse us too. Plus, it makes little sense in an economy where manufacturing is a smaller and smaller share of GDP, and some goods are made to order anyway. We live in a service economy, and a made-to-order economy, where the inventory story makes no sense anyway. So we should stop doing it. And textbooks should stop doing it too.
This is what we should do instead.
Think about services, not manufacturing. Think haircuts, not widgets. A haircut is produced and sold at the same time. It is produced to order. There are no inventories of haircuts. Don't even mention the "I-word". Leave it to the chapter on investment demand, where it belongs.
In micro we make a careful distinction between three quantities: quantity supplied, quantity demanded; and actual quantity bought and sold. Quantity supplied is what sellers want to sell; quantity demanded is what buyers want to buy; and these are both conceptually distinct from the actual quantity bought and sold. There's Qs, Qd, and Q.
We should make exactly the same distinction in macro, only we are talking about aggregates of goods.
[Update: National income Accounting is about Q — how to measure Q, and how to define Q and add up different components of Q in a logically consistent way. Economics is about what determines Qs and Qd, and how Qs and/or Qd determine Q.]
The AE curve in the Keynesian Cross diagram is a demand curve. Sure, it's a weird demand curve, because it slopes up, and has weird things on the axes, and quantity demanded is on the vertical axis (actually, where it should be, given it's the dependent variable) and not on the horizontal axis like normal. But it shows aggregate quantity of goods demanded as a function of aggregate income from the sale of goods. So it's a demand curve. (OK, it's an aggregate Engel curve, if you insist).
The 45 degree line is a semi-equilibrium condition. Full equilibrium requires Qd=Qs=Q. The 45 degree line shows only that AE=Y, which is just the aggregate version of Qd=Q. It ignores the supply side equilibrium condition that Qs=Q. Of course it does. The Keynesian Cross model is part of a model of aggregate demand. It ignores Aggregate Supply. That's OK. We bring in the supply-side later.
At the equilibrium point, where the AE curve crosses the 45 degree line, income from the sale of services (and goods) Y* is at exactly the right level for the demand for services (and goods) at that level of income to equal that same level of income. We are at a point "on" the demand curve.
And here's a story to tell about how we get there, if we start out from a level of income below equilibrium. Suppose people expect to be able to sell Y1 services (and goods), and so expect to have income equal to Y1, in aggregate. They demand AE1=AE(Y1) of services (and goods). And AE1>Y1. Firms (because they want to sell even more) are willing to produce and sell AE1 worth of services (and goods). So actual income exceeds the income people expected to earn. So people increase their demand for services (and goods) based on that higher level of expected income. Etc.
Then run the same thing in reverse, if we start out at a level of income above equilibrium.
Don't mention inventories at all.
Oh, and one more thing. There is absolutely no need to talk about income rising because firms hire more workers, or income falling because firms lay off workers. It's wrong. That has absolutely nothing to do with the reason income rises or falls. Suppose the demand for haircuts falls. So fewer haircuts get sold, and the output of haircuts falls. It makes absolutely no difference whether the hairdressers sit idle in the salon, or get laid off and sit idle at home. In both cases the output of haircuts has fallen and so the income from cutting hair falls by exactly the same amount.
The only difference it makes, whether the hairdressers sit idle in the salon rather at home, is to the distribution of income between wages and profits. If they are laid off it's wage income that falls; if they sit idle in the salon, but continue to get their wages, it's profit income that falls. Either way, output and income fall. We should stop this labourist bias. People's income and wage income are not the same thing. Capitalists are people too.
(I just tried it out in my second year macro class. It seemed to go OK.)
Worthwhile Canadian Initiative 
Nick,
How does this fit with an economy where international trade in manufactured goods is important? (I’m not sure this is at all related, but hey …I’m a micro economist!).
Just be sure to tell them that this is a toy, and the real world doesn’t look like that.
