PK and the literature

I always laugh when Paul Krugman urges others to read the literature. He’s a brilliant economist, but he’s not exactly the goto guy for literature reviews…

Today he says that Sumner and Avant should read the literature on macro vs micro labor supply elasticities. Well, ok, he says they should only read half of that literature. Over the last decade, Prescott has been doing a lot of work showing that differences in taxes explain differences in employment and hours worked between Europe and the US. I think his Nobel speech was about this.

The micro people threw fits though because their estimates of the response of labor supply to tax changes is much less extreme than Prescott’s finding suggest. They basically find labor supply curves are vertical. This would mean that taxes simply can’t have an effect on labor supply.

For a while, these guys had me convinced because, in general, micro/labor types do a much better job of identification and I trust their estimates more than I trust macro estimates. More recently, however, macro people ((A also vagualy remembering a paper that uses the PSID to estimate the two types of elasticities, but I can’t remember what it was.)) have been making the case that the “labor supply elasticity” estimated by the micro people is different from the “labor supply elasticity” the macro people estimate. The difference isn’t due to statistical methodology, we were just calling two different things the same thing.

Of course, its the macro elasticity that matters for tax policy, though. Prescott’s work (and not the paper that PK links to) is the place to go for understanding differences between Europe and the US. He says that difference is due to differences in tax and transfer policies.

UPDATE: AC found the PSID paper.

The unemployed aren’t the only ones seeking jobs

Work with me here. In normal times, say 2004 through 2007, suppose 10% of the working population are looking for jobs while still employed (do you know of a better estimate?). This means about 13 to 14 million employed workers are “job seekers” in normal times.

“Quits” are voluntary separations from jobs and folks do that because they’re leaving the work force (e.g. retiring) or because they found another job. Now look at quit rates over the last couple of years:

Quits have declined by about 40% compared to normal times. From here, about 50% of quits are retirements. If the retirement rate has stayed the same, then quits due to job changes went from 1.1% to about 0.5%. This suggests the number of “job seekers” among the employed has gone down by at least half.

How many job seekers are there right now? Supposing all unemployed workers are “job seekers” then the total number is about 20 million people. In normal times, that number is about… 21 or 22 million people (unemployed plus 10% of the working population). By this measure, the job market now is less congested than usual!

Making the assumptions that I made above, I constructed a “job seekers” per job opening time series:

If you assume none of the currently employed workers are “job seekers” then the graph above looks like the one put up by EPI. If you assume 20% of the currently employed are “job seekers” then even the up-tick seen in the later part of the time series goes away.

Japanese unemployment


In November, Japanese unemployment was at 5.2%.

Many folks are comparing the Fed’s action today to those of the Bank of Japan in the “lost decade” (e.g.). I can’t quite peg down the comparison being made by Sumner, but Yglesias articulated it most clearly:

[Bernanke] knows that unemployment is a problem now and he believes that he could fight it, but that fighting it more aggressively would elevate the risk of inflation in the future and he thinks that reducing the possibility of future inflation is more important than reducing the reality of current unemployment. I think that’s nuts. But it’s an attitude the Bank of Japan has consistently maintained since the 1990s.

Here’s the employment to population ratio:
Its safe to say employment wasn’t a problem in the “lost decade” and there wasn’t too much weight put on inflation relative to unemployment.

Costs of discretion

In the previous post, I argued upping the inflation target would have no impact on unemployment and it would risk unanchoring inflation expectations. If expectations became unanchored we would be trading off a permanent increase in expectations for a temporary decrease in unemployment.

Assume we gave half of those currently unemployed a job, output would rise by about 5%. Without retargeting, I’m guessing it’ll take 5 years for unemployment to get below 5% (unemployment has a two-year half life). So we’re getting about 13% of GDP by retargeting.

The costs of 10% of inflation are estimated between 1% and 5% of GDP per year. Woodford estimates that under a discretionary policy, inflation would be about 10% (I’m looking at figure 7.1 in his textbook). Discretion isn’t really permanent, its just really hard for the Fed to get expectations reanchored. Suppose x is the number of years of unanchoredness and so its the number of years we’d have 10% inflation.

