Reihan Salam replies

The view that higher tax rates have very little effect on the economy will continue to persist, regardless of the arguments and evidence we bring to bear. I do recommend reading Arpit Gupta’spost on reconciling Edward Prescott’s macro estimates with the work of Raj Chetty et al. on Danish tax records:

On some level that is of course right but I do think there is room for constructive debate. As a point of evidence, I would offer that I used to believe the exact opposite of what I know believe, so some change is possible.

Also, what I think of as a very GMU-y approach to the subject helped solidify my present perspective. That is, say the model we are thinking about is true. What does that tell me about the world generally? Do those observations seem to be true?

I’ll note that this cuts opposite ideological ways on two big economics controversies.

First, the minimum wage. Its just going to take a lot of work to convince me that the minimum wage does not depress employment opportunities. Its not impossible, but you are climbing a steep hill and to my mind Card and Krueger and their intellectual decedents just haven’t gotten there yet. They might get there but there is a serious “lying eyes” problem.

The second, is the issue at hand – marginal tax rates and labor supply. Here the hill is likewise large and the evidence so far weak. You have to beat back enormous amounts of introspection, basic observations of human behavior and 200 year trends in work habits. That’s a big hill.

The evidence presented so far just doesn’t get very far up that slope.

Let me talk a little bit about Prescott’s work because its similar to stuff I have done myself. Chetty’s paper is a tour de force and the issue of frictions is important. However, its not clear to me that in the case of marginal taxation there is really anything to be explained.


Prescott uses the following utility function


Which is similar to the one I model with. The separable log is attractive because it the right properties and is analytically relatively easy.

However, with this function almost all of the effect of a larger government sector is done through government spending or in modeling terms Transfers (T). In his later equation it is this element.


If you work the problem that way then you are basically pushing down the income effect through the use of the term T.

The built-in assumption is that the government taxes money away from you and then gives it right back to you. So, the net effect is essentially only the substitution effect.

If I remember correctly following is true:

Once you have chosen to use separable log utility the income and substitution effects necessarily cancel out. That is, an increase in the real wage must increase consumption and leisure by the same out. The distribution of time between consumption and leisure is determined only by (1) the co-efficient on the leisure’s log term, in this case ALPHA; (2) the amount of hours in your endowment, in this case 100; and the exogenous income, in this case T.

The relative price between labor and leisure cancels out.

Now all of this is actually the origin of my position that its government benefits, not taxes that cause the decline in labor supply. However, when you start to think of this in a human context it makes sense as well.

Its also consistent with broad observations about taxes raised to in an effort to cut deficits. Since, in that case the labor reducing effects of Transfers would have already occurred.