You can follow this conversation by subscribing to the comment feed for this post.

Nick,

I have a question about the Phillips Curve.

When Fisher discovered the macro correlation between inflation and output, he assumed it went from inflation to output. When Keynesians discovered the Phillips curve, they assumed it went from output to inflation. Why did they assume that one causes the other? Why not just say that changes in AD cause both simultaneously.

In micro we don't say that a change a change in demand for apples causes prices to change and then THAT causes output to changes, or vice versa. We just say that demand causes both to change simultaneously. Yet, in macro we change the rules. Why?

Macro models always say that how much changes in AD affect output by "expected versus actual INFLATION." But thats not how we do it in micro. First, if we directly took it from micro we would say "expected versus actual MV or AD." We would say that it is demand, MV or AD, that changes both prices and output. NOT that prices changes output. MV and AD are for macro what the demand for apples is in micro.

Second, why the "expected" part. In micro we do not say that changes in the demand for apples cause changes in prices/output because of actual versus expected expectations. Expectations have nothing to do with it. We just say that demand changes them automatically.

I've never heard anyone bring this up, yet it immediately occurred to me when I was going through Mankiw's textbook.

In fact, let me be even more radical. Why not just skip stickiness, the Phillips curve and expectations all together. Just say that changes in AD affect both prices and output JUST LIKE the micro demand for apples. For example, if the CB hands me a billion dollars, I go to the grocery store and purchase apples. What we learned in micro courses tells us that prices and output will go up. No macro gimmickry is needed. That's the same story of a macro change of AD, but I did it without stickiness, the Phillips curve and expectations. I don't understand why micro 101 isn't good enough for macro models.

Also, it just occurred to me that it is strange that micro courses don't discuss stickiness......

Sorry for the rambling.

Here something along those lines.

I remember someone (Andalfatto?) writing that he didn't understand why changes in inflation would causes changes in spending. It then occurred to me that the micro effects of expectations are the same as those we teach in micro.

When you explain the demand curve in econ 101 you give a list of effects that shift the curve. One of those is expectations of prices and output. For example, expectations of the price of apples falling in the future, will cause the demand for apples to fall NOW. But that's the exact same thing as inflation expectations! If people expect the price of money to be lower in the future, they will attempt to get rid of it now. This increases spending.

Yet, I have never ever heard anyone explain expectations from this basic micro 101 form. Everyone always writes these huge blog posts, and articles and papers trying to explain the macro effects of inflation uses sophisticated macro concepts, when its really just basic micro 101.

A demand for goods is not a demand for labor, and, economist tell us that technological innovation if the key to economic growth, technological innovation that increases output and very often reduce the input of labor.

So real increases in growth and increases in aggregate demand very often DO NOT go hand in hand with increases in employment, as we have seen many times.

Greg: this may or may not be true. There are actually many ways how real GDP may grow without any significant technology innovation. Actually most of the human history is about growth in the absence of innovation. If African nations would start to actually implement the technologies already potentially available to them, the world would live through a period of real GDP growth even if nothing new gets invented. This may actually be the very example of Nick's self-fulfilling prophecies - if investors start to believe that Africa is going to be next Asia they may start investing there which may result in these expectations to cause themselves.

Greg:

This isn't right for the entire economy. What happens in micro is that the improved technoglogy allows more of a particular good to be produced, but because people would rather spent the extra real income this generates on other goods as well, labor and other resources are freed to produce those other goods.

In aggregate, this doesn't happen. Total output and real income both rise. Now, if people want more leisure, because of the greater real income, then less labor will be used.

And it is possible that the change in income will shift the functional distribution of income away from labor to capital. But it could go the other way.

Joe Mac:

I would tell the simple micro story like--shifts in demand lead to shortages or surpluses. Those lead to changes in prices. The changes in prices lead to changes in quantities supplied and demanded.

And so, changes in prices cause the changes in production. They also dampen and partly offset the initial shift in demand. Or rather, the change in quantity demand partially offsets the change in desired purchases implied by the initial shift in demand.

On the other hand, I don't necessarily treat that story as realistic. The way I see it, in the typical market entrepreneurs simultaneously set prices and determine production based upon expectations of demand and their costs, (which is what is represented by supply.)

