I like macroeconomics with monopolistic competition rather than perfect competition. It helps me make sense of the world. (Most modern macro models assume monopolistic competition.)
I like Milton Friedman's "Plucking Model". It seems to work well empirically.
But the two seem to me to be in conflict. I can't consistently believe both.
Let me start with three pictures of business cycles.
(They should be curvy rather than spiky, but I'm not good enough with Paint.)
The first picture is the one I normally draw when teaching intro macro. There's a reasonably smooth trend line for Real GDP (or some other measure of real economic activity). But actual RGDP fluctuates around that trend line. There are both booms (when it's above trend) and recessions (when it's below trend). It's symmetric. Some booms are big and some are small. Some recessions are big and some are small.
The second picture doesn't have any booms. It only has recessions. It's asymmetric. RGDP can be below trend, but it is never above trend. Some recessions are big and some are small.
The third picture (just for theoretical completeness) doesn't have any recessions. It only has booms. RGDP can be above trend, but cannot be below trend. It's asymmetric in the other direction. Some booms are big and some are small.
The first question you ought to ask is: "How can we tell the difference between those three pictures, if we don't actually see that trend line?"
If all booms and recessions were exactly the same size, we wouldn't be able to tell the difference. A trend line across the peaks (picture 3); a trend line across the troughs (picture 2) and a trend line through the midpoints between peaks and troughs (picture 1); would all fit the data equally well.
But if booms and recessions were all different sizes, we would be able to tell the difference. A trend line across the peaks works best in picture 2; a trend line across the troughs works best in picture 3; and a trend line through the midpoints works best in picture 1.
But there's a better way of telling the difference between those three pictures. It's what Milton Friedman used.
Imagine you started with a perfectly flat garden. In 1 you dig random sized holes, then use the soil you dug out to create random sized mounds. In 2 you dig random sized holes, and sell the soil you dug out. In 3 you buy some extra soil, and dump it in random sized mounds. If an ant were crawling across the surface of your garden, how could the ant tell the difference between 1, 2,and 3?
In 2, the ant would notice a strong correlation between how far he descended from peak to trough and how far he subsequently ascended from that trough to the next peak. But the ant would notice no correlation between how far he ascended from trough to peak and how far he subsequently descended from peak to trough. Big descents are always followed by big ascents, but big ascents aren't always followed by big descents.
In 3, the ant would notice a strong correlation between how far he ascended from trough to peak and how far he subsequently descended from that peak to the next trough. But no correlation between how far he descended from peak to trough and how far he subsequently ascended from trough to peak. Big ascents are always followed by big descents, but big descents aren't always followed by big ascents.
In 1, the ant would notice roughly the same weak correlation between descents and subsequent ascents as there is between ascents and subsequent descents.
Empirically, the business cycle in the real world looks closer to picture 2 than picture 1. More accurately, it's somewhere between pictures 2 and 1. Picture 3 doesn't work at all. (Which also means the more extreme Austrians, who say the seeds of the recession are always sown in the preceding boom, seem to be wrong.)
To use Milton Friedman's metaphor, it's like a string stretched tight along a board. You can pluck the string away from the board but it returns to the board when you let go and can't go past the board. Or to use Scott Sumner's metaphor, it's like valleys in a high plateau. What look like mountains are really just places where there is no valley. What look like booms are really just times when there is no recession.
Now look at the very simple macro model in picture 1m.
On average the economy is on the vertical LRAS curve. If there is a sudden unexpected fall in AD, the economy goes into recession to the left of the LRAS curve. If there is a sudden unexpected rise in AD, the economy goes into a boom to the right of the LRAS curve. Model 1m gives us picture 1. It's symmetric.
Now look at the very simple macro model in picture 2m. It works just the same as 1m if AD falls to produce a recession. But the SRAS curve either stops dead when it hits the LRAS curve, or else turns vertical when it hits the LRAS curve. So if there's a sudden unexpected rise in AD, there is no boom. Either prices fail to rise enough, and so there's excess demand for goods (when the SRAS curve stops dead), or else prices instantly rise so there's inflation but no excess demand or boom (when the SRAS curve turns vertical). It's asymmetric.
Perfect competition leads to a model like 2m. In long run equilibrium, when prices and wages have had long enough to adjust, firms and workers are selling as much output and labour as they want to at those prices and wages. You can't force them to sell more output and labour than they want to, even if demand does increase and prices or wages are sticky. You can't create a boom.
