A new paper from the National Bureau of Economic Research has made waves by claiming that 1.8 million jobs were created by the expiration of unemployment benefits at the end of 2013. What went into that estimate? Will it stand up to criticism?

First, some perspective: 2014 really was a good year for the labor market. The U.S. economy created 3 million jobs in 2014, up from 2.3 million in 2013 and 2.2 million in 2012. From its nadir in December 2009 to December 2013, the share of 25- to 54-year olds working grew 1.3 percentage points. In 2014, the same ratio grew 0.9 percentage points.

But 1.8 million is a lot of jobs to ascribe to a single policy change, even in a good year. Population growth accounts for about 1.3 million jobs a year, and some employment recovery would be expected regardless of policy. In addition, concerns with the data source used by Marcus Hagedorn, Iourii Manovskii and Kurt Mitman in the NBER paper mean that the dramatic estimate probably will be revised downward when better data becomes available.

Another problem is that although Hagedorn, et al.’s theory is sound, their empirical methodology may not match the theory. They want to use neighboring counties in different states as a pseudo-experiment, with each county experiencing a different policy treatment. But county pairs can either be too entwined, so that the policy treatments spill over across the border, or too distant, so they can’t offer a meaningful comparison. (For more details on the paper, see Patrick Brennan’s excellent write-up.)

Some critics of Hagedorn, et al. have pointed out that micro-based research, looking at the incentives for job searchers rather than overall jobs data, finds much smaller effects on employment. But Hagedorn, et al. explicitly agree: The macro effects, they argue, far outweigh the micro effects. So there is no intellectual contradiction between the new and old research, even though the policy implications are very different.

Manovskii and Hagedorn’s long-term research program is steadily building a theoretical and empirical case that there are large macroeconomic costs to unemployment insurance. Their former students, Mitman and Stanislav Rabinovich, have one of the most empirically successful models of unemployment, which finds that the post-1990 “jobless recovery” phenomenon is largely due to the expansion of unemployment insurance (a finding that the group vigorously defended against the Council of Economic Advisers).

The macroeconomic costs to unemployment insurance come from the demand side. Employers have a narrow benefit from each marginal hire, so hiring can respond a lot to a small shift in wages. Unemployment insurance gives potential employees more bargaining power, raising the equilibrium hiring wage a little, but lowering the equilibrium number of hires a lot.

Manovskii and Hagedorn’s research is a good reminder that economic incentives are as important on the demand side as on the supply side, and in imperfect markets as well as frictionless ones. While economists promising free lunches will always find a politician willing to buy them lunch, the research is perpetually reminding policymakers that there are real tradeoffs.

We know this: Unemployment insurance is a good way to soften the personal blow of job loss in a recession, but at the cost of slowing the recovery. How great those costs are exactly, we’re still discovering.

 - Salim Furth is a senior policy analyst in macroeconomics at the Heritage Foundation.