The slow speed of recovery, PSST, plucking, and skewed news

Judging by output, employment, and unemployment, the Canadian economy hit bottom around the middle of 2009. Two and a half years later, output, employment and unemployment have recovered a lot. But like most observers I believe, and certainly hope, that we have not yet had a full recovery to the long run sustainable path.

Why is recovery taking so long?

You might argue that the economy recovered slowly because Aggregate Demand recovered slowly. Never mind the reason why AD originally fell, if the Bank of Canada had followed a more aggressive monetary policy beginning in mid-2009, the economy would have recovered more quickly. But the data on inflation, for both total and core inflation, suggest that AD was not the whole story. If deficient AD had been the only thing slowing the recovery, I would have expected inflation to have been below the 2% target, or even falling relative to target, since mid-2009.

I want to suggest a supply-side explanation for the slow recovery. I am not saying that a supply shock caused the recession. I believe the recession was caused by an AD shock. I am not even saying that there was a supply shock. I don't believe there was. But that doesn't mean we can ignore the supply side of the economy when we want to understand the slow speed of the recovery.

Suppose a big negative AD shock hits the economy. Given nominal rigidities, that has real effects. Patterns of Sustainable Specialisation and Trade get disrupted. Workers lose their jobs, firms go bust, and teams get broken up. A recovery of AD is a necessary condition for recovery. But it takes time for Humpty Dumpty to recalculate and put himself back together again. At the really micro-micro level, right down at the level of individuals, we won't expect to see the same people back in exactly the same jobs in the same firms selling to the same customers as before the recession. It takes time for new workers to be interviewed and hired and trained on-the-job, for new firms to get started or surviving firms to expand to replace those that disappeared, and for those firms and customers to find each other.

Arnold Kling doesn't need a supply shock to make PSST and recalculation relevant.

If the original negative shock to AD is big enough, and lasts long enough, you can't suddenly increase AD back to normal and expect the economy to put itself back together again instantly. If you increase AD too quickly, the only result will be inflation.

The standard natural rate model with an approximately linear Phillips Curve says that the average level of output, employment, and unemployment will be approximately independent of the variance of output, employment, and unemployment, if that variance is caused by AD shocks. Milton Friedman's "plucking model" disagrees. Here is Mark Thoma's (2006, positively ancient by blogging standards) exposition of that model, and its empirical support. The plucking model says that an increase in the variance of unemployment will increase the mean rate of unemployment. The business cycle is asymmetric. Loosely speaking, in the plucking model there are recessions, but no booms.

Let me now try to put these three themes together.

If good monetary policy gives a steady growth in AD, there will be no disruptions of PSST, other than those caused by the steady stream of little supply shocks. If bad monetary policy causes a high variance of AD, PSST's will be disrupted in negative AD shocks, and will only restore themselves slowly over time even if AD recovers quickly. Negative AD shocks will do harm to PSST, but positive AD shocks cannot do good to PSST. In any designed or evolving system, like PSST, cars, and organisms, the news is skewed. There can be bad news, but the only good news is the absence of bad news. As Frances Woolley said, her Toyota Matrix might unexpectedly stop working, but it won't unexpectedly start driving like a Lamborghini. Frances' Toyota is like Milton Friedman's plucking model. Jeremy Fox tells me how Fisher explained why genetic mutations are mostly like that too.

(Scott Sumner gave me the idea for the link between the plucking model and the news being skewed.)

Comments

[marris]

> But it takes time for Humpty Dumpty to recalculate and put himself back together again.

I think I see your overall point. One problem with the Humpty Dumpty (HD) analogy is that each HD piece *knows* it's role in the combined form. It's the same role it played before the smash.

PSST recovery is more like the T1000 in Terminator 2, right? Once it's smashed into tiny metal slivers, it will try to recombine. But who knows what form it will take? Maybe it will look like Robert Patrick. Maybe like Linda Hamilton. Maybe like Arnold. But until the recombination is done, it will just look like metal mush.

[david]

Kling might not like the argument. If there's recalculation to get to a low-AD equilibrium, then there is symmetrically another recalculation to get to the high-AD equilibrium, so even in a hidden technological shock a la Kling, it's very plausibly welfare-improving to just stay at the high-AD equilibrium and never come down at all. You save yourself one recalculation, at the expense of debt or inflation.

[david]

Also... PSST is not quite like a car falling apart. A car is consciously put together. PSST is not consciously put together; indeed that's the Austrian extent of it (pick your favorite sunspot-equilibria, industrial policy alternative otherwise). For PSST to be the long-run attractor, it has to be the result of non-conscious falling forward.