Monopolistic competition leads to a model like 1m. In long run equilibrium, when prices and wages have had long enough to adjust, firms and workers are not selling as much output and labour as they want to at those prices and wages. Because those prices and wages are above Marginal Costs. But individual firms (or workers) would have to cut prices (or wages) to sell more output (or labour), and they don't want to cut prices (or wages), because Marginal Revenue equals Marginal Cost. But if aggregate Demand increased, and if prices or wages were sticky, firms and workers would sell more output and labour. So we get a boom.
So how do I reconcile monopolistic competition and the plucking model?
Perhaps this is the answer: if monopolistically competitive firms are hit with a small positive demand shock, and their prices are sticky, they will increase output to satisfy demand. But if monopolistically competitive firms are hit with a large positive demand shock, and their prices are sticky, they will not increase output to satisfy demand, because doing so means they would go past the point at which price equals marginal cost. Small booms are possible, but large booms are impossible. But both small and large recessions are possible. The model is a hybrid of 1m and 2m. The SRAS curve continues a small way to the right of the LRAS curve, but then stops dead. And the outcome would also look like a hybrid of pictures 1 and 2.
You've simplified out new markets. When credit is easy and optimism is high, it's easy to get funding to expand in some new direction, if there is some obvious direction to go. To get a true boom, you need an opportunity to invest in the next new big thing - canals, railroads, radio, the internet, biotech, tulips, whatever (and there's always real estate. There not making any new land, you know. Soon it will all be gone...).
If you look only at booms of this type, you'll get a pattern. All business cycle upswings are not booms of this sort.
The other important point is credit. Some booms are financed from existing capital and some are financed through credit creation. Booms of the latter sort are much more likely to end in tears.
As for busts, some are the tears that credit booms are fated to end in, some are due to shocks, and most unfortunately some are the result of both at the same time. I believe 1929 and 2008 are busts of the third kind.
Credit problems for some reason tend to paralyze governments and lead them to take counterproductive steps - austerity, raising tariffs and taxes, price supports for commodities, tightening credit (is this really the right time to switch the banks to Basel III?).
Posted by: Peter N | June 01, 2013 at 12:45 PM
Nick, Excellent post, and I agree with the shape of the SRAS curve you end up with. I also agree that monopolistic comp. is the dominant market structure. But, I'm not clear as to exactly what you mean by "perfect competition." Why would a drop in AD reduce output with perfect comp? At one point you mention that wages and prices could be sticky with perfect comp. In that case would a rise in AD lead to shortages? We typically don't see many shortages in perfectly competitive industries, so I'm a bit confused by how you are describing perfect comp.
Posted by: Scott Sumner | June 01, 2013 at 12:56 PM
Peter: You are thinking micro. You need to think macro. At the individual level, if my firm gets more credit I can hire resources away from other firms and away from other firms' customers, and increase my firm's output. Now tell me a story where we start in equilibrium and firms in aggregate somehow find more real resources and produce more real output. Why weren't they already using all the available resources, including hiring all the workers who wanted to work and producing and selling all the output they wanted to produce and sell?
Posted by: Nick Rowe | June 01, 2013 at 01:02 PM
Scott: thanks! If we take a simple model in perfectly competitive Long Run Equilibrium, and then hold prices and wages fixed, and then reduce AD, firms can't sell as much stuff so stop producing as much stuff, so we get a recession. If we go back to LR equilibrium, hold prices fixed, and increase AD, firms are able to sell more stuff, but don't want to produce and sell more stuff, because they were already producing and selling as much stuff as they want to. So we don't get a boom, just shortages. Under monopolistic competition, we do get a boom, and we don't get shortages.
The problem is that we rarely see shortages, and yet the plucking model says we don't see booms either.
Posted by: Nick Rowe | June 01, 2013 at 01:10 PM
Stupid TypePad puts every single one of my comments in spam!
Yep, I had to fish this one out too!
Posted by: Nick Rowe | June 01, 2013 at 01:12 PM
Why not straightforwardly invoke asymmetric rigidities combined with the monopolistic competition, since there is plenty of micro-level evidence for asymmetric price and wage adjustment?
Posted by: david | June 01, 2013 at 01:51 PM
david: OK. If an individual firm will never set a price below its profit-maximising level, but may have a price above its profit-maximising level, that would give you the kinked SRAS curve like in my picture. But the micro evidence talks about resistance to *absolute* cuts in prices (or wages), I think. While what we need here is rsistance to having a price below the profit-maximising level. And the two aren't exactly the same, if the target rate of inflation is positive. (But then maybe over the period Friedman looked at to test his model, the average inflation rate was close to 0%, so this might not make much difference in practice.)
Posted by: Nick Rowe | June 01, 2013 at 03:40 PM
Nick Rowe,
What does your explanation imply for the cyclicality of productivity, to relate back to an interesting recent discussion on here?