I like Krugman’s take on models. They’re illustrative, not definitive. Given that the external perception of economists is that they think it’s the other way around, hammering the point home throughout economics instruction sounds good to me.
Linda: maybe manufactured goods are a larger share of trade than they are of GDP, but in teaching the income expenditure model it’s the interaction between the demand for GDP and actual GDP that’s important. And inventories are such a small part of that interaction, and not an essential part either.
Jim: You mean you think our models are not the Truth, Whole Truth, and Nothing But the Truth?
I get that point by teaching about the SRAS curve. If you start with a standard textbook labour market, assume sticky wages, the SRAS curve should not slope upwards the way the textbook shows it. It should slope up till it hits LRAS, then suddenly bend back, because the labour market is now supply-constrained. (And if you assume sticky prices and a competitive output market the SRAS curve should not be horizontal the way the textbooks show it. It should be horizontal till it hits LRAS curve, then stop dead, because firms don’t want to sell any more.). Then I say the world doesn’t seem to look like that. Then we try to patch up the theory. It’s always a patch job.
“… it’s profit income that falls”
Isn’t it revenue that falls? Could a decrease in revenue could play out ways that doesn’t imply less profit income? Seems to me that the answer is ‘yes’ but I’m still mulling it over …
Nicely put; I have always disliked the inventory story. I will try this on my first year class this week. Fits in nicely.
Patrick: Revenue from the sale of newly-produced goods and services is income, so income falls. If the salon keep all the hairdressers, for the same hours and the same wages, labour income stays the same. So profit income (using “profit” in the accounting, not economic sense) has to take the hit.
G Voss: Thanks! I never understood why we ever started talking about inventory in this model anyway. I expect it was all part of that ex ante ex post muddle in the 1950’s. We never make this sort of fuss in micro. There’s Q, and Qd.
Thank you, Nick. This is exactly what is troubling me every time I teach the Keynesian Cross to students. I was just thinking about this problem again today. I agree with you, I also don’t like the mentioning of inventories, it’s just confusing. But can I ask you something: How do you explain to students that firms produce more goods and services if AE>Y, when we just explained them that factors of production are fixed (ch. 3 in Mankiw)? This is something that has been always troubling me. I would be curious how you teach that.
Hi Mariko!
I think there are two ways to handle this problem:
1. You can say “Assume that the price level is fixed, and that there are lots of unemployed resources, that are only too willing to work if they can get someone to hire them, and firms are only too willing to hire them if only they could sell the extra output they could produce”. Then draw a horizontal SRAS curve in {P,Y} space.
2. Or you can say “This is a theory of aggregate demand. We are ignoring the supply-side of the economy for the moment; we will introduce that later. We can derive a demand curve in micro without talking about the supply curve, and we are doing that here”
Conceptually though, there is one difference between a micro partial equilibrium demand function and a macro aggregate demand function:
When we draw the demand curve for apples, it depends on income, but we ignore the fact that the income of some of the people demanding apples depend on the quantity of apples sold. And that’s OK as an approximation, because income from selling apples is such a tiny part of all household income it’s safe to ignore it. So the quantity of apples actually sold has a negligible effect on the demand for apples.
But in macro, when we draw the demand curve for everything, it depends on income, which is equal to the quantity of everything sold.
What this means is that the IS curve and AD curve in macro is not strictly a demand curve. It’s a semi-equilibrium condition. It tells you what demand would be if quantity sold were equal to demand.
“Quantity supplied is what sellers want to sell; quantity demanded is what buyers want to buy . . . .” Do you mean, at the prevailing price? (Of course, it is an oversimplification to suppose there is just one price at a given time.) Or do you mean, at what each person expects to be the price by the time he can consummate a transaction? (Different people may have different expectations.)
I understand your argument on inventories re manufacturing. But, how about resource extraction where “inventories” are not assets, but rather liabilities (contained externalities)?
