Yesterday we kicked off the 24th NYU Tax Policy Colloquium, with Stefanie Stantcheva of the Harvard Economics Department presenting Taxation and Innovation in the 20th Century. Here are some of the main points that occurred to me about the paper. (I don’t comment here about the discussions, in order to preserve their off-the-record status – not that this often matters, but so that no one in attendance will ever need to worry about this question.)
1) Strong empirical results – This paper is a major success (for Stantcheva and her co-authors: Ufuk Agcigit, John Grigsby, and Tom Nicholas) because it generates strong and robust empirical findings, using new datasets that permit them to examine how state and local personal and corporate income taxes since 1920 may have affected innovation, as defined by patent quantity and “quality” (measured by citations). They find very significant tax elasticities for innovation quantity and quality, as thus measured, that are robust across a range of different specifications. And while a lot of what they find appears to be “business-stealing” – i.e., mere shifting of activity from one tax jurisdiction to another – they conclude that this is not the entire story; there appear also to be (lesser) effects on overall activity levels.
While the paper is science not advocacy, it strikes me as potentially important for a broader set of tax policy debates. A whole lot more would be needed to support some of the implications that I discuss below, and the authors make no claim that such support is likely to be found, but it’s nonetheless worth spelling out here why I think this line of research is of broader interest.
Other empirical results in the paper that are of interest include (a) its comparison of the effects of corporate versus personal income tax rates, and (b) its finding that agglomeration effects, such as the concentration of IP researchers in Silicon Valley, reduce the tax elasticity of the measured behaviors.
2) Contra Diamond and Saez, Ocasio-Cortez, et al? – Our era of rising high-end inequality has helped to shift the Overton window for academic debate, and perhaps even public political debate, regarding how high marginal rates at the top should be.
Optimal income taxation (OIT), a literature to which Stantcheva has (elsewhere) made significant contributions, was long thought to suggest having relatively flattish rates, for what is essentially a technical or logical reason. OIT aims to maximize a measure of social welfare, typically based on utilitarian or other summations of individuals’ welfare, by trading off redistributive benefits (from declining marginal utility or a pro-egalitarian aggregation rule) against efficiency costs. However, the fact that rate brackets at the very top of the income distribution are inframarginal (because one is guaranteed to be above them) for so small a proportion of the people subject to them, as compared to rate brackets lower down the distribution, pushes against steep rate graduation. E.g., if I know that I will be earning $1 million or more, and the choices that I am considering relate only to the question of how much more, then rate brackets for the first $1 million of my income have no distortionary effects (i.e., only income effects) on what I decide to do.
An influential paper from several years back, by Peter Diamond and Emmanuel Saez, nonetheless used an OIT methodology to support top marginal rates as high as 70%. Their conclusion relied heavily on (1) empirical evidence of low short-term labor supply elasticity, which suggested that the deadweight loss from so high a rate at the top might not be so great, and (2) a view that any lost utility to people at the very top from the high rates would be subjectively, and/or as a matter of social welfare function weighting, indistinguishable from zero. Under the Diamond-Saez paper’s model, one therefore might choose the revenue-maximizing rate for people at the top of the income distribution. One still, however, wouldn’t want a more than revenue-maximizing rate.
An admitted problem with this conclusion was our lack of good knowledge about long-term labor supply elasticities. However, one may derive additional support for very high rates, and indeed for more than revenue-maximizing rates at the top, if one believes (as I do) that extreme high-end wealth concentration can have negative externalities, e.g., by reason of its (in the words of an NYT op-ed today by Saez and Gabriel Zucman) undermining "democracy against oligarchy” and creating “corro[sion of] the social contract.” Under that view, extremely high wealth and income concentration at the top can be like pollution, which a Pigovian tax would price at marginal cost rather than aiming at revenue maximization.