Is the apparent procyclicality of productivity ALSO a case of the plucking model, i.e. recessions hurt productivity, but booms just restore its growth rate to its secular trend?
Posted by: W. Peden | June 01, 2013 at 08:02 PM
"Now tell me a story where we start in equilibrium and firms in aggregate somehow find more real resources and produce more real output. Why weren't they already using all the available resources, including hiring all the workers who wanted to work and producing and selling all the output they wanted to produce and sell?"
I'll agree that a model consisting of large stagnant monopolies operating at full capacity without disruptive innovation or speculative investment won't produce the kind of booms I'm talking about. However it also isn't a very good model of the US economy. For instance the dotcom and biotech manias would have been impossible.
From "High-Growth Firms and the Future of the American Economy" by the Kaufman foundation
"Every year, roughly half a million new firms are started in the United States; not all of these will survive, of course, and survival rates across time are remarkably stable.4 In the first two years, roughly a third of these companies will fail and, in five years, just under half (48 percent) will remain. Starting from a base year of zero, by the fifth year we will have two million firms younger than five years old—of these, around 809,000 will be between the ages of three and five.5 According to Census data, the top-performing cohort of this group (43,000 firms) accounts for 10 percent of overall net job creation in the economy. This is not a static figure, as each year means more new firms come into existence, more young firms are operating, and a steady number of fast-growing companies are creating jobs. High-growth firms, that is, accumulate over time, continuously adding new jobs, subtracting old jobs, and challenging incumbent companies. The firms that survive and grow more than make up for the companies that fail."
http://www.kauffman.org/uploadedfiles/high-growth-firms-study.pdf
models should be "as simple as possible, but no simpler"
Posted by: Peter N | June 01, 2013 at 10:25 PM
Peter N: "I'll agree that a model consisting of large stagnant monopolies operating at full capacity without disruptive innovation or speculative investment won't produce the kind of booms I'm talking about."
No, you don't *agree* with me there. You are saying the exact opposite to what my models are saying, and to what I am saying in this post. I am saying you *can* get booms with monopoly, and you *can't* get booms with perfect competition.
And that kaufman quote is micro, not macro.
W Peden. Hmmm. Dunno. First answer: since monopolistic competitive firms produce on the downward-sloping bit of their ATC curves (upward-sloping AP curves) total factor productivity should be procyclical. And since perfectly competitive firms produce on the flat bit at the bottom of the U on their ATC curves (flat bit of their AP curves) total factor productivity should be roughly acyclical (strictly, it would fall a little in big recessions).
Posted by: Nick Rowe | June 02, 2013 at 05:48 AM
Nick:
I think Peter has something like the following in mind.
In booms, credit is available and so innovative firms pull resources away from old, unproductive firms. The resources are used more productively, and output grows more quickly. Potential output grows more rapidly.
In recessions, credit dries up, new firms cannot get startup money, and the resources return to the old, unproductive firms. They get more resources, but they are so unproductive, that output falls. Potential output falls. (Like any real business cycle theory, falling output rather than just slower growth is a bit implausible.)
I think many people telling this story don't have the old, nonproductive firms getting more resources and so the reduction in output is due to unemployment of resources. The reduction of startups means that the resources that those startups would use remain unemployed.
Even if the nominal quantity of money adjusts so that total spending on output remains the same, or else prices and wages are perfectly flexible so that the real quantity of money adjusts to the demand, and real output remains equal to potential, when funds go to startups, potential output grows more rapidly. When funds don't go to startups, but rather go to unproductive established firms, then potential output grows more slowly. It could even shrink.
Of course, employment of resources wouldn't fall in recessions. And the recessions wouldn't involve inability to sell stuff. To get the reduced provision of resources you need the usual shape to labor supply functions. Funds don't go to startups, real wages are lower, so people play golf rather than work. (And if 20% of total wage income is going to people who have already made enough money that they could retire at any time, this could have a big effect on real output numbers, though maybe not much of an effect on employment numbers.)
Of course, rather than go with trying to make recessions about absolute decreases in potential output, instead let booms be extra rapid increases in potential output, and recessions be due to spending falling and output falling below potential.
If we had nominal GDP targeting and a laissez faire approach to finance, then when finance is working well, we can have a boom because funds go from savers to the most productive investors. When finance works poorly, we have stagflation. Startups can't get funds, and spending goes to incumbent firms. They have higher prices, profits, and produce and employ more. The inflation is obvious, but the increase in output (and increase in resource demand) does not fully make up for what the new startups would have produced.