Of course, I’m going to argue oil sands tailings ponds. A deferred liability that makes today’s investments more economic. (btw did you notice Avatar director Cameron appearing with Alberta’s “blue people” in a press conference today?)
Probably not at all related to your blog topic – is this more micro?
I’m pretty sure when I learned the Keynesian Cross etc I just ignored all the inventory stuff. I don’t think any of the exams had any questions on the inventory aspect of it either, of if they did they were pretty minor questions.
Of course, I didn’t do so well so maybe I shouldn’t have ignored inventories.
Philo: the standard micro model assumes that buyers and sellers of a good today know the price they are paying and getting today. So quantities demanded and supplied are functions of that “prevailing” price (and of other things). Macro makes the same assumption.
JVFM: Those aren’t really inventories, which are stocks of goods that have been produced but not yet sold. More micro, really.
Michael: I’m not sure how many teachers and textbooks talk about inventories. My impression is that it’s common, unless they jump straight to the equilibrium point and never discuss the adjustment process at all. The inventory story came up again in the blogosphere about a year or so back, in one of the fights over Say’s Law. What remined me of it was preparing for my lecture, and reading Greg Mankiw’s and Bill Scarth’s intermediate macro text.
We used Blanchard’s text, and it was definitely covered in my course. There was a lot of talk about it during the lecture on the Keynesian Cross, but the adjustment process didn’t really turn up much in tutorial questions or exams.
Had to remind myself of the accounting. It’s a nitpick and doesn’t change anything but this is what I was thinking: Gross profit is revenue minus cost of goods sold (inventories are accounted for in COGS btw). R&D and some other expenses are in there. Operating profit is gross profit minus expenses but not taxes and interest. So … Recession hits, revenue falls, you decide to keep all your labour but you reduce other expenses to compensate. Operating profit can remain constant. Of course that just means somebody else’s has less revenue, so it comes out in the wash.
“OK, it’s an aggregate Engel curve, if you insist.”
That’s what I was told when I started learning economics, but I don’t think I’ve ever seen anyone put it in writing before. I’m not sure it’s true. An Engel curve is derived on the assumption that the consumer is not rationed in any market. Macro only makes sense if suppliers are rationed in some way.
Kevin: Funnily enough, I had never heard the AE curve described as an Engel curve before. (Embarrassing confession: and I had to Google “Engel Curve” just to check that an Engel Curve was what I thought it was!)
“I’m not sure it’s true.” Good point, but I disagree. There are “notional” demand curves (or functions), which assume buyers are unrationed in any other market, and there are “constrained” or “effective” demand curves which incorporate constraints on purchases or sales in other markets. I don’t see any reason why we can’t also distinguish notional vs effective Engel Curves too. The AE curve is an aggregate constrained Engel Curve.
Patrick: I think your nitpick is correct. But it does assume the firm wasn’t maximising profits before the recession.
Nick, I agree that if we’re talking about effective demand the Engel curve idea makes sense. The derivation needs a bit of work though. Funnily enough the economist who told me, circa 1980, that the simple Keynesian consumption function is essentially just an Engel curve had previously taught in Canada, so I’m wondering whether it might be a relic of a local verbal tradition?
Kevin: Did that economist previously teach at University Western Ontario? I first learned these ideas from David Laidler and Peter Howitt at UWO. Clower is the origin of thinking of the AE as a quantity constrained demand function. But I don’t remember anyone specifically mentioning Engel curve.
Good point. I’ve bookmarked this post and will see about working this in next time I teach macro.
“Leave it to the chapter on investment demand, where it belongs.”
In a world of hair-cuts, isn’t labor more analogous to inventories than investment? The short-term adjustment to swings in demand takes place through hours, not capital spending.
kharris: Interesting. I can see the analogy. Firms producing widgets keep inventories of widgets so they can keep production smooth even if sales fluctuate (or, can produce in batches, so production fluctuates, evn though sales are smooth). Firms producing haircuts may hire more workers than they need, because demand for haircuts fluctuates, and it’s costly to hire more labour at short notice.
