1. Because it helps us understand micro better too. For example, suppose we found that macroeconomic models worked a lot better if we assumed that firms were monopolistically competitive rather than perfectly competitive. Even if you were a microeconomist with no interest in macroeconomics, that tells you something useful. It's one more bit of evidence for monopolistic competition.
2. Because it's there. Even if macroeconomics were of no practical use whatsoever, I would still want to understand it. The macroeconomy is what people do. I want to understand why people do what they do. Just because. It's interesting. And I want to push that understanding down to understanding the interaction between individuals, if I can. You tell me there are emergent properties at the macro level? Neat! Can we understand how emergent properties emerge? (Is money an emergent property?)
Others have made sensible arguments for doing microfoundations. See Noah Smith for some, and for some links to others. I just wanted to add my twopence-worth.
Worthwhile Canadian Initiative 
Determinant: Ferroelectricity is a particular kind of hysteresis. That’s why we have a special name for it. We don’t just call it “hysteresis” because that’s far too general. There are a variety of possible sticky price models, some of which may be similar to ferroelectricity. I haven’t thought about it. But “ferroelectricity” would be a bad name for it. “Sticky” seems good.
Guys! I want to clear up one point on the relation between: involuntary unemployment; wage W; and value marginal product of labour VMP.
Lets define “involuntary unemployment” in the everyday sense of “I can’t get a job but I would be more than willing to work at a wage that people just like me are earning”.
In the simplest competitive labour market model, the height of the labour demand curve (at any point) is VMP. The height of the labour supply curve (at any point) is the Marginal Disutility of Labour MUL (more strictly, the Marginal Rate of Substitution between Leisure and consumption). With no involuntary unemployment, at equilibrium where the S and D curves cross, VMP=W=MUL.
Now lets assume that W is stuck above that equilibrium point, due to stickiness, monopoly power, minimum wage laws, an efficiency wage model, or whatever. Employment gets determined by the quantity of labour demanded at that wage. Quantity of labour demanded is less than quantity of labour supplied. The market is “on” the labour demand curve and “off the labour supply curve. There is involunatry unemployment in the sense defined above.
Here’s the bit you are getting wrong: W is still equal to VMPL. What involuntary unemployment “means” is that W is greater than MUL.
“Involuntary unemployment” in the everyday sense means “I would strongly prefer the utility of a paypacket to the utility of leisure, but I can’t do the swap”. That says W greater than MUL. It says nothing about W and VMPL. The economy is off the labour supply curve. It says nothing about whether the economy is off the labour demand curve.
I hadn’t noticed before, but yes; economists do use the words “sticky” and “hysterisis” in very idiosyncratic ways. But it works for us. There’s a logic to it, even though it’s not obvious.
Suppose wages and prices are perfectly flexible. The economy starts in equilibrium, then gets hit by an earthquake. It takes 10 years for the economy to get back to where it was. We do not use the word “hysterisis” there.
Suppose wages and prices are sticky. The economy starts in equilibrium, then gets hit by an AD shock. It takes 1 year for wages and prices to adjust and become perfectly flexible, but it takes 10 years for the economy to get back to where it was. The effects in year 1 are due to sticky wages and prices. The effects in years 2 to 10 are due to hysterisis.
Weird, huh?
K: “Determinant: Ferroelectricity is a particular kind of hysteresis. That’s why we have a special name for it. We don’t just call it “hysteresis” because that’s far too general. There are a variety of possible sticky price models, some of which may be similar to ferroelectricity. I haven’t thought about it. But “ferroelectricity” would be a bad name for it. “Sticky” seems good.”
Aha! For us, “stickiness” is a very particular kind of slowness to adjust to shocks. And “hysterisis” is a slowness to adjust to a very particular kind of shock due to stickiness.
Only prices (and wages etc) can be “sticky”. And “hysterisis” means slowness to adjust to the consequences of sticky prices+shock even after the stickiness of prices has gone away.
Nick: Sorry for the slow responses….been working.