If inflation is targeted, then spending on output slows to avoid this inflation, and so shift of resources to and increase in production from the old, nonproductive firms is even less. In fact, it really shouldn't happen at all until money wage growth slows.
Therefore, the only answer is to save the banks.....
I think these are the sorts of things Tyler Cowen has in mind.
Posted by: Bill Woolsey | June 02, 2013 at 07:34 AM
Nick Rowe,
Interesting: that presumably means that markets that are close to perfect competition (like agricultural markets) should have a very different business cycle from those that are closer to monopolistic competition.
A testable prediction from a plucking model of productivity would be that there should be evidence for Verdoorn's Law, but not in cases of major demand shocks i.e. increased demand has a positive impact on productivity only insofar as both are below trend. There are cyclically-driven productivity busts, but not cyclically-driven productivity booms.
I double-checked Verdoorn's Law on wikipedia, to see if I'm remembering it correctly, and Verdoorn's own description (admittedly out of context, but it's not specified elsewhere in the article) is awful: "in the long run a change in the volume of production, say about 10 per cent, tends to be associated with an average increase in labor productivity of 4.5 per cent". Nominal production? Real production? Real productivity? From the period of the history of economics that gave us 'demand' for 'real output', I'm not surprised.
Posted by: W. Peden | June 02, 2013 at 08:04 AM
Bill: Aha! That makes a lot of sense about what Peter might be saying.
W. Peden: in that definition, Verdoorn's Law sounds a lot like Okun's Law "fluctuations in real output are twice or three times as big, percentage-wise, as fluctuations in employment", with one big exception: Verdoorn is talking Long Run and Okun is talking short run. This is short run macro.
(Could somebody put a comment on Frances' Friends of Myles post saying that both Bob Smith and Frances are stuck in spam. I can't do it because all my comments go straight to spam, and I can only retrieve them from my own posts!!!)
Posted by: Nick Rowe | June 02, 2013 at 08:49 AM
Nick Rowe,
Thanks, that helps clarify the difference and makes Verdoorn's Law look much less plausible.
So the following three propositions are logically compatible, based on your last half-dozen or so pieces:
(1) Monopolistic competition is the norm.
(2) The plucking-model is the best approximation of fluctuations in output.
(3) Productivity is subject to the plucking-model as well as output.
Posted by: W. Peden | June 02, 2013 at 10:14 AM
Excuse my ignorance, but what does MC have to do with the shape of the LRAS curve?
As an example, in an increasing returns economy, say Y = L^2, there is a low-wage/low-output equilibrium, and a high-wage/high-output equilibrium. But output is capped by population in all cases. The wedge-shape has nothing to do with perfect competition.
Posted by: rsj | June 03, 2013 at 02:41 AM
rsj: I'm not sure if by "MC" you mean Marginal Cost or Monopolistic Competition! I think the latter.
Assuming long run flexible prices, and no money illusion, the LRAS curve is vertical whether you have perfect or monopolistic competition. But it will (usually) be further to the right under perfect competition than under monopolistic competition. And under perfect competition the SRAS curve will stop dead when it hits the LRAS curve, while under monopolistic competition it will (usually) keep going to the right of the LRAS curve for some way.
Posted by: Nick Rowe | June 03, 2013 at 05:54 AM
@Bill Woolsey,
Thanks, that's a good part of what I've been saying. There's a lot of turnover among companies, and most of the action is in companies 5 years old or less. This isn't a good match for a model of monopolies in equilibrium. You have to model the economy you have, not the one whose model is the most elegant.
There's also the question of credit.
Additional credit produces additional GDP once the borrowed money is spent on new goods. If you look at the broadest monetary aggregates like M4-, you'll see the steady growth paralleling that of the economy. There's no point in arguing which is a cause and which is an effect, as long as you agree that it would be a serious shock to the economy should this growth stop, which it most certainly did in a most dramatic fashion.
Here's David Beckworth on the subject:
http://macromarketmusings.blogspot.com/2012/07/safe-assets-money-and-output-gap.html
Posted by: Peter N | June 03, 2013 at 08:22 PM
I'm sorry I still don't understand.
I was imagining that with MC, you have a dotted line that is to the left of your PC line, sloping and then smoothly bending before going vertical.
You are saying it overshoots, corrects, and goes vertical. What is the basis of the overshoot?
Posted by: rsj | June 05, 2013 at 09:00 PM
rsj: it's difficult to explain. It took me a long time many years ago to get my head around it. My best attempt at explaining it is in this old post. And this one too.
Posted by: Nick Rowe | June 05, 2013 at 10:44 PM