To me, it seems like economists love to assume that patents are good for innovation, because they can measure patents and thus produce papers. Based on my (software) industry experience, I think it would be more accurate to treat patents as a negative factor that slows down innovation.
The vast majority of patents occur when a company has *not* innovated anything, and is copying exactly the obvious things that other companies are doing. Technically this is against the rules of patents, but the patent reviewers are widely understood to be completely incompetent, and this is easily evaded by inserting meaningless jargon. The patent is like a land mine, interfering with other companies that are treading the same ground that has been covered dozens of times before.
Trying to boost innovation by encouraging more patents is like trying to boost housing construction by encouraging more CEQA lawsuits.
I will somewhat defend patents as a (weak) measure of innovation though, especially in this paper specifically.
The association of patents with innovation is more believable here because the intervention has nothing to do with patents ex ante. Like, when people find that longer patent protection or development grants leads to more patents and call that an innovation effect it's a bit rich. But here there's no direct change to cost/benefit of patenting. You're just giving people more money and they choose to invest in more R&D, producing patents as a correlated result.
I don't think that argument is even saying patents are a good *measure* of innovation, just that they *tend to correlate* with innovation.
As an analogy, consider air resistance. It's proportional to velocity (for laminar flow). For a fixed object, it's just a law of physics. The faster you go, the more air resistance you have. But it would be completely backwards to use air resistance as a measure of velocity. Imagine trying to maximize air resistance when trying to make an airplane go as fast as possible!
I contend that patents are like air resistance. Unfortunately, they may correlate with innovation. But you're better off trying to *reduce* patents, and when you find a particular intervention that increases patents it is likely to be a bad thing.
IBM has many, many patents in recent decades. SpaceX has very few patents. Do we need more IBMs and fewer SpaceXs?
I agree that patents are a poor goal to optimize for. And in many cases they are a misleading measure (many interventions like extending patent terms add air brakes to the plane and celebrate the increased air resistance).
But in this case all we're doing is adding fuel or cutting weight or some such (reducing incomes taxes), and the air resistance goes up. In a case like that, air resistance is a serviceable measure of velocity when actual velocity is very hard to measure.
> Federal income taxes paid by the top 10% of earners is 1.6 trillion. So decreasing the marginal tax rate for those earners by 1% costs 16 billion.
I wasn't able to decipher the paper well enough - does reducing the marginal tax rate from 30% to 29% count as a 1% or 3.3% decrease? If the former, the cost would be 53B, not 16B.
This is a good question and I definitely agree that it's not very clear in the paper. In section 4.1.2 they say "A one percent decrease in MTR90 (equivalently, a one percent increase in the net-of-tax retention rate at the 90th percentile)"
I think this means they are looking at when the amount of tax dollars collected goes down by one percent, or the amount of money kept by people increases by one percent. If they were talking about changing the nominal rate by one point I think they would say "A one percentage point decrease."
Another clue is that they are running every specification in logs which is looking at percentage changes. The log just treats the tax rate as an integer so 29 ==> 30 is a 3.5% increase, so a 1% increase would be 29 ==> 29.29.
Thanks for clarifying! Surprising results, I didn't expect people to respond so strongly to tax rates. If true definitely a very strong argument (among others) for lower taxes.
Indeed. Wonderful piece here Max. My position is, and remains, that the cornerstone of modern society should be Land Value Tax. To the maximum extent possible, we should eliminate taxes on production and earnings, LVT fits the bill: https://www.lianeon.org/p/just-tax-the-land
This research just illustrates how broad and deep the effects of income taxation actually run….I am wondering what the effects of a DBCFT, as a replacement to the corporate income tax for example, might be on research output. Did you have any thoughts?
BTW, you left out of the objective of taxation transferring consumption from people who on the margin value it less to those who value it more. This is the raison d’état of pensions, health insurance, unemployment (and a child allowance if we had one)
On the Adam Smith criteria:
• Canon of Equity: Taxes should be progressive (proportional to income) [No. “Progressive” means that marginal rates rise with the amount of the base, not remain constant.]