But I think the analogy stops there. The labour of an idle hairdresser today is wasted; it can’t be stored up to double the number of haircuts she can produce tomorrow.
Nick,
A thank you:
I love your posts, regardless if I agree with everything you say or not, they have sped up my learning of economics 100 fold. This type of free education is something that I have really appreciated (Of course the same goes for Stephen, Frances and Mike), so thank you for your time and effort.
A comment:
Macro is aggregate, so of course their is no distinction between wage and profit at its most general, and making a distinction doesn’t offer any insight into the introduction of the topic, especially when you start by teaching static equilibrium models. However, income distribution does matter, i.e. the difference between wages and profits, when you look at the growth of the economy and welfare of the people who make it up. I think you have an unfounded willful ignorance bias, unless of course you are tongue in cheek, but a lot of people share your “opinion” as an excuse to take advantage of others so I think its worth calling you out every time….
And, I would argue that now a days capitalists, by percentage of wealth not numbers of individuals, are not “people too”; instead I would argue that they are mainly institutions.
And as for AE > Y and why this leads firms to produce more (of course in the future): doesn’t this follows from the definition of Y = AE + G + Trade, because terms on the right can be negative, and supply depends on AE rather than Y…
Just some more on Mariko’s question, and Kevin’s point about deriving the AE curve.
For Y less than Ys, there’s no problem in deriving the AE curve, IS curve, and AD curve. Maximise utility or profits subject to all the usual stuff plus a constraint on sales of output. But for Y greater than Ys, this doesn’t really make sense. You can’t force people and firms to sell more output than they want to. Actual Y would be stuck at Ys.
This means the AE curve stops dead when it hits Ys. And the IS curve, which normally is a locus of points in {Y,r} space such that Yd(Y,r)=Y, will have a sharp kink in it at Ys, plus a different interpretation, It’s no longer a semi-equilibrium condition. It’s now a demand curve Yd(Ys,r). Intuitively, it’s because the multiplier stops dead when firms are no longer willing to increase sales of output if demand exceeds output. So the IS curve gets much steeper to the right of Ys. Same with the AD curve.
Rick: Thanks!
Yes, the distribution of income matters, for many things. But does it matter for the AE curve? That’s what’s at issue here. It matters if there are different marginal propensities to consume (or invest) out of different types of income. And there may be (though it’s not there in the simple models I’m talking about, with their simple consumption functions). But that doesn’t affect my point though: If AE falls, income falls because sales and production falls, not because firms lay off workers. Any distribution effects of layoffs would only matter in “second rounds” of the multiplier.
Some profit income goes to institutions, but it’s people who own those institutions. You’re talking about my pension plan!
I think instead that inventories are perhaps not given enough attention in economic teaching. Take the magnitude of the downturn in international trade in the US over this past recession. Several authors have labelled it a “puzzle” and have been busy looking for new explanations as to why it occured, when in fact it and similar episodes in previous recessions can satisfactorily be explained with explicit attention to the dynamics of inventories (due to larger trade frictions in international transactions than domestic ones, etc). It is only because many people have not properly been taught (or have forgotten) about the importance of inventory dynamics that they did not recognise a common pattern and began to look for new explanations. I could go on, for instance on the over-optimism at the beginning of the year regarding the recovery when much of it was in fact being driven by (temporary) inventory adjustments…
Rick: “And as for AE > Y and why this leads firms to produce more (of course in the future): doesn’t this follows from the definition of Y = AE + G + Trade, because terms on the right can be negative, and supply depends on AE rather than Y…”
No! It does not follow from the definition. And that’s not a definition; it’s a semi-equilibrium condition. Suppose initially AE=Y, then AE increases. Will Y increase? This is like supposing that initially Qd=Q, then Qd increases, will Q increase? It depends. If Y is less than Ys, so there’s an excess supply of output, and firms would love to produce and sell more, if only there were more buyers, then yes, firms will increase Y, just as soon as it’s technically feasible to do so. But if Y=Ys initially, then AE increases, all that happens is we get line-ups of customers at the salon, not an increase in haircuts.
