Bad economics from McArdle!?

I know. I know. I’m thinking I misunderstood her in some way. She says:

When the unemployment rate is up around the 10% line, the problem is not people waiting around for perfection; it’s that there aren’t any jobs for them to take. Increasing peoples’ incentive to take a job offer will not do anything to increase the number of job offers; at best, you’re just shuffling the unemployment around some. Indeed, at a time when aggregate demand has collapsed, more generous unemployment benefits could plausibly make the unemployment rate lower.

The way I parse this paragraph — and I’m very suspicious of my parser given what its forcing me to conclude — she’s saying the labor supply curves for the unemployed don’t slope up and she’s saying UI is a more efficient policy than straight transfers (the alternative policy to UI) or normal aggregate demand policy.

Let me first state the obvious: because there are few job openings out there, doesn’t mean there are no job openings. Can we all please get aboard the marginal analysis bus?

Consider the decisions an unemployed person has to make regarding finding a new job. He must decide how hard to work at finding a new job and he must decide which job offers to take. Now, I have a very loose definition of job offer. The help wanted poster outside McDonald’s is very likely a job offer even though no one explicitly offered the job to the unemployed person. The harder he looks for a job the more implicit and explicit offers he will get.

Working backwards in time: will the unemployed worker be less likely to take a random job offer if UI benefits increase? Yes because the cost of taking that random job offer goes up as he has to give up more UI benefits.

Will the unemployed worker search less hard due to increases in UI? If he has no control of the quality of offers he gets (which is likely given the loose definition of a job offer), because random job offers are less likely to be taken, the expected benefits of searching are reduced. Looking harder for new jobs, for example, will produce more offers but those offers are, on average, worth less. With the same cost (loss of leisure time and searching for jobs is a pain in the neck) but less benefits, he’ll reduce his search effort.

How would either of these decisions depend on the business cycle?

(It would be interesting to see if people in their last week of UI benefits find jobs at higher rates than the rest of the unemployed. This is true in “normal” times, but I wonder if its true now.)

The increase in UI is inefficient because if you gave that money to the unemployed person but didn’t make that payment contingent on finding a job, you’d have more people employed quicker, thus more production, with the same cost of the policy. If you’re trying to manage aggregate demand, UI is a terrible policy instrument.

UPDATE: In comments I mention a study by Card and friends that used a UI extension in New Jersey as an experiment to see its the effect on unemployment. They find extending UI by 13 weeks increases average unemployment durations by a week and increases the number of chronically unemployed by 7%. I messed up the calculation to see what this translates to in unemployment percentage points. The answer, as it turns out, is very sensitive the separation and hire rates (the percentage of the employed that become unemployed and the percentage of unemployed who become employed). In normal times, these are about 1.3% per month and 28% per month respectively. The Fed says unemployment duration is about 26 weeks as of last month. This suggests a hire rate of 15%. You can infer the separation rate if you assume we’re in the steady-state (alarm bells!). Anyway, if you plug in the “normal times” separation rates (1.3%), you get about 1.5% additional unemployment from the UI extensions and if you plug in the implied rates using steady-state assumptions (1.6%) you get 0.2% additional unemployment from UI extensions. A big range.

8 thoughts on “Bad economics from McArdle!?”

  1. I think what Megan is saying is that, given that the number of unemployed people per job offering has tripled in the past year and a half as a result of the business cycle, the high unemployment effect’s on the likelihood of pulling a job from the job bucket, and an agent’s expectation of that being rough throughout 2010, dominates the worker’s decision to the point where adverse incentives from UI is a rounding error. Or at least that’s my take when the discussion is brought up.

    I assume the agent’s expectation of the job value function has to have broad variables for numbers of others searching (high) in the denominator and number of firms hiring (low) in the numerator that move with the business cycle – or is it a magic bucket of jobs? – but I’m not there yet.

  2. Also, fwiw, “Food services and drinking places” sector is down jobs YTD, which is 45% fast food, 45% restaurants. I wonder how that looks underneath, because restaurants are taking a bath in ’09 but fast food stocks have been on a tear with the economy in the toilet, as everyone wanting to get as many calories for $1. (Is a Big Mac giffen?)

    I assume there are more, perhaps even a lot more, jobs at McDonalds than there were before – but would a former executive with a college degree be a shoe-in for the hire? Or would there be a bizarre adverse selection from the POV of Mayor McCheese?

  3. Mike, what you say is true. The theory has the macro variables in there (if you think unemployed workers produce externalities in their searching for other unemployed workers), but the UI variable is in there too. Holding the macro variables constant, UI increases unemployment.

    I’m not convinced the effect size is a rounding error. Studies I’ve read have shown significant effects of extending UI benefits. (An “experiment” in New Jersey suggested increasing UI length by 50% increases the number of long-term unemployed by 7% and increases the average unemployment spell by a week… 1.5% in unemployment percentage points if I’ve got my math right). They’ve estimated these effects controlling for the level of unemployment and their sample contained the nasty early 80’s recessions.

  4. As usual, my arithmetic was wrong and qualitatively so… the study I mentioned suggests unemployment in today’s environment would be increased by about 1/2 a percent by increasing UI 13 weeks (as is being contemplated in congress).

    UPDATE: see the update to the main post on these calculations. Does anyone report hazard rates implied from the JOLTS surveys?

  5. It matters when you extend the benefits. At this point, anyone who is taking a job quickly rather than waiting around for the perfect job is not noticeably reducing the friction in the labor markets, because jobs are being net destroyed. Taking a job means that you’re unemployed for a week less, but with two few jobs to go around, it means someone else is unemployed for a week longer.

    As unemployment regresses to the natural rate, frictional/structural unemployment dominates cyclical unemployment, and providing more unemployment benefits does indeed lengthen unemployment.

  6. I’m not aware of any model that has that prediction and most empirics I’ve seen finds negative effects of UI controlling for the level of unemployment.

  7. I think that Megan’s result is based on the idea that the unemployed, especially the poorer among them, are constrained somehow in credit markets, so giving them money while they’re unemployed will encourage them to shift consumption to current periods. This is an empirical question as to how large the effect might be. At the very least it will remove a fair amount of risk from the loss of income, so it might be a good idea welfare-wise, to the extent that people are simply risk averse.

  8. Chris, you can take the UI money and give it to the unemployed without making it contingent on finding a job. This way you deal with the welfare issue but you don’t give disincentive to find work.

Comments are closed.