• Canon of Certainty: Taxes should be clear, not arbitrary or hidden
• Canon of Convenience: They should be easy to pay (i.e. no filing fees)
• Canon of Economy: They should be cheap to administer and collect
• [I would add, have low deadweight losses. (Taxing negative externalities is creating negative deadweight losses)]
A combination of VAT, tax on net emissions of CO2 and methane, and progressive consumption tax meet these criteria. I think the difficulty of valuing “land” in order to tax its rental income makes the LVT (as I understand it) fail on the “certainty” and “economy” criteria.
My thought on the DBCFT is that we should not "replace" the corporate income tax but eliminate it. As a way of raising revenue it is just a VAT with arbitrary coverage. Let's go for the real thing.
Great read. I'm really surprised by the magnitude of the effect at the individual level. I have some reservations (like, the three-year lag seems somewhat arbitrary and maybe an opportunity for p-hacking?), but really it's just difficult for me to imagine individuals changing their patent-output based on small adjustments to state taxes. I'm curious whether more crucial innovation is less likely to be effected. Also, don't we currently incorporate tax-reduction strategies for boosting innovation via R&D tax credits?
I'm a big fan of tax cuts. But only if they are paired with spending cuts. We are wasting about a trillion dollars a year now (and rising) for interest on the debt. Now maybe if that debt was all being used to make positive NPV investments in the future it would be ok. But the majority of that is being used to fund present consumption.
That's just not sustainable long term. Moreover, 80%+ of the budget goes to entitlements, defense, and interest on that debt. So any material cuts are going to be pretty hard to make.
Sadly what that means is the most likely fiscally sane solution is going to be more taxes combined with entitlement cuts.
Spending cuts are good. I can think of several expenditures with NPV << 0 Ethanol subsides , agricultural price supports, enforcement of sundry regulatory errors. Unfortunately these will not reduce the deficit very much
Oh please. First, as noted above, patents are not the same as innovation. Second, what’s the theory of patent innovation, and why would state taxes be a significant factor in that mechanism? If your regression analysis leaves out key explanatory variables, you’ll get poor results. Third, does the location of a patent cause incomes to rise in that location and that location only? If not, who cares which states produce more patents?
It pops out immediately that it's an unreasonable assumption to make that the "fixed effects" discussed are not in any way related to tax rates/spending.
Without any analysis of where the "fixed effects" might come from, the conclusions of this research have to be disregarded as they're based in unjustified assumptions.
It may well be that 30% of states with high taxes spend tax money in ways that foment innovation, while 70% of states don't. Surface level analysis would then tell you that there's a negative correlation between high taxes and a high "fixed effect", which might lead you to unjustifiably conclude that there is no relationship.
But the reality could very well be that there are very effective programs, education systems, and/or incentives in some states, which are funded by taxes, which encourage innovation.
That might NOT be the case, but without any actual analysis on this issue, it is terrible scientific practice to simply assume that it's not, and that the "fixed effects" can simplistically be assumed to have nothing to do with govt. programs funded by taxation.
Anyone capable of reading this research without noticing this glaring problem with their analysis is likely either scientifically illiterate or under the influence of confirmation bias.
This is a valid insight as it applies to what kind of taxes make the most sense, clearly not taxes on business income. I do hope that it is not misused to justify setting revenues at less than the amount of NPV > 0 expenditures. My guess is that it not of much relevance in deciding what amounts of expenditures with NPV < 0 cum with what rates of taxation of consumption would be desirable.
I don't understand why blanket tax cuts on the rich, most of whom do not spend their money funding R&D, is the way you would want to encourage R&D. Couldn't you make R&D expenses tax deductible, or some other more refined policy measure, to encourage innovation, without exacerbating income inequality?
Yes, from my understanding it seems like this would probably be especially helpful to start-ups, but even tax credits for R&D or subsidization would seem to be more effective at incentivizing R&D specifically than simply cutting taxes.
I find it useful to stress test findings like these. A 1% increase in income tax rate decreases innovation by 2%. I interpret this to mean:
innovation at tax rate 2 = innovation at tax rate 1 x (tax rate 1/tax rate 2)^2, i.e.
innovation with a 1% tax increase = 1.00 * (1/1.01)^2 = 0.98 (2% decrease)
From the business cycle peak in 1937 to the 1981 corporate tax rate averaged 45%.
