Thread by @SHamiltonian

Twitter has been all about the wealth tax lately. Some have claimed that all capital taxes are bad. Others that Warren Buffett’s lower tax rate than his secretary is wrong. Underpinning it all is the theorem I think about more than any other: Atkinson-Stiglitz. A thread. 1/22

I teach A-S three times a semester: once when comparing the efficacy of income and consumption taxes, another when considering the efficacy of capital taxation, and last when considering the efficacy of estate taxation specifically. It’s a lot of A-S. But it’s important! 2/22

First, income taxes. The 1970s were all about optimal income taxation. (So much for inequality being a “new” field of study 🙄). We model inequality as a type distribution, which generates a wage distribution. Workers then choose their labor supply given the wage and taxes. 3/22

As you raise the marginal tax rate at a point it discourages workers at that point but taxes lump-sum all workers above. The optimal tax schedule depends on that response, as well as the income distribution and the welfare weights assigned to workers across the distribution. 4/22

The idea is that the nonlinear income tax can be set optimally to balance redistribution and efficiency concerns. The key paper on this is Mirrlees (1971), for which along with other contributions he won the Nobel Prize in 1996. Sadly, he passed away last year. 5/22

So what do A-S have to do with this? They asked a natural question: if we have the optimal income tax, is there any additional need for a consumption tax? If you think about it, it’s pretty weird to tax your money both on the way into and on the way out of your bank account. 6/22

Their answer was “no”... under some assumptions:

1) the income tax really is optimal;
2) labor/leisure and consumption are neither complements nor substitutes; and
3) everyone has the same preferences.

If you want to justify having both taxes, you must break one of these. 7/22

Saez (2002) showed that if people consume different things, then consumption taxes can target “high types”. Imagine two people on $1k/week, one on $25/hr and one on $50/hr. Who is more likely to buy caviar? Conditional on income, can consumption reveal something to target? 8/22

And there have been many others. Now you’re probably thinking: what on earth does this have to do with capital taxation? Well that’s interesting. On an infinite time scale, all savings must be consumed. So capital is simply consumption in the future. So should we tax that? 9/22

A-S say: “no”... under some assumptions. If we have an optimal labor income tax but also tax capital, then we are discouraging saving (and investment: see Chamley-Judd), which generates a deadweight loss, but for no redistributive benefit. Same applies to estate taxation. 10/22

The Saez (2002) argument would say: do savings patterns, conditional on income, reveal something about peoples’ “types” that we can target over and above their observed income? Maybe. But in general I find a lot of these kinds of arguments a little delicate or esoteric. 11/22

To my mind, the critical violation of A-S is that we certainly do not have the optimal nonlinear income tax. It’s obvious to me that the income tax should be more progressive. More problematic though is the inability of the IRS to distinguish labor from capital income. 12/22

If we believe Warren Buffett has unique qualities that generate super-normal returns, then those are part of his type. Those returns are to his labor, not capital. But he receives them in a form that makes them appear to be capital income. The same is true for other CEOs. 13/22

So while we might like a zero tax on capital income and the optimal nonlinear labor income tax to target the wealthy, in response they will take their labor income in the form of capital income and we won’t, in effect, have the optimal income tax after all. A-S breaks. 14/22

That means the optimal tax rate on capital income isn’t zero. But because there are limits to the ability of workers to relabel labor income as capital income, it won’t be as high as the optimal labor income tax rate either. So the key task is to find the right discount. 15/22

We preferentially tax dividends and capital gains, and that seems consistent with this. Of course, the discount we apply to capital gains by way of allowing inheritances to escape taxation entirely (i.e., a zero capital income tax rate) is too generous. So we should fix it. 16/22

Where does a wealth tax fit in? Well, it’s hard to justify in the usual optimal tax framework, on top of optimal taxes on labor and capital income. But there is one justification consistent with the above, and that’s to make up for insufficient redistribution in the past. 17/22

If for years we haven’t had the optimal nonlinear income tax, as I suggest, then that will have allowed the wealthy to accumulate excess wealth. This calls for a one-off tax to correct for past mistakes and move forward by properly taxing the flows. But I said “one-off”. 18/22

We already have a one-off tax that targets the wealthy and makes up for insufficient taxation over their lifetimes: the estate tax. Going back to the well year after year is impossible to justify if we simply correct the way we tax estates and capital income going forward. 19/22

That’s why I agree with people like @wwwojtekk and @LHSummers who have advocated focusing our efforts on improving our existing taxes. It is not about redistribution vs no redistribution; it is about the best way of achieving a desired degree of redistribution. 20/22

So that’s how I’ve been thinking about wealth taxes. There are a range of problems I haven’t discussed with taxing the stock of wealth year after year and myriad implementation concerns that make the case against even stronger. But A-S provides a nice conceptual foundation. 21/22

I hope you enjoyed the thread, and perhaps learnt something interesting if you’re not a tax person! Feel free to comment along the way with any links or cites I might have missed. Hopefully that will be it from me on wealth taxes for a while!

✌🏻

22/22