“…your questions would imply that all models are useless. We just look at the data, and stop there. But we don’t do that. We try to fit patterns to the data. … We place restrictions on what patterns we try to fit. Microfoundations are one of those restrictions. ”
So, you agreed that microfoundations place no real restrictions on the data or the patterns that we see….and then argue that microfoundations are useful restrictions. Obviously, I’m confused….but even more broadly, if you think that (1) we need restrictions, and (2) the micro approach gives us some interesting ones, I don’t think your argument is done. I thought you were arguing that microfoundations give essential restrictions — that any approach without it is inferior. (Perhaps I’m reading too much into what you wrote…but what I’m describing is certaintly a popular view in macro.) Your logic seems to fall well short of that. What am I missing?
I don’t teach Arrow. I leave that to the micro people. But then a determined insistence on Arrow would rule out almost all policy recommendations (there’s always someone who wants the world to end!). You can argue that good monetary policy improves the welfare of the representative agent, then if someone wants to argue that it might make distribution worse, you think about that too, and so on.
“If we only accepted perfect models and ironclad policy recommendations,…some other idiot would start making even less perfect models and even less ironclad policy recommendations. We are going to believe something and we are going to do something.”
There’s an old episode of Yes, Minister/Yes, Prime Minister where someone’s logic is reduced to
1) We must do something.
2) This is something.
3) Therefore, we must do this.
Your argument reminds me of that. There’s many arbitrary ways to generate policy recommendations; microfoundations are one of those ways. But this does not explain why we have to foresake all other illogical or unreliable methods for this one.
(And don’t get me started on why we are keen to generate policy advice from models that we admit may easily be giving the wrong advice.)
Sorry….shouldn’t have included that paragraph about Arrow in the above (that was from your earlier post.)
Simon: “What am I missing?”
Nuance. 😉 The word between yes and no. The idea that (nearly) all evidence, arguments, and restrictions, are tentative and provisional, rather than watertight. Other things equal, a model that also makes sense at the micro level is more credible than one that doesn’t, or one that we don’t know whether it makes sense at the micro level or not. We are trying to fit a Quinean web of belief onto a lot of “facts” that we don’t even know are facts. Other things equal, the wider that net, the better we can believe it makes sense of the world.
Let me put it this way instead. The wider the data set that is encompassed by the model, the better the model, other things equal. A model that explains both X and Y is better than a model that explains X and says nothing about Y. Why should we draw a line between macro Y and micro Y?
Yes, I’m not as hardline as those who say that only microfounded macro models should be accepted. We should never be satisfied with a non-microfounded model, but we can’t always get what we want.
Nick,
I found the paper that describes the labour equilibrium I was thinking about when I wrote the above comment. Take two monopolistic producers of two imperfectly substitutable goods, diminishing revenues from output above a threshold, a Walrasian labour market, and workers with zero disutility of labour; incredibly you can still get unemployment at any positive wage, and the rest of the workforce employed below the VMPL, a nasty GE result indeed. Is it realistic? It doesn’t seem absurd to me that we have strongly declining marginal utility for the current consumption basket. Maybe we are short on stuff we want to buy more of. This is obviously the important assumption as it makes it pretty easy to imagine that the profit optimizing output might involve a limited quantity of labour.
Are the assumptions otherwise inconsistent with real world data, such as income distributions? I haven’t thought about it enough – the paper doesn’t address the details of how Say’s law ends up balanced. So perhaps it implies some particular unrealistic implications for distribution of capital ownership and government redistribution. Worth thinking about though.
On hysteresis:
“Hysteresis” just means memory. The state of the system depends not just on exogenous variables, but on previous values of the state variables. Of course, any system with continuously changing state variables (e.g. mechanical inertia) could be described as falling in this category, so the name is usually used when there are closely related systems that don’t exhibit this property in the same variable (paramagnets have it, diamagnets don’t). Prices that can’t just change to the Walrasian equilibrium in response to whatever exogenous changes in supply/demand are definitely an example of hysteresis in the price variable. The previous level of the price is an important state variable in the system. My only point was that “hysteresis” is far too general a concept to use as a name for “sticky price processes.” Solid state physics is full of examples of materials with “hysteresis”. Many of them are special in some particular way, so physicists give them special names. Economists should do the same, and, of course, they do.