For the peak in 1989 to 2023, corporate rate averaged 32%. The expectation is that corporate innovation rates proceeded at about half the rate in the earlier period (32/45)^2 = 0.51.
We should expect such a dramatic change in innovation rate to show up in economic growth rates. So I compared real per capita GDP growth rates: In the earlier, higher tax period, real per capita GDP grew at 2.8% while during the later lower tax period it grew at 1.5%.
This is the opposite of what we would expect. The reason, of course is not that the effect seen in the paper didn't happen, but rather its effect it was overwhelmed by other factors that are more important. One obvious factor is WW II. If we select the 1948 to 1981 period to exclude the war, the growth rate falls to 2.25%, which is still higher than the recent low tax period. Average corporate tax rate during this period was a bit higher at 49%. So even after taking out the war, the effect is still opposite of what is expected.
Studies with things like cigarette taxes show that if you tax something you get less of it, just as economic theory would predict. And this study shows a similar effect of small changes in tax rates leading to small changes in innovation rate. But when we look at large changes made in the past the effect was not apparent. Tax cuts have other effects, monetary, financial stability, executive behavior, etc. that may be larger than the effect measured in the paper.
For example, corporate rates were cut by 40% in 2017, which implies companies would enjoy greater after tax earnings, which the paper suggests would translate into more investment in innovation. So what did they do with these earnings? Over the 2018-23 period 99.2% of the earnings of the S&P500 companies were paid out as stock buybacks (59.6) and dividends (39.6), leaving only 0.8% for investment. It does not look like that rather large decrease in corporate taxes led to more investment or innovation. Again, the tax effect measured by the paper was overwhelmed by other things.
To me, it seems like economists love to assume that patents are good for innovation, because they can measure patents and thus produce papers. Based on my (software) industry experience, I think it would be more accurate to treat patents as a negative factor that slows down innovation.
The vast majority of patents occur when a company has *not* innovated anything, and is copying exactly the obvious things that other companies are doing. Technically this is against the rules of patents, but the patent reviewers are widely understood to be completely incompetent, and this is easily evaded by inserting meaningless jargon. The patent is like a land mine, interfering with other companies that are treading the same ground that has been covered dozens of times before.
Trying to boost innovation by encouraging more patents is like trying to boost housing construction by encouraging more CEQA lawsuits.
I agree. I've written against IP laws before for this reason e.g https://www.maximum-progress.com/p/patents-vs-prizes or https://www.maximum-progress.com/p/the-printing-press-nfts-and-the-end-of-copyright
I will somewhat defend patents as a (weak) measure of innovation though, especially in this paper specifically.
The association of patents with innovation is more believable here because the intervention has nothing to do with patents ex ante. Like, when people find that longer patent protection or development grants leads to more patents and call that an innovation effect it's a bit rich. But here there's no direct change to cost/benefit of patenting. You're just giving people more money and they choose to invest in more R&D, producing patents as a correlated result.
For a case for using patents in research more generally see Matt Clancy https://mattsclancy.substack.com/p/patents-weakly-predict-innovation
I don't think that argument is even saying patents are a good *measure* of innovation, just that they *tend to correlate* with innovation.
As an analogy, consider air resistance. It's proportional to velocity (for laminar flow). For a fixed object, it's just a law of physics. The faster you go, the more air resistance you have. But it would be completely backwards to use air resistance as a measure of velocity. Imagine trying to maximize air resistance when trying to make an airplane go as fast as possible!
I contend that patents are like air resistance. Unfortunately, they may correlate with innovation. But you're better off trying to *reduce* patents, and when you find a particular intervention that increases patents it is likely to be a bad thing.
IBM has many, many patents in recent decades. SpaceX has very few patents. Do we need more IBMs and fewer SpaceXs?
I agree that patents are a poor goal to optimize for. And in many cases they are a misleading measure (many interventions like extending patent terms add air brakes to the plane and celebrate the increased air resistance).
But in this case all we're doing is adding fuel or cutting weight or some such (reducing incomes taxes), and the air resistance goes up. In a case like that, air resistance is a serviceable measure of velocity when actual velocity is very hard to measure.
