I am not a taxation expert but I would like to suggest an interpretation of the @gabriel_zucman - @wwwojtekk
diatribe (a good summary of links here by @Undercoverhist: ). Hope this helps the few readers patient enough to follow /1
diatribe (a good summary of links here by @Undercoverhist: ). Hope this helps the few readers patient enough to follow /1
TL;DR: The question "Who pays how much taxes?" is
ill-posed. Better question: "who pays how much taxes RELATIVE TO WHAT SCENARIO?" There is an even better question but I'll write it later /2
ill-posed. Better question: "who pays how much taxes RELATIVE TO WHAT SCENARIO?" There is an even better question but I'll write it later /2
Zucman/Kopczuk are ultimately debating over which counterfactual makes more sense. But counterfactuals are a bitch... I thought a lot about counterfactuals in my research. They require clarifying what is changing and what is fixed. That is almost never obvious without a model /3
"Who pays how much taxes" compares a baseline world with taxes with a counterfactual world with no taxes. Taxes change, everything else remains fixed, right? Wrong. What is Zucman fixing? What would Kopczuk fix? I will describe two of the most controversial points /4
A) Should we fix EITC at baseline levels? Zucman says yes because EITC is a transfer. If not: slippery slope: why not also change other expenditures, harder to allocate? He would have to make assumptions about how to allocate even defense expenditures: very difficult! /5
Zucman's counterfactual is to change taxes only, and the EITC is formally a transfer, therefore he keeps it fixed. He adopts the legal definition of EITC: this has the advantage of clarity and simplicity /6
@LHSummers/@Kotlikoff and others find this silly, because EITC is directly tied to taxation - calling it a transfer and not a tax-rate change is semantics. My issue instead is with the nature of the counterfactual: in a world with no taxes how are you paying for the EITC? /7
Zucman's counterfactual is interesting but IMHO with little practical meaning from a distributional-policy POV: it's hard to set income taxes to zero without changing at least transfers directly connected to them. But ok, this is opinable /8
B) How about tax incidence? Zucman allocates the corporate tax to shareholders. His counterfactual sets these taxes to zero and assumes that in the baseline they are paid by those cutting the check to the govt. This is clean but we teach undergraduates differently. /9
@wwwojekk, @LHSummers, @Kotlikoff and others argue that corporate taxes are partly paid by workers, owners of other forms of capital, etc... Their counterfactual in this case is even harder than in A) because to account for incidence, more hypotheses are needed. /10
What people really pay depends on elasticities. Quantifying them requires models, functional forms... This counterfactual needs fixing the model and its parameter estimates. What about GE effects? Do we ignore (fix) them? /11
Zucman chooses fewer assumptions but his results are less relevant because they map into a less realistic world and an answer to irrelevant question. We should not care about the taxes paid by those that formally send the checks /12
This reply ignores the real issue. Of course we care about how taxes affect wages! I am not even taking into account welfare (leisure?), dynamics and political-econ issues raised by others discussing about wealth taxes (@fatihguneven, @IvanWerning). /13
Because, IMHO the meaningful, important question[s] is not "who pays how much?", but "how does the consumption distn' change when taxes change [fixing everything else, assuming balanced budget, in partial or general eq...]". /14
To conclude: counterfactuals vary some parameters, keep others fixed, and keep implicitly fixed a framework that governs what happens under the two scenarios: the structure of the model, other institutional details, functional forms.../15
Debates arise not only when people have different counterfactuals in mind, but also when people have altogether different frameworks in mind. I.e. it's not just a matter of what parameters to change (e.g. issue A) but also what model governs outcomes given parameters (B) /16
Economics is hard. If one cares about distributional consequences the counterfactual must be appropriately designed. Zucman and coauthors are doing great data work that is much needed to answer meaningful questions about the distributional effect of taxation.
Obviously that is not the only possible goal. In any case Z. et al. provide a much needed first step but I liked @wwwojtekkk's call for humility in selling results to the press and the public (livestream.com/sipa/events/88…). I hope this clarifies why /18
PS: This episode of @elucidationspod podcast may
help clarifying the conceptual challenge of thinking about counterfactuals lucian.uchicago.edu/blogs/elucidat… /END
help clarifying the conceptual challenge of thinking about counterfactuals lucian.uchicago.edu/blogs/elucidat… /END
@ElucidationsPod gee I mistyped @wwwojtekk 's at least a couple of times. Well you know who he is.