Thanks K. I have skimmed it. I am annoyed they didn’t reference my 1987 paper which showed exactly the same thing, in a fully macro model! Instead they reference Weitzmann 1982, which is wrong because he forgets to include profits in the budget constraint! Ah well.
Here’s the jist. If we replace perfect competition with imperfect, the height of the labour demand curve becomes Marginal Revenue Product rather than Value Marginal Product.
VMPL=P.MPL
MRP=MR.MPL= (1-(1/e)).P.MPL where e is elasticity of demand and MR is Marginal Revenue.
So, with imperfect competition, W will be less than VMP.
Imperfect competition (generally) makes the demand (curve) for labour lower than it otherwise would be, and equilibrium wages and employment lower than they otherwise would be. It doesn’t really get you involuntary unemployment, contrary to the assertion in that paper (OK, I didn’t actually read it). In their example (like in my 1987 model) it does push equilibrium wages down to zero. But that’s only because they (and I) assumed a very clunky reverse-L-shaped labour supply curve. With W=O, nobody who is not working prefers working. That’s not like real world involuntary unemployment, where some people are working at W, and others (just like them) want to work at W and can’t get a job.
Even more annoying, IIRC, the QJE turned down my paper, even though it was better than that one, and earlier!!
Nick,
“I am annoyed they didn’t reference my 1987 paper which showed exactly the same thing”
Sorry for mentioning it…
“[Weitzmann] forgets to include profits in the budget constraint”
…that guy’s such a dufus…
“the QJE turned down my paper”
…son of a #%@$!!! 🙂
” the height of the labour demand curve becomes Marginal Revenue Product rather than Value Marginal Product.”
That makes sense. Thanks. Do you have a link to your paper? I’d love to read it.
“that’s only because they (and I) assumed a very clunky reverse-L-shaped labour supply curve”
That labour supply curve may not be *that crazy. If we assume that those people earning that low equilibrium wage own little capital (and need a roof and some food), the labour supply curve could be downward sloping at the bottom. Vertical might be a conservative assumption. Some kind of indentured servitude equilibrium? Foxconn?
@Nick: you’re right. i didn’t talk about marginal productivity theory, only ways of getting involuntary unemployment
I think I’ve just come to understand consumer debt as a way to make labour supply curves slope downwards. Debt is like food or housing, but unlike basic goods it ratchets upwards to whatever your wage level might be. Then, when negative wage/income shocks hit you must work more in order to service the debt. I’m sure this was obvious to lots of people, but seeing it as a negatively sloped labour supply is new to me. Humans are such a bunch of suckers!
K: thanks, but my old paper was quite good 25 years ago, but not now. My post here says the same thing more cleanly. Just pretend the MC curves are labour supply curves, the demand curves are VMP curves, the MR curves are MRP curves, replace Y with L, Pi/P with W/P, and you have the model.
“Other things equal, a model that also makes sense at the micro level is more credible than one that doesn’t, or one that we don’t know whether it makes sense at the micro level or not.” Let’s leave aside arguments about “credibility.” The point made by many outside of macroeconomics is that what macroeconomists call “micro-foundations” are not, in fact, supported by microeconomics. Microeconomists do not assume that social welfare can be judged by the welfare of one individual (“representative” or not.) Microeconomists do not think that aggregate levels of production, in a world with diminishing marginal returns to factors, will be independent of how those factors are distributed. Macroeconomists who start to look at micro data (e.g. on labour flows of individuals or firm-level productivity) often find that much of what matters at the aggregate level is not well-captured in a representative agent framework. So I’d agree that models with good micro-foundations are more credible…..I just don’t see macroeconomists use or teach them.
“Let me put it this way instead. The wider the data set that is encompassed by the model, the better the model, other things equal.” It almost sounds like you’re saying that the best model is one that can never be empirically falsified, but I don’t think that’s what you have in mind.