Wednesday, November 13, 2019

Tax policy colloquium, week 11: empirical paper on Belgian IP box rules

Yesterday at the colloquium, Stacie LaPlante presented The Effect of Intellectual Property Boxes on Innovative Activity and Effective Tax Rates (coauthored by Tobias Bornemann and Benjamin Osswald, both former students of mine as I teach a mini-course every 3 years at Vienna University's DIBT program, from whence they both recently graduated).

This paper looks at the patent box that Belgium enacted, effective 2008, which presented a nice research opportunity due to its design. It was for patents only, not other IP, making empirical measurement easier, Plus, it required new patents, and gestured in the direction of requiring activity in Belgium, although as we'll see this may not have been much more than a gesture.

It also was exceptionally generous. Belgium had a 34% corporate income tax rate (okay, strictly speaking, 33.99%, but for a corporate income tax "99 pricing" strikes me as a bit idiotic and pointless). The patent box retained the full tax rate on the deduction side, and provided an 80% exclusion for gross income (generally) from patents.

Thus, suppose one spends 100 developing a patent that ends up earning 70. The former can presumably be expensed as R&D, leaving 66  after-tax given the 34% rate. The latter is taxed at 6.8% (20% of 34%), leaving 66.17 after-tax. So the 30 percent pre-tax loss becomes an after-tax gain, under the Belgian patent box. That is not exactly ungenerous.

As background before discussing the paper's empirical findings, why would one have a patent box? It's one mechanism among many for increasing "innovation" that is thought to have positive spillovers. These might be of two main kinds: (1) the global benefit from increasing knowledge that leads to further knowledge expansion, practical applications that benefit people, etc.; (2) local spillovers from having the activities take place in one's own jurisdiction. Here the idea is that everyone wants their own Silicon Valley, on the view that it enriches and otherwise benefits the jurisdiction and its residents.

Alternative ways of increasing valuable innovation include (to name just two among many) (1) patent law and other associated legal protections, and (2) up-front tax benefits, such as R&D expensing or credits, perhaps made refundable so that innovators can benefit even if they don't have current net income.

For real world patent boxes, the particular motivations might include (1) nobly and disinterestedly wanting to benefit everyone in the world by at least slightly and incrementally increasing global innovation activity, (2) more selfishly (and via tax competition) aiming to become the host of a new Silicon Valley, (3) revenue piracy - which I don't mean to condemn via the label - meaning that one gets patenters to assign legal and tax claims to one's jurisdiction so that both they and oneself will benefit - less global taxes for them, more revenue for oneself than if it hadn't offered accommodation services, and (4) simple Ramsey pricing, whereby one lowers the tax rate (for efficiency reasons) on activity that is relatively mobile and thus elastic.

As we'll see, the paper's findings reinforced my view that, while in principle countries can benefit from offering patent or broader innovation boxes, on one or more of grounds 2 through 4, in practice this never seems actually to be the case. This in turn leaves a question that I'll address at the end: Why, then, do patent boxes seem to be so popular with national policymakers. But first, let's look more closely at the paper.

The paper has 6 main empirical findings, each of which I'll accompany here with my own commentary.

1) Effect on patent applications and grants - Using a difference-in-difference research design and with multiple controls, fixed effects, alternative specifications, entropy balancing, etc., the paper finds that the Belgian patent box increased patent applications by 0.4 to 1.8%, and patent grants by 0.4 to 5.1%. This is consistent with concluding that the patent box increased innovation activity in Belgium, although this could of course involve shifting from other countries, rather than new activity.

There's a vast IP literature, making the point that it's tricky to go straight from more patents to more spillover benefits from innovation, for a number of reasons. For example, strategic patents and greater activity by patent trolls often are not good things. But, in a study like this, increased patent applications and grants is verging on necessary, even if not sufficient, to suggest a strong case that there might be increased innovation, at least as to that which is being (perhaps formalistically) assigned to the country with the patent box.

The paper aims to offer only a lower bound on the response. For example, its panels exclude firms that weren't around for all of the years under study, thereby omitting the creation of new firms in response. But even granting that, the positive response, while unsurprising given the very substantial benefits Belgium was offering to patents that ended up earning gross income, seems rather small.