Yvan: fair enough. I should have titled my post “Why we should stop talking about inventories when explaining the conceptual distinction between AE and Y, and how AE and Y are brought into equality in the Keynesian Cross model”. But it was a bit too long.
Inventories belong in the investment chapter, under accelerator model.
Because I never starting talking about inventories in macro, I won’t be able to stop. Seriously, this is exactly correct, and the income/jobs point is also a good one
Nick:
You don’t own your pension plan. You have an equitable interest in the income produced by the assets of your plan, but you do not own the underlying assets. You also can’t exercise the rights of ownership of those assets either.
“Did that economist previously teach at University Western Ontario?”
No, but near enough by the standards of Canadian motorists: Waterloo. And in the days when doing econometrics meant begging or buying time on a water-cooled mainframe I suppose the estimation of Engel curves and consumption functions was done by the same people for the most part, so it would be natural to make the connection.
Nick: “Capitalist are people too.”
Per dollar, not very.
I would just point out that Mankiw’s Principles text doesn’t have the Keynesian cross. I’ve always found it confusing and unhelpful for freshmen/sophomores. The Laurence Ball Money & Banking text Mankiw recommends doesn’t either.
In my intro Macro class I brought the incongruousness of talking about inventories in a world where dell didn’t even buy the parts until an order is received. I remember trying to explain to my proffesor that a company like WAL-MART has no inventory beyond what is in a store because of their supply chain expertise. He gave me a weird look and couldn’t explain what inventories mattered that much anymore. Of course my father was in charge of modernizing one of the worlds most complex supply chains (US military) and had gone to various companies to look at how they did it which we often discussed when he got back and was uniquely armed for the discussion.
That was the moment I began to doubt how much intelligence a PhD requires. I eventually came to the conclusion that a doctorate means you are at least knowledgeable, not necessarily a good teacher.
I’ll continue my barrage 🙂
That things are made to order or delivered as a service does not mean that there is not an industrial process behind the service delivery. I think it’s fair to say that a copy is made on demand, but the ability to create the copy takes years of investment.
Take health care which is primarily service based. Nevertheless there is an industry of medical equipment manufacturers, pharmaceutical companies, custom IT vendors, etc. The cost of developing a drug and time involved is enormous, even though the actual pill can be produced on demand at very low per unit cost. A friend of mine worked for a firm that created software for CT scanners, and the development process — including certification — took years. Once a version was qualified, they could make as many copies as they could find customers at little cost.
Or for example, software as a service. Or film — any media, really. In each case, you get a copy on demand, but the ability to create the copy requires a long period of investment. I once worked for an IT vendor whose development cycle lasted about 3 years, and the QA cycle lasted 2 years. That was all fixed costs. Once the product was released, we had low per unit costs since manufacturing was outsourced. We could order as many just in time copies as the customer wanted. I think our gross operating margins were about 60%. The per unit overhead in the form of customer support also scaled favorably.
So I don’t think that haircuts are the appropriate model here. Not unless you include a few years fixed investment before being able to cut hair. And part of the just in time requirements are that prices are negotiated in advance, at least within a range.
Perhaps the better replacement for inventories would be capacity? When there is excess capacity, and fewer people order medical services, then eventually the hospital orders fewer CT scanners, the CT developers invest less in new CT design, the firm that writes the software for the CT scanner invests less in developing their new version, the firm that supplies the authoring tools for the CT scanner author invests less in developing future versions of that tool, etc. Either all these firms are extremely forward looking or capacity is going to adjust very slowly, even though the number of actual copies ordered adjusts quickly. In this sense — viewing capacity as a generalized form of inventory — I think we have less supply flexibility than we did in a less specialized era.