What Do Unions Accomplish?

Some useful back-and-forth in a number of places on the effects of unions. Part of the discussion spins off a recent paper arguing that in closely contested representation elections, the union/no-union outcome seems unrelated to subsequent wages. My take (also posted in comments at Marginal Revolution):

Note, though, that the election outcome, including the margin of victory/loss, is endogenous. Workers at firms at which unions are more likely to have an impact on wages and benefits are also more likely to vote for a union. In relatively evenly divided votes, my prior is that workers are less certain (and, in a rational expectations world, correctly less certain) that unions can provide benefits.

I would think that comparing unions that win close certification elections with unions that win easily might come closer to the issue. Controlling, of course, for the other factors we are likely to think matter (capital per worker, industry/market structure--yes, I'd expect unions to be able to capture some monopoly profits, worker characteristics, and so on).

UPDATE: More here and here.