2) Effect on patent quality - Combining several standard measures of this that are used in the IP literature, the paper found a decline in patent quality by reason of the patent box (although the effect's magnitude is hard to quantify, given the squishiness of the metric). This is hardly surprising under a design that allows investments earning a 30% pretax loss to be profitable after-tax. Of course, if profitability rather than spillovers were the sine qua non for the patenting activity that one wants to encourage, there'd be no need to do more than, say, address liquidity problems. Still, the provision's allowing a tax arbitrage between deductions at 34% and gross income taxed at 6.8% seems unlikely to be a strong positive inducement to "quality" of any kind.

3) Intra-Belgian shift in firms' use of employees with college degrees - The paper found a quite substantial increase - in contrast to its generally otherwise modest results - in the extent to which Belgian firms that could make practical use of the patent box, as distinct from those that couldn't, increased their relative levels of employment of individuals with a college education. The paper uses this just as an indicator that something is happening in these firms.  If they were just mailing patent applications to a Belgian rather than non-Belgian address, there might be no need for greater relative use of college grads. But the data don't permit analyzing whether these were, say, engineers or tax planners.

From a nationally self-interested standpoint, it might be great for Belgium if its enacting a patent box induced a good number of educated, high-value employees to move to Belgium from, say, neighboring EU countries. The tax revenues alone might be significant in such a case (although they wouldn't be scored under Belgium's corporate income tax). But data limitations made it impossible to test for this intriguing possibility. But that said, if I were trying to attract high-skilled workers from neighboring countries, I rather doubt that a patent box is where I'd start out.

4) Rate of increase in patent activity vs. other EU countries - The paper compared Belgian patent trends to those in 3 peer countries in the EU: Germany, France, and Sweden. Germany and France are of course both neighbors, and neither changed their rules with respect to patent boxes during the period under study. Germany never had a patent box during the period, while France always did. Sweden is included because of similarities to Belgium in terms of overall size and that of its IP sector.

The paper finds that Belgium's highest rate of relative increase in patent activity pertained to Germany, as opposed to France or Sweden. While the cause and significance of this can't be nailed down definitively, I view it as consistent with (and perhaps mildly supportive of) a switching story. Germany was the one country of the three that combined being adjoining with not having its own patent box regime.

5) Relative effective tax rate (ETR) effects - Overall, the paper found that firms taking advantage of the patent box saw their ETRs drop by 2.2 to 2.4% absolutely, or 7.2 to 7.9% relative to their prior ETRs. However, the degree of average benefit varied by the type of firm. It was highest for multinational companies (MNCs) that had limited profit-shifting opportunities out of Belgium, intermediate for MNCs that had profit-shifting opportunities (and that thus had already been shielded from actually paying an ETR in the ballpark of Belgium's 34% statutory rate), and lowest for firms that were Belgian only.

I would think it's reasonable to presume that the MNCs were predominantly owned by non-Belgians. After all, they're presumably on at least EU-wide or even global capital markets, and Belgium is too small for one to think that its residents would typically own a high percentage. By contrast, I'd presume that the Belgian firms were mainly owned by locals. So the tax benefit was going far more to companies whose shareholders were non-Belgians, than to those whose shareholders were Belgians.

6) Revenue effect - The paper estimates that the Belgian patent box resulted in a revenue loss of €68 million per year, representing 0.63% of Belgian corporate income tax revenue.

This is a very bad result, from the standpoint of the provision's merits from a Belgian national welfare standpoint. It means that, far from representing successful revenue piracy - or, to put it more charitably, hitting the peak of the Laffer curve, the provision is actually losing money. So rather than making money out of being an accommodation party, Belgium is paying.

One needn't generally demand of tax breaks that they better than pay for themselves. But here it's very plausible that they should. Again, this money is coming mainly from non-Belgians, as to whom it's plausible that Belgium benefits from revenue-maximizing, subject to modification by reason of positive externalities created by luring patent activity inward. It's plausible that the incidence of the tax cut actually lay with the MNCs' predominantly non-Belgian shareholders, given the recent prevalence of extra returns (presumably reflecting market power) to MNCs generally, and those involved in IP activity particularly. For Belgian residents, by contrast, one would have Ramsey tax motives of trading off revenue against deadweight loss, and thus of adopting more efficient tax levels even if one thereby loses revenue. But deadweight loss incurred by non-residents is normatively irrelevant under a selfish national social welfare function.