In short, you can't compare it to a Toyota falling apart and failing to turn into a Lamborghini - that is exactly what is argued to happen. The analogy doesn't work. Compare rather a regrowth after a forest fire burns away a non-fire-resistant ecosystem. The new ecosystem is different, and it takes time to move to a stable state.

[Ralph Musgrave]

Given the reduced reliance we will hopefully place on irresponsible lending and borrowing, Humpty Dumpty will doubtless be put back together in a slightly different way to the way he was constructed prior to the crunch.

On the other hand, you’d expect the main group of people to be adversely affected by the crunch to be construction workers. Yet there is evidence that this group has had no more difficulty finding work than other groups. See (p.8 in particular):

1. http://www.rooseveltinstitute.org/sites/all/files/stagnant_labor_market.pdf

2. See under heading “Reason for job loss…” here:

http://www.clevelandfed.org/research/trends/2010/1110/01labmar.cfm

[Jodi Neufeld]

Tough questions. There are a couple of interesting related papers by Jordi Galí et al at http://research.barcelonagse.eu/Working_Papers.html

[Nick Rowe]

Briefly

marris: if there has been no change in underlying tastes and technology, the economy should return to the same shape it was before. But different individuals may be in different positions in that shape than they were in before.

david: in my version, the negatice AD shock breaks up the old PSST's, then that same PSST (but with different individuals perhaps switching places) reforms after. There is only 1 recalculation, which calculates the same old answer. And yes, it would be welfare improving to have no AD shock.

An organism is not *consciously* put together, but the Darwinian process means it is as if it were designed (blind watchmaker, etc.). See the link by Jeremy Fox. PSST is a mixture of conscious design and unconscious evolution. It's both car and blind watchmaker.

Ralph: US construction? Dunno.

Jodi: thanks. I will check later.

[K]

Nick,

Does this mean that while nominal wages may be sticky down, it's real wages that are sticky up?

[Nick Rowe]

K: I don't think so. Think more of a SRAS curve that comes to a dead stop, just a little bit to the right of the LRAS curve. Starting in LR equilibrium, a negative shock to AD causes Y to fall, but a positive shock to AD barely causes Y to rise at all. (Yes, this does contradict my old monopolistic competition story a bit, which says that booms and recessions are roughly symmetrical, at least within bounds.)

[K]

Yeah, what I said was dumb. The SRAS shifts left (NAIRU rises) as the economy breaks in response to the demand shock. And there is little resistance to a left shift compared to a right shift (as you say, firing workers and destroying value, while painful, is a piece of cake compared to hiring new ones and creating value). But the SRAS *will* shift right again in response to increased demand. Just slowly. It's employment (the SRAS), not wages, that's sticky.

But if we have an SRAS problem, shouldn't we at least push on AD hard enough to see a little bit of wage inflation? If not, are we pulling the SRAS to the right at all? Given that nominal wages are so sticky downward, we could be way below equilibrium and not even know it. A bit of inflation is the only way to know for sure that we are pushing in the right direction.

[K]

Here's another point, that I think is important. People (employers) have no idea where the LRAS *is*. How would I know what limits the consumption potential of my customers, never mind the whole economy? If they knew, and they were really confident in that knowledge, they might invest with the expectation that AD would follow. But since they have absolutely no idea they need demand as a signal that they are underinvested. Without that, I don't see why the SRAS wouldn't get stuck in a very suboptimal state essentially forever, as knowledge/information is permanently lost. I.e. if the SRAS doesn't shift to the LRAS, eventually the LRAS will shift to the SRAS.

NGDPLT, by the way, seems like a good way to constantly be testing the position of the SRAS, so I may just be channeling Scott here (or maybe he'll correct me).

[Patrick]

Isn't it the death spiral feedback of an AD shock into the supply side that's so insidious? e.g. AD gets clobbered (for whatever reason), incomes and spending decline, firms contract and people loose their jobs, lather rinse repeat.

[David Pearson]

Nick,
"If good monetary policy gives a steady growth in AD, there will be no disruptions of PSST"

I think the evidence argues otherwise. We had a pattern of specialization and trade develop during the latter part of the Great Moderation: increased use by the financial system of s.t. funding backed by housing collateral. This pattern suffused the system, reaching everything from start-up finance (using housing net worth) to consumer durables purchases (again using housing NW) to the s.t. repo funding that dominated the shadow banking system. The period of stability in AD contributed the this pattern. There is a stability/resilience trade-off in the credit channel: the more central bank-induced AD stability, the less actors will insure against liquidity, duration and credit risk; the more fragile the system becomes.