But the simple story is that quantity supplied depends on price. So price changes, and then production shifts with quantity supplied.

Excess supply of money. Shortages of goods. Prices rise to close the shortages. Firms responde by producing more. Of course, because the prices of all alternative products rose, cost should rise, and so supply should decrease in every market too, which should result in less production, just offseting the increase in quantity supplied, leaving output the same and prices higher.

And this is the phillips curve, short run aggregate supply curve puzzle. Why aren't they vertical.

On the other hand, I agree completely with Rowe's account as a representation of reality. Leaving aside the (correct) account of expectations, what happens is spending increases or grows more rapidly, and firms respond to the more rapid sales by raising prices and production both. More price increase, less production increase and vice versal.

Yeager too that position.

I actually think Matt has it completely backward: holding the interest rate constant (e.g. at zero), inflation (expected inflation, that is) does create jobs. When you tax non-job-creating assets (money) by means of inflation, it creates an incentive to invest in job-creating assets. It's a non-linear story, in the sense that the time sequence is reversed, but it is the expectation of inflation that is the exogenous causal factor.

JoeMac: "When you explain the demand curve in econ 101 you give a list of effects that shift the curve. One of those is expectations of prices and output. For example, expectations of the price of apples falling in the future, will cause the demand for apples to fall NOW. But that's the exact same thing as inflation expectations!"

Bingo! It's the same thing. Except it's not the exact same thing. In micro, we are talking about the expected relative price of apples, in terms of other goods. In macro, we are talking about the expected price of everything, in terms of money. You gotta watch out for fallacies of composition.

"I've never heard anyone bring this up, yet it immediately occurred to me when I was going through Mankiw's textbook."

Bingo again! It is indeed very rare for macroeconomists to talk about this. But I remember David Laidler talking about this in my MA macro class. (And I had a bit of a discussion of it with Scott and Bill, IIRC, a couple of years back, somewhere).

There's an AD curve, and there's that other curve, for the short run, that I'm going to call "the other curve". What is that other curve? Like you, Laidler distinguished between the Fisher story, and the Phelps story, of that other curve. Lucas followed the Fisher story. Keynesians tend to follow the Phelps story. Friedman could be read both ways.

In the Fisher story, it's a supply curve, of some sort. And we want to tell a supply curve story in terms of an increase in price causing an increase in quantity. But that isn't really quite right. (And Marshall himself interpreted supply and demand curves the opposite way round from what we do now, which is why we draw the supply and demand curve diagram with the axes the wrong way round, which always annoys the engineers). Because an increase in price causes an increase in quantity supplied, which is conceptually distinct from an increase in quantity sold (as illustrated for example by what happens when you raise price by law when there's a binding price floor, when you get an increase in quantity supplied but a fall in quantity sold). Strictly, in the Fisher story, it's simultaneous causation.

The Phelps story interprets the "other curve" as a disequilibrium price adjustment story. And we want to tell that causal story the other way round. An increase in AD causes an increase in Qd at the existing P, which causes excess demand, which causes P to start rising. But even here it's not quite so clear-cut. Take the Calvo story, for example, which is at the root of modern New Keynesian models. Every period, a random fraction f of all firms are allowed to change price. If AD increases, that fraction f of the firms increase their prices in response (and those firms may either increase or decrease Q, it depends, but they will usually decrease Q). The remaining 1-f of the firms can't change price, so their Q increases. So again, strictly it's simultaneous causation.

But you can imagine a world in which all firms set prices one period in advance, in which case an unexpected increase in AD causes Q to increase first and P to increase later. Though even here I would object to saying it's that initial increase in Q that causes the later increase in P. Rather, it's the increase in AD that causes both. Because if firms were unable to increase Q immediately (because it takes time to crank up production), they would still increase P later, provided they think that AD won't go back down again.

And yes, I skated right over all this in my post.

I basically agree with Bill, I think.

"I don't understand why micro 101 isn't good enough for macro models."

Because the supply curve of apples in micro slopes up because an increase in the relative price of apples causes an increase in quantity supplied, as resources move away from producing bananas into producing apples. If the prices of apples and bananas and everything else all go up (by the same %) why would any producer want to do anything different? If the macro supply curve slopes up, rather than being vertical, the reason must be very different from the reason why the micro supply curve slopes up.

