I’m a big fan of means-testing (see here, here, and here for starters).  Analytically, though, stringent means-testing is indistinguishable from high marginal tax rates.  In both cases, the government takes away a big chunk of every dollar you earn.  Philosophers may say, “There’s a moral difference between taking away what you’ve earned and giving you less of what you haven’t earned.”  Selfishly speaking, though, they are the same.

For dogmatic neoclassical economists, this settles the question.  But no one should be a dogmatic neoclassical economist.  Empirically, the equivalence remains an open question.  Behavioral economics has repeatedly shown us that framing matters; people may simply fail to mentally equate “marginal benefit reduction” with “marginal tax increase.”  This is especially plausible when the marginal benefit reduction happens far in the future, and the benefit formula is poorly defined in any case.

Take me.  Conceptually, I know that my future Social Security benefits have something to do with how much I pay in Social Security taxes.  Yet when I weigh whether to pursue extra income, I’ve honestly never considered this effect.  In my mind, I round that eventual gain down to zero. 

Of course, if means-testing gets adopted before my retirement, my rounding down to zero will turn out to be justified.  But as far as I can tell, most people round down to zero – even if they plan to retire long before means-testing is likely to be adopted. 

On what basis do I say “As far as I can tell”?  Simple: I’ve never heard anyone – even a fellow economist – claim to factor extra Social Security benefits into a personal cost-benefit calculation.  Indeed, the only time the issue even comes up is in discussions of public policy.  Many people, in contrast, openly discuss how taxes affect their behavior.

While I believe in the power of introspection, I’d definitely like to supplement my introspection with careful empirical research on this topic.  Yet Google Scholar seems oddly empty of articles on the topic.  Question: Is there any scholarly evidence – pro or con – that I’m missing?  If so, please share.