Part of the reason for the slow recovery is the damage done to credit intermediation. This is particularly true in start-up finance, which reaches across all sectors of employment but is particularly important in the smaller-firm dominated services sector.

We have a PSST problem in part because of AD stability, not because of a lack of AD stability.

[Nick Rowe]

K:

Paragraph 1: Hmmm. Yes. I think that's a good way of thinking about it.

Paragraph 2: Hmmm. Maybe yes. And I think it's what NGDP level path targeting would say.

marris again: I haven't seen Terminator 2. I think the thing will recombine into roughly its same original shape, but different bits may be in different places in that shape. The left knee is now the right knee, and so on.

[Nick Rowe]

K @ 10.21: Hmmm. Another good point, I think.

Patrick: That's still a demand-side AD process. You are describing the Old Keynesian multiplier.

David Pearson: OK, maybe. You are adding Minsky to my story.

[Determinant]

Partly the recovery is taking so long because we have a framing and measurement problem. Stephen has blogged steadily about Canada's performance; I believe he has been entirely too overoptimistic.

Unemployment and job recovery need to be adjusted for working-age population growth and output growth needs to be adjusted for productivity. We will not recover one percentage point until the economy grows faster than population growth and productivity growth combined. It hasn't done this since 2008.

Adjusted by these criteria we have been standing still since 2008. The Globe & Mail had an excellent article on this a few days ago. Now I know why I disagree with Stephen.

We really are suffering a Keynesian Multiplier problem like Patrick described.

[Bill Woolsey]

Nick:

Before I buy this, I need to see NGDP back to trend.

If the result is just more inflation, then perhaps the problem is the disruption of PSST.

I don't think failing to see as much disinflation as we might like is enough of a test.

The whole, the central bank responds to disinflationary pressure by being a bit more expansionary policy just seems a bit problematic to me.

I should raise my prices more slowly, because costs are rising slowly and I can expand sales. But when I (and lots of other people do this) demand will grow a bit quicker, and I really wouldn't have needed to slow my price increases after all. Of course, sales will be even stronger if that happens.

[Wonks Anonymous]

I thought one of the features of the plucking model is that a sharp downturn tends to predict a subsequent sharp upturn. So the recent prolonged recessions with "opportunistic disinflation" look less like the recessions Friedman had observed earlier.

[Nick Rowe]

Determinant: "We really are suffering a Keynesian Multiplier problem like Patrick described."

Then why hasn't inflation fallen more, if slow growth in AD is the only problem?

Bill: I can see your story working if it's the failure of the price level to revert to trend (rather than the inflation rate falling below trend) that is consistent with AD growing too slowly. This goes back to my old post on whether it's the price level or the inflation rate that belongs in the SRAS/SR Phillips Curve.

Wonks: the way I interpret the plucking model is like this: adjusting for trend, the magnitude (not the speed) of the upturn should match the magnitude of the previous downturn; but the magnitude of the downturn should not (always) match the magnitude of the previous upturn. Look at Mark Thoma's picture to see what I am talking about.

[Determinant]

One, oil, the critical commodity, has not fallen and has steadily increased and stayed high.

Second, downward pricing pressure is counterbalanced by the erosion of margins. In a world where business relies on banking to manage cash flows, maintaining margins in order to maintain credit quality is critical. Banking acts like a high-pass (low-cutoff) filter on business activity.

Between the need to maintain margins and the continued high price of oil, you can easily get modest price inflation even in an otherwise depressed economy.

Instead of deflation and then default in the classic Keynesian view you would get modest inflation where firms struggle to maintain margins and credit quality but you still get a high level of bankruptcies. You'd get more called loans in this case than voluntary, debtor-led bankruptcies.

Or call it survivor bias. Inflation numbers are dominated by firms that survive. If they survive they have more pricing power than bankrupt firms. Dead companies don't figure in price inflation statistics. Say survivors have a 1% pricing power advantage over bankrupt companies. Then you would expect to see 1% price increases in inflation statistics.

So don't look at inflation per se, look at business bankruptcies, layoffs and hiring figures.

[Scott Sumner]

Nick, I like this model, but I'd add one point that I think is very important (I'd love to know what you think.) I think the PSST model works way better for Canada than the US. I'm talking order of magnitude better. Suppose the PSST model was 100% wrong for the US. It's all AD. But suppose Canadian industry is heavily connected to the US market. Canadian firms make auto parts for US firms, etc. Then if the US has a recession for 100% AD reasons, then Canada could suffer a PSST recession.