This pushes pretty far towards one concluding that Belgium hurt rather than helped itself by adopting the patent box. Again, global positive spillovers might not do much for Belgium in particular even if one did not suspect that there's more switching than fresh innovation activity in response to the provision.

But what about the local spillovers? I suspect that this would require a lot more substance than the Belgian patent box appears to demand. To qualify, the taxpayer must have a Belgian "qualified research center" (QRC). But this need only be Belgian-owned, which can be close to meaningless given the flexibility of corporate residence. And even if the QRC is actually in Belgium, it need only be sufficiently capable, staffed, etc., to "supervise" the research that's going on. Maybe this calls for renting an office that's staffed by a few college grads who look over the docs, but I seriously doubt that it calls for anything close to even requiring the first faltering steps towards the true establishment of a Belgian Silicon Valley.

So Belgium appears to have been losing tax revenues, giving the money to foreigners, subsidizing projects that might even have had significant expected pre-tax losses, and failing to encourage any significant local positive spillovers (apart, perhaps, from encouraging the firms to hire a few college grads who might instead have been toiling in some other office in Brussels or wherever).

This appears to be a very typical story regarding patent boxes. So why they are so popular? I'm not cynical enough, at least in this particular case, to attribute it mainly to lobbying and interest group influence. Rather, the thing sounds trendy and new ("wow - a patent box - where do they keep the darned thing?"). So it might mainly be self-branding by policymakers who want to associate with something that sounds hard-headedly cool, even if it actually isn't. 

Law school naming gift?

Listening to Daft Punk and Diana Ross's Inside Out on an elliptical machine at the health club this morning, it occurred to me that, if I had a few billion dollars lying around, I'd need to consider a naming gift to establish the Nile Rodgers School of Law.

Monday, November 11, 2019

P.G. Wodehouse and Neil Young: decades apart, but geographically almost together

P.G. Wodehouse was living in what is now the Washington Square Hotel, two short blocks away from NYU Law School, when he invented the character of Reggie Pepper, the prototype for Bertie Wooster.

When you add to this the famous Neil Young album cover photo for After the Gold Rush, showing him striding along the fence on the west side of the main law school building, it becomes clear that at least two major, albeit admittedly heterogeneous, cornerstones of my personal aesthetic have historical roots here.

Thursday, November 07, 2019

Revised version of my source / digital service taxes paper

I have posted on SSRN a revised version of my paper, "Digital Service Taxes and the Broader Shift From Determining the Source of Income to Taxing Location-Specific Rents." In a broad sense, it's basically the same as the earlier draft. At the same time, however, due to numerous helpful comments that I have received (as credited in my acknowledgement up front), I also feel it's significantly improved.

You can find it here.

Wednesday, November 06, 2019

Illustrating stylized normative views of tax policy

The Fleurbaey paper that we discussed in yesterday's NYU Tax Policy Colloquium describes 4 normative views that it proposes to embody in social welfare functions. Since these views each have some following or potential plausibility, it might be useful (or at least interesting) to set out how these views might apply to a toy hypothetical involving an 8-person society. Hence the following:
DESCRIPTION
WAGE RATE EX ANTE
WAGE RATE EX POST*
HOURS WORKED
INCOME
A Talented, lucky, hard-working
100
150
40
6,000
B Talented, unlucky, hard-working
100
 50
40
2,000
C Talented, lucky, lazy

100
150
 4
  600
D Talented, unlucky, lazy
100
 50
 4
  200
E Low-talent, lucky, hard-working
 10
 15
40
  600
F Low-talent, unlucky, hard-working
 10
   5
40
  200
G Low-talent, lucky, lazy
 10
 15
  4
    60
H Low-talent, unlucky, lazy
 10
   5
  4
    20