> Federal income taxes paid by the top 10% of earners is 1.6 trillion. So decreasing the marginal tax rate for those earners by 1% costs 16 billion.
I wasn't able to decipher the paper well enough - does reducing the marginal tax rate from 30% to 29% count as a 1% or 3.3% decrease? If the former, the cost would be 53B, not 16B.
This is a good question and I definitely agree that it's not very clear in the paper. In section 4.1.2 they say "A one percent decrease in MTR90 (equivalently, a one percent increase in the net-of-tax retention rate at the 90th percentile)"
I think this means they are looking at when the amount of tax dollars collected goes down by one percent, or the amount of money kept by people increases by one percent. If they were talking about changing the nominal rate by one point I think they would say "A one percentage point decrease."
Another clue is that they are running every specification in logs which is looking at percentage changes. The log just treats the tax rate as an integer so 29 ==> 30 is a 3.5% increase, so a 1% increase would be 29 ==> 29.29.
Thanks for clarifying! Surprising results, I didn't expect people to respond so strongly to tax rates. If true definitely a very strong argument (among others) for lower taxes.
Indeed. Wonderful piece here Max. My position is, and remains, that the cornerstone of modern society should be Land Value Tax. To the maximum extent possible, we should eliminate taxes on production and earnings, LVT fits the bill: https://www.lianeon.org/p/just-tax-the-land
This research just illustrates how broad and deep the effects of income taxation actually run….I am wondering what the effects of a DBCFT, as a replacement to the corporate income tax for example, might be on research output. Did you have any thoughts?
BTW, you left out of the objective of taxation transferring consumption from people who on the margin value it less to those who value it more. This is the raison d’état of pensions, health insurance, unemployment (and a child allowance if we had one)
On the Adam Smith criteria:
• Canon of Equity: Taxes should be progressive (proportional to income) [No. “Progressive” means that marginal rates rise with the amount of the base, not remain constant.]
• Canon of Certainty: Taxes should be clear, not arbitrary or hidden
• Canon of Convenience: They should be easy to pay (i.e. no filing fees)
• Canon of Economy: They should be cheap to administer and collect
• [I would add, have low deadweight losses. (Taxing negative externalities is creating negative deadweight losses)]
A combination of VAT, tax on net emissions of CO2 and methane, and progressive consumption tax meet these criteria. I think the difficulty of valuing “land” in order to tax its rental income makes the LVT (as I understand it) fail on the “certainty” and “economy” criteria.
My thought on the DBCFT is that we should not "replace" the corporate income tax but eliminate it. As a way of raising revenue it is just a VAT with arbitrary coverage. Let's go for the real thing.
Thanks for the correction. I edited that essay just now.
Given the mobility of IP and of IP developers, what I would like to see is some kind of control against gross wages trends in each area.
But I agree this looks like fairly good evidence that there is a casual interaction.
Great read. I'm really surprised by the magnitude of the effect at the individual level. I have some reservations (like, the three-year lag seems somewhat arbitrary and maybe an opportunity for p-hacking?), but really it's just difficult for me to imagine individuals changing their patent-output based on small adjustments to state taxes. I'm curious whether more crucial innovation is less likely to be effected. Also, don't we currently incorporate tax-reduction strategies for boosting innovation via R&D tax credits?
I'm a big fan of tax cuts. But only if they are paired with spending cuts. We are wasting about a trillion dollars a year now (and rising) for interest on the debt. Now maybe if that debt was all being used to make positive NPV investments in the future it would be ok. But the majority of that is being used to fund present consumption.
That's just not sustainable long term. Moreover, 80%+ of the budget goes to entitlements, defense, and interest on that debt. So any material cuts are going to be pretty hard to make.
Sadly what that means is the most likely fiscally sane solution is going to be more taxes combined with entitlement cuts.
Spending cuts are good. I can think of several expenditures with NPV << 0 Ethanol subsides , agricultural price supports, enforcement of sundry regulatory errors. Unfortunately these will not reduce the deficit very much
What are yours?