Of course in the real world PSST would also apply somewhat to the US, as we have auto parts makers for cars assembled in Ontario. But I think the US is more AD driven in it's shocks. Maybe that's why our inflation has only averaged 1.2% since September 2008.

Apologies if other commenters already made this point.

[Nick Rowe]

Scott: I think that comment is spot on. It hasn't been made, but (you have to take my word for this) I was thinking about exactly that point last night. And I think it fits the data too, because Ontario has been one of the worst hit, I think, and I think it does more trade with the US. (I could be wrong on the empirical stuff, but we should expect to see worse unemployment in areas, both geographical and occupational, that are most linked to US trade.)

[JP Koning]

Thanks for this. There are a lot of us-vs-them macro arguments in the econ blogosphere, nice to see how different macro theories can be integrated.

When you talk about AD above, and given your recent posts (http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/12/the-trade-cycle-vs-is.html), are you referring to the total demand for not just final goods but all things that can be bought with money? This would be a redefinition of AD, right?

Also, if you are more concerned with the trade cycle and transactions than before, do you still think NGDP targeting is useful, since NGDP as a category is comprised of only final goods?

[edeast]

Video time?
Psst reminded me of micro-tubule catastrophe and recovery, dynamic instability of cytoskeleton within cell(economy) but after looking into it, the analogy doesn't go too far. Thought maybe the lack of money similar to not enough gtp or something, but just some pretty pictures.

[Nick Rowe]

JP: Hmmm. I was wanting to duck those questions! I think I can set them aside for this blog post. But they are floating around in my mind, unresolved. NGDP can only be a proxy. Better than most. But it is flawed for the reasons I gave in that old post on the Trade Cycle.

edeast: they are quite amazing!

[JP Koning]

"I was wanting to duck those questions!"

Looking forward to posts on this in 2012. I'm throwing a lot at you as in that comment since it's more geared towards the sum total of your last five or so posts + the interesting debate you stirred up regarding debt rather than this particular post.

[Scott Sumner]

Nick, I'm not surprised you had the same thought. I wonder if Kling has seen your post.

[JP Koning]

There you go, Arnold Kling responds.

http://econlog.econlib.org/archives/2012/01/easier_to_break.html

[Nick Rowe]

Thanks JP!

[henryc]

This may or may not be significant. Kling states "So far, the market has managed to figure out that we do not need so many lawyers. What it has not figured out is where the excess lawyers should go and how to make the transition happen."

But here is a recent report on legal jobs in the US:

"According to 2010 data reported in the U.S. Bureau of Labor Statistics Current Population Survey, the national unemployment rate was 9.6% for all occupations, while the unemployment rate for lawyers was 1.5%. The vast majority of management and professional occupations had higher unemployment rates, and many had unemployment rates that were much higher, for example:

• Astronomers and physicists had double the unemployment rate of lawyers;
• Computer software engineers and accountants had more than triple the lawyer rate;
• Environmental engineers had more than four times the lawyer rate;
• News analysts, reporters, and commentators had more than five times the lawyer rate;
• Advertising and promotion managers had nearly six times the lawyer rate; and
• Architects had nearly seven times the lawyer rate.

Among the ten categories of management and professional occupations established by the Bureau, legal occupations had a combined unemployment rate of 2.7%, the second-lowest rate. Health care and technical occupations topped the list with a slightly lower combined rate of 2.5%.

Bureau data also showed that between 2001 and 2010, the economy supported an additional 246,000 jobs for those in legal occupations, including 123,000 additional positions for lawyers. Even during the recession and modest recovery, the number of employed lawyers grew by 3.9% or 1.3% per year. From 2009 to 2010, the number of unemployed lawyers fell by 33.3% -- from 24,000 to 16,000 nationwide."

[Lars Christensen]

Nick, I disagree.

I think the slow recovery has nothing to do with PSST. Rather I believe that David Eagle has the answer. Bank of Canada is an inflation targeter - rather than a NGDP or Price level targeter. That mean that BoC will tighten monetary policy earlier than under NGDPLT or PLT - and that will obviously also be reflected in expactations.

David’s research that the speed of recoveries in the US economy to a very large degree can be explained by whether the Fed has targeted growth or the level of price or NGDP.

Had BoC changed it's target to PLT or NGDPLT I am pretty sure that the Canadian recovery would have been much stronger - recessions and inflation is monetary phenomena, but so is the speed of recovery.