How big can x be and still make retargeting cost effective? Using a low-end estimate of cost of inflation of 1%, x is 17 years. Using a high estimate of 5%, x is 2 years.

Accelerating output and prices equals additional stimulus?

The Fed made a mistake last year. No doubt the Fed chairwoman 70 years from now will give a speech admitting fault for the “Great Recession”. Fine.

It does NOT follow from this that we need more aggregate demand stimulus now. Obvious, right?

Now, unemployment is high. Empirically, we know two things about unemployment. It goes down when output is increasing (the so-called Okun’s law) and it lags output. We have increasing output and inflation. If those empirical regularities rear their ugly heads this time around, we know that unemployment will decline eventually, too.

Politically, I get why presidents react to bad-sounding headlines and hold “job summits” when unemployment goes above magic numbers. I don’t get the argument from economists for more stimulus. What theory, what mechanism of lagging unemployment would suggest we need increased stimulus even though output and inflation are rising?

I can think of a few things:

  1. Laid off workers are just waiting to be rehired by their former employers
  2. There wasn’t enough stimulus and we’re moving towards long-run equilibrium too slowly
  3. Variant: policy makers were too concerned about inflation and they should put more weight on unemployment going forward

It seems the internets are discovering the fact that this recession hasn’t been characterized by a large number of job losses. Instead, unemployment has been increasing because its harder for people who would have lost their job anyway to find new jobs. The JOLTS data, in other words, suggests that even in the recession laid off workers weren’t temporarily let go. They were fired and no amount of stimulus will get their old employer to hire them back. I’m willing to wager most fired employees have no expectation that they’ll be rehired by the same employer that fired from. In fact, I’m looking at data that suggests they shouldn’t even expect to be rehired in the same occupation; they’re going to have to go out and acquire new skills and find a whole new line of work.

Number one is bunk. What about number two? The problem with the “recovery is too slow” argument is that we don’t know what “too slow” means. Looking at the model I shared the other day, even under the optimal policy that makes the proper trade-off between unemployment and inflation, it takes two years to get half way back to normal unemployment rates. If you believe in the “too slow” theory, then you have to write down a model that would generate faster decreases in unemployment (and not just increases in inflation).

Number three? According to the Gali/Blanchard model, if the Fed is heartless and only cares about inflation, then unemployment is much worse at the time of the shock. This is true. However, as the first graph I showed indicates, the decline in unemployment is much faster. From the peak of unemployment it takes less than two years for unemployment to get half way to its normal level.

Under the third scenario, you might say the Fed should switch to the optimal policy. It should put more weight on unemployment and so it should ease policy. Well, according to the Taylor rule that approximates the optimal policy in the Gali/Blanchard model, interest rates should be about 0.5% 0%. Even under optimal policy where both inflation and unemployment are targeted, no further easing is necessary (and current policy might even be slightly inflationary).

UPDATE: as always I got my arithmetic wrong… post updated with strikes to show what changed.

A model

Blanchard and Gali (2008) incorporate unemployment into the standard model. They have some interesting findings, but this one stuck out:
Unemployment under inflation targeting
This is the response of the unemployment rate over time to a 1% decrease in productivity of the economy (a “real” shock) WHEN the Fed has a inflation-only target (i.e. it doesn’t care about unemployment). Blue is the response in the model when its calibrated to look like the American economy and red is the response when its calibrated to look like Europe. Unemployment keeps increasing after the shock for a couple periods and then gradually declines. Look familiar?

Here’s the same graph but for when the Fed has the *optimal* policy of targeting a weighted average of inflation and unemployment:
Unemployment under optimal policy
Unemployment still jumps but it doesn’t have the hump shape. More importantly, notice how little the unemployment increases when the Fed is following optimal, or best, policy… almost an order of magnitude difference in the response of unemployment.

What does this mean? This paper gives us two ways to interpret what happened since last Fall. One, there was a small to medium sized real shock but because the Fed cares too much about inflation, unemployment sky rocketed. Reality looked (and looks) like the first graph. Or two, the Fed is following optimal policy but there was a huge real shock. Our reality is more like the second graph, but amplified.

