It is time for a showdown between the findings of two separate strands of empirical literature. On one side, studies of jobs and unemployment find that giving more to the unemployed cuts the number of jobs and raises unemployment. On the other, as we have seen, studies of the effects of total social transfers on the growth or level of GDP find no reliable statistical effect. The conflict stares at us directly in the raw data and is not just a subtlety revealed by the buildup of statistical studies. Just looking at the postwar record, we can see that unemployment rose dramatically in many countries after the 1960s, yet their GDP did not visibly drop relative to countries with less unemployment.

How can these two strands be tied together? How can GDP not be cut if jobs are cut? Is it just that transfers to the unemployed cut jobs and output, while other transfers actually raise output? If the story of no clear GDP cost is correct, did more generous unemployment compensation really not destroy any jobs, contrary to past findings? If subsidizing the unemployed makes fewer people have jobs, is the GDP literature overlooking true costs? The reconciliation cannot simply hinge on differences between the GDP effects of the dole and the GDP effects of total social transfers, since Chapter 18 found that even the dole itself did not have a significant GDP cost. Alternatively, could more unemployment compensation remove only completely unproductive workers, whose marginal product is zero?