"Also, it just occurred to me that it is strange that micro courses don't discuss stickiness......"

Yep. They duck the topic. Macroeconomists can't duck it, because it's hard to explain why increases in AD cause increases in Q without talking about stickiness.

I actually think Matt has it completely backward: holding the interest rate constant (e.g. at zero), inflation (expected inflation, that is) does create jobs.

Holding which interest rate constant, and for how long?

Thank you for your responses. Two quick questions.

Bill,

When you say, "shifts in demand lead to shortages or surpluses. Those lead to changes in prices. The changes in prices lead to changes in quantities supplied and demanded."

Is it possible that a shortage or surplus will first lead to a change in production and THEN a change in prices. Or, is it always prices that changes first.

Then, you say, "Excess supply of money. Shortages of goods. Prices rise to close the shortages. Firms responde by producing more."

If so, then doesn't that contradict stickiness?

Nick,

What if we were to treat the AS curve as one long upward sloping curve that eventually becomes perfectly inelastic at the right. The elasticity of the AS curve would be determined by how much of all available resources (land, labor, capital) are being used for production. At full output all resources are being used so an increase in AD just causes all markets to compete for the same limited stock of resources. Therefore, all you get is increase in general prices. This would not require Friedman's story of expectations from is presidential address.

In a recession, the AS curve would be much more elastic and therefore an increase in AD would cause limited increase in prices relative to output within individual markets. No supplier could raise prices much in response to demand because they would easily get outbid by competitors who can easily bring in new cheap resources at the margin. But as more resources become used the AS curve becomes more inelastic and firms can more easily raise prices.

I stole this story from Armen Alchian's textbook. He has an appendix in his textbook where he explains how changes in the money supply affect prices and output, but he does it without stickiness, expectations, or the Phillips curve.

JoeMac: Start at (say) full employment (however you define that). Assume the AD curve slopes down (to keep it simple). Now suppose the AD curve shifts left. Does that cause a drop in employment? If so, why don't prices of all resources instantly fall, and keep on falling, moving down along the AD curve, until you are back at full employment? The standard story is: because prices are sticky, so it takes time for them to adjust.

Nick, I don't think you really mean to say that the causal nexus between inflation and jobs is a matter of "simultaneous causation." The story you tell contains a great deal of complicated non-simultaneity.

I also don't understand what role non-linearity is playing in your account. Where does appeal to a non-linear functional relationship between causal factors and effects come into play? Feedback does not require nonlinearity.

I also don't think you help yourself when you say "an increase in expected inflation causes an increase in the aggregate demand curve." Curves are representational devices, not events or conditions. What actual conditions, events, or behaviors do these curves describe? How are these real non-representational phenomena affected by changes in expectations among various types of economic agents?

Dan: "Nick, I don't think you really mean to say..."

Yes I do mean to say that.

"I also don't understand what role non-linearity is playing in your account."

"Non-linearity" in the economist's/sciency sense, i.e. "Y=a+bX^2 is a non-linear functional relationship", plays no role in my story. I'm using the word in what I call the "Artsie" sense, hence my very first line in this post, and the embedded link.

"Curves are representational devices, not events or conditions. What actual conditions, events, or behaviors do these curves describe? How are these real non-representational phenomena affected by changes in expectations among various types of economic agents?"

??? Economists have used curves for hundreds of years. We shift them, and look at what happens to the points where the curves cross. Those curves represent conditions. When they shift, and the points where they cross move, those represent events. Must we stop using curves?

Should be, "my *second* line in the post".

When Nick says "non-linear", translate that to "solution to an equation of several variables". There isn't any need to get nit-picky and point out that such solutions can be linear or non-linear.

Others will say -- how does the economy arrive at such a solution. What are the forces acting on it from which you derive the laws of motion. And previously Nick has argued that knowing these laws or questioning whether the forces actually exist isn't important, although in other blog posts he talked approvingly of disequilibrium economics. I suspect that there is an element of being an agent provocateur.

Am I right?

This post relates to a perception that I have encountered among opponents of monetary stimulus, mainly that it is somehow different from fiscal stimulus in terms of the "mix" of output and inflation that each generates. I started to try to use what I thought was the standard AD-AS model to explain to them that the type of stimulus should not matter and then discovered almost none of them knew what I was talking about. This got me thinking, who uses the AD-AS model?