Ironically, those making a bunch of noise about banking regulation, centering the blame for the recession on the financial sector, are arguing for the second graph. My impression is that those people are also more likely to be agitating for more aggregate demand policy. The second graph has the Fed acting optimally, i.e. there’s no need for further stimulus because the Fed is doing everything necessary. If you buy the logic of this paper, however, you can’t have it both ways.

PS – This model also has a positive response of inflation to the real shock. Given we saw a negative response last Fall there’s still room for a monetary shock in the story.


The problem with most stories is that they’re just words. With just words its easy to sound like you’re making sense, that the one thing you said leads to the other. For example, Ryan Avant says we need more stimulus and so does Scott Sumner (e.g. ?).

So, we have increasing inflation and GDP. Understanding the limited ability of policy makers to fine tune the economy, what model under conditions of increasing inflation and increasing GDP recommends more stimulus?

Speaking of which, I’m embarrassed to say that I have no idea why unemployment is a lagging indicator, but I do know that it has been so for a long time. This means high (but decreasing?) unemployment isn’t necessarily an indicator of insufficient aggregate demand.

I wish this editorial wasn’t at CATO

“Conservatives” hate the unemployed so the facts on display in this editorial are obviously ideologically tainted. (h/t Wilkinson)

Look at the JOLTS data yourself: unemployment is being driven by a slow hire rate (fire rates are at about their normal level). Hiring rates could be depressed because of low demand for workers or for low supply of job hunters. GDP is rising which suggests demand for workers should be increasing. Unemployment benefits, on the other hand, have risen substantially so there are incentives to stay unemployed. Finally, we know the disincentive effect is large because of research by conservative stalwarts such as Larry Summers and Alan Krueger ((both somehow sneaked their way into high posts in the Obama administration)).

Why is the unemployment “rate” called a rate?

Its so much fun to accuse a Nobel Laureate of getting basic economics wrong. Paul Krugman confuses levels with rates.

The unemployment “rate” is the number of unemployed divided by the labor force. This is not a rate in the rate vs. levels sense. Its a level. Just like the real money supply is a level, the unemployment “rate” is a level. It has just been normalized.

Suppose there’s a bathtub with water flowing out through a leak and water flowing in from the facet. The water level is constant if the amount of water poured in is the same as the amount poured out. You will notice, though, even if the water level is constant the water itself is not. In such a dynamic system, eventually all the water will be replaced by new water. The water level tells us nothing about that.

The number of unemployed is like the water level in the bathtub. There’s always newly unemployed workers being poured into the unemployment tub and there’s always workers leaking out when they find jobs.

Ryan Avent (somehow people discovered the true identity of The Economist blogger) and Paul Krugman are wringing their hands over the slow pace in the reduction of the unemployment during recoveries. Similarly, Brad Delong is always talking about a “jobless recovery”.

But do high unemployment rates mean people can’t find jobs? No. This is like suggesting that because a bath tub is full there’s no leak. Would you pat a plumber on the back and tell him “good job” because he “fixed” the leaking problem by turning on the facet?

Despite the headlines and the breathless way some bloggers write about it, people don’t care about the unemployment rate, they care about whether or not people are finding jobs.

From the JOLTS data, it looks like unemployment in this recession is being driven by an unusually low hire rate (firing rates are more or less the same as normal). Suppose the recession ended and the job finding rate immediately resumed its usual, non-recession value. How long would it take for the unemployment rate to get from the current 9.8% to normal levels? Doing the math, it would take about nine months. On the other hand, if it takes about a year for the hire rate to get back to its normal values, we wouldn’t see normal unemployment rates for a year and a half. So the longer it takes for the hire rate to get back to normal, the longer it will take to see normal unemployment rates.

That last sentence might have seemed obvious, but consider the implications. With improving hire rates, the average amount of time spent unemployed would be dropping. This answers Avent’s questions about the political economics of persistent unemployment. If people are finding it easier to find jobs, they’re less likely to throw fits in the ballot box, no?

Also it suggests we shouldn’t care much for how fast the unemployment rates get back to normal after recessions.