*Wage rates ex post differ from ex ante because each individual makes an irreversible occupational choice. This either pays off and yields a 50% increase in the wage rate, or backfires and results in a 50% reduction. The ex ante wage rate is an expected value prior to one’s making this choice.
Utilitarian: Absent incentive effects, equalize everyone. But may need to consider incentive effects on both wage rates ex post (if dependent on choice under uncertainty but some information) and hours worked.
Resource egalitarian (Dworkin version): Wage rate ex ante is brute luck. Suppose we agree that wage rate ex post and hours worked are option luck. In that case, only want to address ex ante differences.
John Roemer 1996 (from Theories of Distributive Justice): Same as resource egalitarian except treat option luck the same as brute luck when not effort-related. OR, same as utilitarian except for effort level. Want to equalize for ex ante AND ex post differences, but not hours worked (= effort level). Note: This is Roemer 1996 as viewed through the filter of Fleurbaey & Maniquet 2018; no guarantees that the actual John Roemer would agree with it.
Libertarian: Don’t want to equalize anything. So no transfers & also no tax, except to fund public goods, based on benefit that might (???) have something to do with income levels.
FIRST-BEST DISTRIBUTIONAL OUTCOMES, IGNORING “SLAVERY OF THE TALENTED” ISSUE
Suppose no public goods to fund, no incentive issues, labor supply is fixed (e.g., the state can’t command people to increase their hours), and full information regarding not just income but wage rates ex ante and ex post, and hours worked. Then:
Utilitarian: Equalize everyone by dividing up the $9,680 of total income so each individual gets $1,210.
Resource egalitarian: Equalize between people with different ex ante wage rates but the same ex post luck and effort levels. So A and E should split their $6,600 ($3,300 each). B and F should split their $2,200 ($1,100 each). C and G should split their $660 ($330 each). D and H should split their $220 ($110 each).
Roemer 1996: Equalize between people with the same effort level. So hard-working A, B, E, and F split their $8,800 ($2,200 each). Lazy C, D, G, and H split their $880 ($220 each).
Libertarian: Leave everything as is.
“SLAVERY OF THE TALENTED” ISSUE
The above took labor supply as given. Suppose we continue to ignore incentive issues, but allow for the possibility that the state could command individuals to work more hours. Then there’s a possible implication that the utilitarian, at least, would consider commanding people with high ex post wage rates to work longer hours, so as to fund greater transfers to everyone. These, too, would be split evenly, unless working longer hours (via command) affected the marginal utility of a dollar for those subject to the command. This possibility might make one uneasy. Ronald Dworkin dubbed it the “slavery of the talented” problem.
Within the utilitarian framework, the only way to rule out the problem (if one is not willing simply to embrace it) is to posit that the utility losses from being thus commanded would exceed the utility gains. In a very simple framework, however, the only utility loss would be from reduced leisure, as distinct from the indignity, etc., of being thus commanded to work longer.
Because the other frameworks are less committed in advance to a determinate framework, they can – for better or worse – accommodate an ad hoc (which is not to say necessarily unreasonable) presumption or side-constraint to the effect that we rule out doing this. This, of course, leaves the question of what underlying meta-framework one is using to determine the set of desirable side-constraints. Arguably, the desirability (if one agrees to it) of this side constraint does not necessarily dictate adopting the other normative frameworks’ approaches to other issues, such as what we think of option luck and/or low effort levels. Note that a utilitarian might also more readily accommodate than the others the view that “low effort” is merely an anodyne example of commodity choice, i.e., preferring leisure to work and market consumption, just as one might have a preference between ice cream flavors.
SECOND-BEST DISTRIBUTIONAL OUTCOMES
Suppose that we can only observe income (and perhaps the overall statistical distribution of types), as opposed to the distinct breakdown items above (ex ante and ex post wage rates, along with hours worked). Suppose, moreover, that we add in incentive issues, as well as public goods that even the libertarian agrees require tax funding. Then the utilitarian approach is to a degree specified, at least within the contours of a simplified model, although it requires other inputs, such as concerning labor supply elasticity and the slope of declining marginal utility. It’s not clear (at least to me) how this might be made equally to hold for the other approaches.

NYU Tax Policy Colloquium - paper by Marc Fleurbaey on optimal tax theory

Yesterday at the colloquium, Marc Fleurbaey presented his recent JEL paper, Optimal Income Taxation Theory and Principles of Fairness (co-authored by Francois Maniquet). The paper is more mathematical and abstract than our usual fare at the colloquium, but it aims to illuminate an aspect of the philosophical debate around tax policy that is certainly of interest.