Well, as many people know it was created by British economist John Hicks and is based on the theory of Keynes as presented in The General Theory. It is used by a broad array of economists, including New Keynesians, Post Keynesians, Monetarists, New Classical, Neo-Keynesians and Market Monetarists.

It is of course rejected by the Austrians. But my general impression is that most, but not all supporters of Neo-Marxism, Sraffrian and Modern Monetary Theory economics also are either not familiar with the model or do not believe in it. I find this somewhat ironic as many of these people identify themselves as "Keynesians" and yet it is the Monetarists that are actually using a model based on Keynesian theory.

So, I guess my question is, why do supporters of MMT think the AS/AD model is either unimportant or invalid?

And I have an ancilliary question directed specifically at Nick. Can you explain to me what the following post really means? I've read it multiple times and still cannot grasp what the MMT people are thinking in IS/LM terms:

Mark: "So, I guess my question is, why do supporters of MMT think the AS/AD model is either unimportant or invalid?"

I think we maybe need to make the distinction (which you may be making above) between the *supporters* (i.e. internet followers) of MMT and the main MMT people themselves.

I suspect that the MMT supporters don't understand AD/AS. More importantly, I suspect they might not understand even micro demand and supply. More generally, they don't seem at all comfortable with simultaneity (and the curves that represent simultaneity). See Dan's comment immediately above, for example, which I found really surprising. They really want explanations of the form: "event A causes event B, which causes event C" in a temporal and causal sequence. Simultaneity sounds like magic.

And this ties in directly with my answer to your ancillary question:

"Can you explain to me what the following post really means? I've read it multiple times and still cannot grasp what the MMT people are thinking in IS/LM terms:.."

Take a standard second year textbook ISLM/AD/AS model.

1. Assume the interest-elasticity of investment and saving is zero, so the IS curve becomes vertical. Monetary policy now has no effect on AD, and fiscal policy has a big effect on AD. Plus, the AD curve becomes vertical, because changes in M/P due to changes in P, even if they shift the LM curve, don't affect the level of Y at the ISLM intersection. Plus, you cannot define a natural rate of interest such that desired S=desired I at "full employment" output.

2. Now assume an infinite interest-elastic demand and/or supply of money, so the LM curve becomes horizontal. IS curve shift have no effect on interest rates. Plus, this would make the AD curve vertical even if it weren't already vertical due to the vertical IS.

3. Forget the distinction between LRAS and SRAS. (Or between the LR and SR Phillips Curve). There's just one AS curve (or Phillips Curve) for both the SR and LR. It is horizontal for Y below "full employment", and vertical at full employment. Shifts in AD affect only Y and not P when we are on the horizontal section, and only affect P not Y when we are at full employment. At less than full employment, all inflation is cost-push.

If you don't like simultaneity, you will really like this model, because it contains no simultaneity (OK, there is still simultaneity with the old first-year Keynesian multiplier inside the IS curve, I expect.) IS determines Y. LM determines r. Cost push determines P. (This assumes we are below full employment, but they think we always are, because there are always some people unemployed.)

Thanks. That was crystal clear.

My only problem now is how could a person believe such things? Even if one has an internally consistent model that displays such characteristics, it's contrary to the empirical evidence.

Nick, my objection has nothing to do with being comfortable or not with simultaneity. It is an objection to employing models that posit simultaneous causal interactions where, as a matter of empirical fact, no such interactions occur.

In the models of classical Newtonian gravitational theory, for example, one can exhibit simultaneous causal interactions. Suppose, for example, one posits a two body system in which the mass of body A is some constant M1 for all times up to and including T0, and then for each time T after T0, the mass of body A is K*M1*(T-T0), for some constant K. So, in other words, suppose we are modeling a somewhat fanciful two-body world in which the mass of one of the bodies increases continuously throughout the interval [T0, ...] By the laws of Newtonian gravitation, and assuming the mass of B is a constant, the force exerted on body B by body A, and consequently the acceleration of B, will change continuously as well. And the change will occur from time T0 forward, with no temporal lag. The strength of the force will vary with the distance between A and B. But in this little Newtonian world, the distance between the bodies has nothing to do with the time it takes for changes in the force on B to be propagated by changes in the mass of A. The change is propagated instantaneously.

