Optimal Taxation and the Domar-Musgrave Effect

In this paper, we find that by replacing the labor income tax by a value-added tax and maintaining a capital income tax, we can improve welfare by a staggering 7%.

It took a long time for this paper to get shaped and published. In 2017, Brendan and I wrote a paper that characterizes the Pareto exponent of the stationary distribution of a Markov multiplicative process with reset. Section 4 of that paper contains an application to the wealth distribution of a Huggett model with constant absolute risk aversion (CARA) utility. We submitted this paper to Econometrica and received a “reject and resubmit”. Reviewers complained that the application generates a wealth distribution with exponential tails, not Pareto tails. Hence, in the revised version, we changed the model to a version of the Angeletos model. When we received a “revise and resubmit” from Econometrica, the editor asked us to cut this application to shorten the paper.

So, the paper became dormant for a while. As I like to recycle materials, I rewrote the paper with a quantitative focus on optimal taxation. Initially, there was just capital and labor income taxes, but when Brendan presented the paper at Duke, John Coleman suggested that we should also include consumption tax (value-added tax). This was an extremely helpful suggestion because the welfare implications changed completely, from a modest (around 1%) increase in consumption equivalent to a staggering 6.6%.

After revising the paper, it was a long shot but we submitted the paper to Journal of Political Economy, and not surprisingly, got rejected. We then submitted the paper to Journal of Public Economics and received a “revise and resubmit”. However, the reviewers asked too many things that we thought were infeasible. Brendan discovered that Economic Inquiry allows the option of “no revision policy” (meaning the paper gets accepted as is or rejected). As we were not willing to spend effort on additional quantitative work, we withdrew the paper from JPubE and submitted it to EI. The review process took only two months and the paper got accepted as is, so I am very happy.

Let me also point out that EI is serious about replicability (so preparing replication files requires effort), which I think is a good thing.