A central premise is that optimal tax theory (OTT), as founded by Mirrlees' famous 1970s work, has made major strides in deploying social welfare functions (SWFs) to support conclusions about, not just optimal, but second-best tax systems. An example is the long-standard recommendation that the tax system use demogrants at the bottom with relatively flat rates, possibly even declining (in theory to zero) at the very top. Diamond and Saez have recently expanded the Overton window by arguing that OTT might instead support a tax rate as high as 70 percent top. A key move that they make in this regard is to argue that the marginal utility of a dollar for the very richest people should be valued at (effectively even if not quite literally) zero - whether as their own presumed subjective measure, or as a social assignment of value in the SWF. With a welfarist SWF,  only people's welfare counts to the bottom line evaluation of a set of outcomes, and it can only count positively, but differential weighting of people's utilities is permissible unless one adopts a utilitarian approach, which requires valuing everyone's utility equally.

The paper notes that utilitarianism (and other welfarism) have not won universal, unchallenged acclaim. Hence, if one considers the exercise of using SWFs intellectually (or otherwise) valuable, one should be in favor of modifying them, so that they can accommodate alternative viewpoints, such as those which value "fairness" defined in one way or another. The idea is that, say, libertarianism or resource egalitarianism (or systems resembling / parallel to them) ought to be expressible in SWF terms, permitting one as well to be, say, partly one or another or both.

The concept of "money market utility," dating back (at least) to a 1974 paper by Paul Samuelson, plays an important role in the analysis, but explaining all that here would be rather complex and take a long post of probably less than general interest. The basic idea behind money-market utility is to surmount interpersonal utility comparison problems by employing complete specifications of people's preferences, stated in dollar terms, in comparing states of affairs. E.g., rather than asking how my utility differs in inferior State 1 as compared to superior State B, we ask how many dollars I would have to be given, in State 1 as compared to State 2, in order to deem them equivalent. The concept's usefulness is undermined by problems such as preference knowledge and preference revelation. But it may help if one considers its existence in principle (assuming that people have consistent and well-ordered preferences) to be important.

But the following two quick points may help to show very generally what the paper has in mind:

1) Technically speaking, most efforts to incorporate non-utilitarian (albeit generally not non-welfarist) values into the SWF have involved differential weighting of people's utilities - for example, to give priority to the wellbeing of the worst-off, at the extreme through the quasi-Rawlsian maximin, in which the welfare of the worst-off individual completely outweighs that of everyone else. Under the maximin, reducing everyone else's welfare by 20 trillion utiles (granting for argument's sake the existence of such a thing) in order to raise that of the (still) worst-off individual by one utile would be scored as a social welfare gain. This might support the tax policy conclusion that everyone above the worst-off individual should be taxed at the revenue-maximizing rate, with the proceeds being used to raise the bottom as much as possible. But the paper argues that differential weighting of utilities generally doesn't get one to the right place, so far as the various fairness theories it explores are concerned. Instead, one has to go the individual inputs (people's utility as determined for purposes of the SWF) and modify them as needed.

2) To illustrate that point, consider what I just called the quasi-Rawlsian maximin. I called it quasi-Rawlsian because, as many have noted, it's not really what Rawls supports even though he advocated absolute priority for the relevant concerns of the worst-off individual. The difference arose in his not being a welfarist. E.g., he spoke of primary goods rather than welfare generally. Suppose, therefore, that one modified the SWF so that the relevant "arguments" (i.e., people's utilities) were based on a Rawlsian primary goods conception, rather than on the notion of utility. Or to put the same point differently, suppose that one defined "utility" for purposes of the Rawlsian SWF in terms of primary goods - on the view that it is simply a marker for what the social welfare evaluator cares about, rather than purporting to represent an objective fact about people's welfare. I suspect that the SWF one thus computed still wouldn't be precisely what Rawls, or various of his followers, might say they want to do, but it would certainly come closer to systematizing, OTT-style, the normative concerns that motivate them.

I'll have a follow-up post to this in which I discuss a road not followed in our discussion yesterday, so that it doesn't go to waste (as it may, I hope, be interesting & useful). It involves a little illustration I prepared, but then elected not to use in the discussion as it proved not to be sufficiently germane, that sketches out how some different philosophical positions discussed in the paper (utilitarianism, resource egalitarianism, an approach taken by John Roemer, and libertarianism) might apply to a particular stylized fact pattern.