Now I would submit that there is no phenomenon in the social world of human beings and human institutions, the world economic science purports to describe, that is accurately described by laws permitting simultaneous causal interactions of this type. The causal relationships of interest among human beings, those of economic interest, involve the propagation of change from cause to effect with time lags that so significant to the policy maker or analyst that models that suppress these time lags can be no more than crude approximations of minimal utility.

Dan: to argue that simultaneous causation is OK in Newtonian mechanics when modelling the interaction between physical bodies, but not OK when modelling the interaction between humans, is a philosophically strange position. If anything, I would say that because humans have expectations, can contemplate counterfactual conditionals, and may be aware of their interaction, simultaneous causation is even more applicable to human bodies than to non-human bodies.

Take for example Nash Equilibrium. How people behave in Nash equilibrium is almost always (except maybe in degenerate cases like Prisoners' Dilemma where there is a single dominant strategy) a case of simultaneous causation.

The fact that there may be lags in human interaction (just as there may be lags in mechanical systems) does not mean there is no simultaneous causation. For example, in a multiperiod game the Nash Equilibrium is a pair of conditional time-paths of actions, that must be solved simultaneously, and solved backwards in time from the end of the game back to the beginning. Since we are talking about purposive behaviour, and not billard balls, you can't solve forwards. Human behaviour is end-oriented.

But if you do reject simultaneity for human interaction, you are rejecting almost all of economics for the last few hundred years. I know of only one economist who deliberately set up a non-simultaneous model -- and he was an econometrician (forgotten his name) who only did it because, in his day, econometric techniques just couldn't handle simultaneous equation estimation (we can now). The very basic supply and demand model is simultaneous. Simultaneity is what we do.

I would say that physics has moved on since Newtonian mechanics. We no longer believe in action at a distance. There must be a force carrying particle that is exchanged. And the exchange of that particle does not happen *instantly*, there is a speed limit, namely the speed of light. Moreover physics has also realized that information is costly, so that the egg doesn't re-assemble itself and jump up into the air.

I was just reading Leijonhufvud's essay on Keynes:

The only thing which Keynes "removed" from the foundations of
classical theory was the deus ex machina - the auctioneer which is as-
sumed to furnish, without charge, all the information needed to obtain
the perfect coordination of the activities of all traders in the present
and through the future.

Which, then, is the more "general theory" and which the "special
case"? Must one not grant Keynes his claim to having tackled the
more general problem?

Walras' model, it has often been noted, was patterned on Newtonian
mechanics. On the latter, Norbert Wiener once commented: "Here
there emerges a very interesting distinction between the physics of our
grandfathers and that of the present day. In nineteenth century phys-
ics, it seemed to cost nothing to get information" [14, p. 29]. In con-
text, the statement refers to Maxwell's Demon-not, of course, to
Walras' auctioneer. But, mutatis mutandis, it would have served admi-
rably as a motto for Keynes's work. It has not been the main theme of
Keynesian economics.'

I would suggest that "sticky" prices, or sticky information is a crude attempt
to bolt onto the utopian model some time-delays. Models of search are another way of dispensing with simultaneous solution of supply and demand curves. You have to do that in order to get anything sensible. And doing that is hard, not easy. Solving simultaneous equations and finding optimal paths was pretty much solved hundreds of years ago. It may have taken economists a long time to learn how to do it, but engineers were doing it quite well in the 19th century.

But dealing with interactive systems is a wide open problem. It is much harder, and mathematical intractability is the only reason you don't do it, not because there is something intrinsic to economics that makes it immune from causality, informational costs, processing times, and reaction times.

rsj: "Models of search are another way of dispensing with simultaneous solution of supply and demand curves."

All search models I can remember seeing are simultaneous. There's the sellers' reaction function (aka the "supply" function), and the buyers' reaction function (aka the "demand" function). And we solve the two simultaneously to determine the equilibrium number of matches, and everything else.

They violate the simultaneous solution of supply and demand curves because they violate the law of one price. Different workers will have different reservation wages based on their own knowledge of the offer market (which is a stochastic learning problem).

? Price discrimination models violate the law of one price. That doesn't mean they are not simultaneous models.

Perhaps we are talking past each other. What would a non-simultaneous model be?