As Paul Krugman and Matthew Yglesias, among others, have noted, this academic research underlies Alexandra Ocasio-Cortez’s recent call for a 70 percent top marginal rate, which I view as desirably expanding the Overton window for public policy debate, although I wouldn’t go all the way there myself for reasons that it would be too digressive to address here.
So, how is the Stantcheva paper on innovation’s tax sensitivity relevant to all this? The answer is that it could point the way towards a new line of argument against very high marginal rates, in particular at the top where one might expect successful IP entrepreneurs to find themselves. The point here is not just that innovation, as measured by the paper, appears to be highly tax-responsive, but also that there are arguments for its having great social value.
3) Is “innovation” special? – In my days (back in the mid-1980s) as a Legislation Attorney at the Joint Committee on Taxation, I had a colleague whose area of responsibility included the R&D credit. He personally hated the credit (not that he could do anything about this preference), despite the fact that our economists loved it, at least in principle, because they viewed R&D activity as having positive externalities. He liked to say, in support of his view, that too much of the activity that would end up qualifying for the credit was of the character of, say, McDonald’s or Burger King working on their sesame seed buns.
But even leaving aside knowledge spillovers from businesses’ research that may be more consequential than improving the sesame seed bun, we know there is some patentable innovation that appears to have strong positive externalities. Consider iPhones, or else advances in medical treatment that prolong people’s lives and wellbeing. There is “innovation” out there that (a) creates new consumer surplus, in that the subjective value to people of a newly available item exceeds its market price, and/or (b) increases workers’ productivity, thereby permitting to reach higher levels of market consumption plus leisure than were previously available to them.
Insofar as high taxes reduce the amount of such socially valuable innovation, we have positive externalities from low rates to think about, not just the negative externalities that Saez and Zucman invoke. Thus, a finding that socially valuable innovation is reduced – not just shifted around between jurisdictions – by higher tax rates would complicate the OIT analysis and push back against the Diamond-Saez-Zucman / Ocasio-Cortez approach.
This is very far from being established by the Stantcheva paper (as Stantcheva herself would be the first to agree). But it helps to show why further research, along the lines that this paper significantly (albeit still preliminarily) advances, is of very broad interest.
One should also note the possibility of negative externalities that might be associated with innovation – e.g., if it increases high-end inequality, creates welfare losses from Schumpeterian “creative destruction,” and/or involves strategic and defensive patenting, patent trolling, etc. But in any event the paper helps to point out the need to think more about possible spillovers from “innovation,” as well as about long-term labor supply elasticities, when debating top marginal rates.
4) The paper’s quantity and quality measures of innovation – In keeping with much other IP literature, the paper uses patents as a measure of innovation quantity, and patent citations as a measure of innovation quality. This is inevitably open to challenge and imperfect, as a rich vein of IP literature has been exploring.
5) Policy implications if high tax rates reduce socially valuable innovation – Even if the paper’s possible and still speculative broader policy implications are confirmed, there is a whole lot of instrument choice to think about. E.g., lower rates vs. exempting the normal return to innovation vs. targeted subsidies for innovation vs. stronger IP protections. Each tool might have its own set of tradeoffs.
6) Fiscal federalism issues – Some of the positive spillovers that innovation might yield may operate at the global level. E.g., insofar as iPhones improve people’s lives, such improvement is not conditioned on the iPhone’s having been invented within one’s own taxing jurisdiction. So just as any one country, if it is acting purely selfishly and unilaterally, lacks the incentive to impose carbon taxes that are priced at global marginal cost, so it might be expected to disregard or at least undervalue innovation’s positive spillover effects on outsiders.
On the other hand, “business-stealing” may be locally beneficial even if not globally so. Thus, local incentives to encourage innovation might end up being either too high or too low in the aggregate, but seem likely to be misdirected from a global social welfare standpoint.
I’ve commented here mainly on some of the broader implications, albeit as still unknown, that cause “Taxation and Innovation in the 20th Century” to be of especial broader interest. But I should also note one more time its yielding strong and robust empirical findings, of a sort that researchers more commonly wish for than achieve.