Oh please. First, as noted above, patents are not the same as innovation. Second, what’s the theory of patent innovation, and why would state taxes be a significant factor in that mechanism? If your regression analysis leaves out key explanatory variables, you’ll get poor results. Third, does the location of a patent cause incomes to rise in that location and that location only? If not, who cares which states produce more patents?
It pops out immediately that it's an unreasonable assumption to make that the "fixed effects" discussed are not in any way related to tax rates/spending.
Without any analysis of where the "fixed effects" might come from, the conclusions of this research have to be disregarded as they're based in unjustified assumptions.
It may well be that 30% of states with high taxes spend tax money in ways that foment innovation, while 70% of states don't. Surface level analysis would then tell you that there's a negative correlation between high taxes and a high "fixed effect", which might lead you to unjustifiably conclude that there is no relationship.
But the reality could very well be that there are very effective programs, education systems, and/or incentives in some states, which are funded by taxes, which encourage innovation.
That might NOT be the case, but without any actual analysis on this issue, it is terrible scientific practice to simply assume that it's not, and that the "fixed effects" can simplistically be assumed to have nothing to do with govt. programs funded by taxation.
Anyone capable of reading this research without noticing this glaring problem with their analysis is likely either scientifically illiterate or under the influence of confirmation bias.
This is a valid insight as it applies to what kind of taxes make the most sense, clearly not taxes on business income. I do hope that it is not misused to justify setting revenues at less than the amount of NPV > 0 expenditures. My guess is that it not of much relevance in deciding what amounts of expenditures with NPV < 0 cum with what rates of taxation of consumption would be desirable.
I don't understand why blanket tax cuts on the rich, most of whom do not spend their money funding R&D, is the way you would want to encourage R&D. Couldn't you make R&D expenses tax deductible, or some other more refined policy measure, to encourage innovation, without exacerbating income inequality?
R&D should be expensed for purpose of imputation of business income to personal income for taxation.
Yes, from my understanding it seems like this would probably be especially helpful to start-ups, but even tax credits for R&D or subsidization would seem to be more effective at incentivizing R&D specifically than simply cutting taxes.
I find it useful to stress test findings like these. A 1% increase in income tax rate decreases innovation by 2%. I interpret this to mean:
innovation at tax rate 2 = innovation at tax rate 1 x (tax rate 1/tax rate 2)^2, i.e.
innovation with a 1% tax increase = 1.00 * (1/1.01)^2 = 0.98 (2% decrease)
From the business cycle peak in 1937 to the 1981 corporate tax rate averaged 45%.
For the peak in 1989 to 2023, corporate rate averaged 32%. The expectation is that corporate innovation rates proceeded at about half the rate in the earlier period (32/45)^2 = 0.51.
We should expect such a dramatic change in innovation rate to show up in economic growth rates. So I compared real per capita GDP growth rates: In the earlier, higher tax period, real per capita GDP grew at 2.8% while during the later lower tax period it grew at 1.5%.
This is the opposite of what we would expect. The reason, of course is not that the effect seen in the paper didn't happen, but rather its effect it was overwhelmed by other factors that are more important. One obvious factor is WW II. If we select the 1948 to 1981 period to exclude the war, the growth rate falls to 2.25%, which is still higher than the recent low tax period. Average corporate tax rate during this period was a bit higher at 49%. So even after taking out the war, the effect is still opposite of what is expected.
Studies with things like cigarette taxes show that if you tax something you get less of it, just as economic theory would predict. And this study shows a similar effect of small changes in tax rates leading to small changes in innovation rate. But when we look at large changes made in the past the effect was not apparent. Tax cuts have other effects, monetary, financial stability, executive behavior, etc. that may be larger than the effect measured in the paper.
For example, corporate rates were cut by 40% in 2017, which implies companies would enjoy greater after tax earnings, which the paper suggests would translate into more investment in innovation. So what did they do with these earnings? Over the 2018-23 period 99.2% of the earnings of the S&P500 companies were paid out as stock buybacks (59.6) and dividends (39.6), leaving only 0.8% for investment. It does not look like that rather large decrease in corporate taxes led to more investment or innovation. Again, the tax effect measured by the paper was overwhelmed by other things.
So, I question how useful this finding is.