For me, the issue is in drawing a supply curve and a demand curve and assuming that the price and quantity transacted are the same as at the intersection of the two curves.

I want to know how the economy arrives at the ideal solution. If there is no auctioneer, and neither side "knows" the curve of the other, then how do they agree to transact at the optimum levels?

Dan: to argue that simultaneous causation is OK in Newtonian mechanics when modelling the interaction between physical bodies, but not OK when modelling the interaction between humans, is a philosophically strange position. If anything, I would say that because humans have expectations, can contemplate counterfactual conditionals, and may be aware of their interaction, simultaneous causation is even more applicable to human bodies than to non-human bodies.

Sorry Nick, but I can't begin to understand this attitude. Changes in psychological expectations have causes, and they also have effects. But the causes of changes in expectations always precede those changes, especially when we are talking about expectations as large macro-phenomena. Likewise the effects of changes in expectations are subsequent to the change. Almost every cause and effect relationship in the world of human phenomena - including economic phenomena phenomena - is like this.

For example, if you tell me that a statement by the Fed chairman on Monday will cause, over the next several days or weeks, a change in inflation expectations, then temporal precedence and asymmetry is obviously part of your explanation. If you then tell me that the change in inflation expectations taking place over the next week will cause a change in spending behavior, temporal asymmetry is also part of that behavior. Nobody ever buys anything as a result of a change in expectations that occurred at the very instant that their purchase occurred. And certainly whole societies whose expectations are represented by some aggregate function are even less plausibly modeled that way.

The problem is that economists often approach problems, and generate predictions and explanations, with a mathematized contraption that is only 1/10th of their actual theory, but which they strangely regard as constituting the bulk of the theory. They trot out the curves and formulas as crutches, but then proceed to tell some additional and much more complex causal story, based on anecdotal or statistical empirical understanding of complex and temporally oriented social processes and events. Their theory is actually the more detailed and causally articulated narrative they are telling, and the functions and diagrams only play a limited and incomplete role as props in that story.

For example, with policy recommendations based on AS-AD analysis, there is always some story told about possible decisions by policy-makers that are supposed to establish the causally sufficient preconditions for some complex and temporally extended social phenomenon represented mathematically by something like "the demand curve moving to the right". All of the real action in the explanation, and debate over that explanation, is involved with that causal story. What we really want to know is whether and how the suggested real-world policy action would have the effect of "moving the curve". In other words, the real core of the explanation lies in intuitive verbal explanations that are exogenous and antecedent to the phenomena represented by the mathematical contraption itself.

Part of the reason I have found market monetarism so unpersuasive is that the MMers keep retreating, when challenged, to textbook models of the role of the central bank as monetary authority, and these models simply adopt as postulates the claims about which the MMers are being challenged. Meanwhile, the seat-of-the-pants causal narratives keep changing. Lately they are turning into philosophical dodges about the world being too holistic and non-linear and feedbacky to even explain how the recommended policy decisions will have the effects their proponents claim they will.

I also don't understand the point of appealing a self-contained model of a static equilibrium when the issues in front of us are about proposed policy changes that by their nature are designed to disrupt some current equilibrium and generate a new one. And even if the point were just to explain the persistence of an equilibrium, we are not talking about genuinely simultaneous causation. If expectations on Monday morning lean to aggregate behavior on Monday afternoon leading to newspaper and other media reports overnight leading to the reinforcement of opinions and persistence of Monday's expectations on Tuesday, that is a complex temporally extended process with a temporally asymmetric direction of causal propagation. Even if the expectations on Tuesday are qualitatively identical to the expectations on Monday, they are not the same event.

Nick

I've been reading Leijonhufvud for the past couple of days, and he seems to contest your emphasis on simultaneity. He says that simultaneity is the method of statics, a Walrasian comparative statics. While any analysis of economic dynamics in the Marshallian tradition which he prefers involves sequential economic decisions, where there may be a quasi-linear chain of causation.

He tends to view simultaneous causation as either a partial equilibrium result, or true in general equilibrium only under the assumption of perfect information and rational expectations. Imperfect information general disequilibrium has a sequence of events.

http://www.econ.ucla.edu/workingpapers/wp186.pdf

And many others along similar lines.

The comments to this entry are closed.

• WWW