David R. Henderson favorably reviewed Ross McKitrick’s presentation (at IER’s Carbon Tax conference last summer) on the “tax interaction effect,” and then in a follow-up post David pursued Greg Mankiw’s writings on the topic.
In the comments of David’s second post, someone wrote:
If my microeconomic intuition — that taxing harmful externalities will enhance net welfare, and to a greater degree than some regulatory regime — turns out to be wrong, then I wonder what else I might be wrong about? After considering tax interaction effects, should we not be so hard on speeders? Maybe even encourage them? And what about theft, of various sorts? Keeping Coase (rather than Pigou) in mind, I sometimes think about uncompensated environmental externalities as a form of theft. I wonder if theft, generally, might turn out to be beneficial in some CGE model beyond my ability to comprehend.
OK, putting aside what seems to be sarcasm, it’s clear that this guy doesn’t really understand what the TIE [tax interaction effect] claim is. That’s not entirely his fault, because it’s a very subtle issue, and I don’t think the actual economists publishing on the topic have spelled it out in crystal-clear fashion. The more I think about it, the more “obvious” it seems to me, but it has taken me a long time to get here. At IER, I will soon have a post trying to walk through the whole thing, but because I explain it from scratch, it’s pretty long. In the present blog hit, let me summarize the key issues (as I now conceive them) in a hopefully intuitive way, for people who are familiar with the standard Pigovian framework.
(Note: Silas Barta had a great idea when he flipped the “stacking” of the taxes; his analysis clarified my own thinking on this.)
(A) Assume people are annually emitting 30 billion tons of carbon dioxide with no carbon tax. Assume the marginal/average “social cost of carbon” for current emissions is the same, at $30/ton. So if nothing is done, Earthlings are causing $900 billion annually in future negative externalities (present discounted value).
(B) With no other taxes, “optimal” thing to do according to textbook is to levy a $30/ton carbon tax, the receipts of which are distributed lump-sum back to citizens. Assume this makes people cut emissions in half, down to 15 billion tons. There is a reduction in climate change damages of $450 billion. However, the conventional economy is hurt; after all, businesses and consumers are now facing a constraint to use less carbon-intensive techniques. (If someone could magically cure climate change, the carbon tax would obviously hurt the economy, so it’s still having that effect now, even though we’re superimposing a stabler climate on top of the hurt economy.) Suppose that the harm to the conventional economy from the optimal carbon tax is $100 billion. Thus the net benefit of the carbon tax is $350 billion; this is the maximum, by the way, because we’re choosing the $30/ton tax optimally. If we had made it $31/ton, then climate change damages would be only $440 billion but economic damages would rise to $111 billion, putting the net gain at only $349 billion. Going the other way, if we had only set the carbon tax at $29/ton, then climate damages would be $461 billion while economic damages would be $90 billion, for a net gain of the carbon tax of $349 billion. So the optimal carbon tax of $30/ton–which is the “social cost of carbon”–maximizes the net benefits of a carbon tax, which makes sense.
(C) Now re-do the analysis, but this time suppose there was an income tax. We all know income taxes are distortionary. Suppose this income tax (considered by itself, without a carbon tax) caused $1 trillion in deadweight loss on the economy. We also know that if you increase the income tax rate, then the deadweight loss increases more than proportionally. So, if you are going to implement this identical income tax code, not starting from scratch, but on top of a carbon tax of $30/ton–which itself is already causing $100 billion in harm to the economy–then the marginal deadweight loss of the income tax code is $1.2 trillion, rather than the $1 trillion if there is no carbon tax. Thus, the “tax interaction effect” here is $200 billion. This is in addition to the $100 billion harm to the economy coming directly from the carbon tax. The tax interaction effect’s $200 billion harm is an indirect effect of the carbon tax, because it makes the deadweight loss from the income tax that much higher.
(D) With the above numbers, it’s still better to impose a carbon tax at $30/ton than to do nothing at all. Note that we are still returning the receipts of the carbon tax back to citizens in lump-sum checks. By imposing that tax, you reduce climate damages by $450 billion, while the economy suffers damages of $100 billion (direct hit) plus $200 billion (tax interaction effect), for a total hit to the economy of $300 billion. But $450 bn – $300 bn = $150 billion, which is the net gain of doing the original $30/ton carbon tax.
(E) Ah, but the $30/ton carbon tax is no longer optimal. On the margin, the carbon tax is causing more damage in the presence of the income tax, than it would in a textbook scenario with no other taxes. So if we reduce the carbon tax down to, say, $15/ton, we now achieve the optimum of a (say) $220 billion gain in welfare, from imposing the tax. This of course is lower than the $350 billion gain we achieved in (B). This makes sense too; since we have to worry about the carbon tax exacerbating the deadweight loss of the income tax, it is no longer as useful a tool, so the total social gains we can derive from it are lower than in a scenario with no tax interaction effect.
(F) Ah wait, we can actually do something better. We can use the receipts from the new carbon tax to reduce the income tax rates. In the absence of the carbon tax, reducing the rates of the income tax obviously reduce its deadweight loss; this is the “revenue recycling effect.” So now, in addition to its beneficial effects in terms of reducing emissions and hence climate change damage, the carbon tax has two other effects that cut in opposite directions: The tax interaction effect makes the income tax worse, pushing us to levy the carbon tax at less than the social cost of carbon. But the revenue recycling effect makes the income tax better, pushing us to levy the carbon tax at more than the social cost of carbon. Which effect dominates? This is an empirical question, but the intuition on this particular point (as to whether TIE <> revenue recycling) is that a carbon tax has a base smaller than a typical income tax, and so on that score you would think intuitively that raising $1 billion from a carbon tax is more harmful to the conventional economy than raising $1 billion from an income tax.
(G) In light of all the above, it is probably the case that (given our other numbers) the new carbon tax, if levied on top of the pre-existing income tax but with all carbon tax receipts being used to reduce income tax rates, will have an “optimal” rate lower than $30/ton but higher than $15/ton. Note that such a number is also higher than $0/ton. Thus, even after you’ve absorbed the standard tax interaction effect analysis: It is still true that a negative environmental externality can be fixed with a Pigovian tax, and it is still true that using carbon tax receipts to reduce other income tax rates (rather than increasing government spending, or even returning receipts back, lump sum) is “good for the economy.” However, it is NOT true that the presence of a distortionary income tax bolsters the “case for a carbon tax,” and it is really really not true that if you disregard environmental benefits, you can still get a “pro-growth” tax reform by implementing a carbon tax and reducing income taxes accordingly.