Wednesday, August 27, 2008

Tax Foundation is NOT a Credible Source

As I explained here, it is a right-wing think tank, part of the billionaire-funded Republican propaganda machine. Here's another post describing some of the misinformation they produce.

Greg Mankiw today links to a brief from the Tax Foundation which tries to claim that tax cuts recover up to 40% of their costs through so-called dynamic effects. Although this is still a long, long way from McCain's deluded claim that tax cuts pay for themselves, it is a much bigger effect than Bush's own Treasury Department estimated (less than 10%). This is an effort to argue that McCain's tax proposal wouldn't create deficits quite as massive as the Tax Policy Center's analysis shows.

Where does the TF figure come from? From yet-to-be-published research, which conveniently makes it impossible to fully critique it. It says it is based on a "large sample of tax returns filed by the same taxpayers between 1998 and 2005 and "examined the change in taxpayers' taxable income as reported on their tax forms and the change in their tax rates, after controlling for a variety of non-tax factors." Immediately, one can see the econometric issues, starting with what is the right lag structure and what is the right set of covariates? The whole analysis sounds pretty fishy. If the data is as juicy as it sounds, why didn't the authors just estimate the direct behavioral effects of the 2001 tax changes, treating them as a natural experiment? The fact that they did not makes me suspect that they massaged the analysis to get the results they wanted. Without seeing the paper, I can only speculate.

My big question is, where on Earth did they get such great data? I have never heard of a publicly available panel data on tax returns, although admittedly, this is not my research area. Will this data be released to other researchers so that they can try to replicate the analysis and examine how fragile the results are? (This has become the research standard for major journals.) I am going to hazard a guess that the researchers will say that the data is confidential, so that we'll just have to trust them. But the Tax Foundation doesn't deserve that kind of trust from anyone, Greg Mankiw included.

UPDATE: (I wrote this in comments on Economist's View):
I want to point out that one of the author's took strong issue with the post in comments on our blog and offered to send me the paper (which I hope to get soon.)

So I may revise my opinion of the paper and even retract my criticisms after further reflection!

But I stand by the fundamental point that outfits like the Tax Foundation don't get the benefit of the doubt when publicizing results based on unpublished research.


Anonymous said...

You obviously didn't read the report because it didn't claim that the tax cuts recovered up to 40% of their costs.

The report, written by a former Treasury official, said that between 25-40 percent of one component of the tax cuts (reducing the top two rates) paid for itself.

There are many other parts of the Bush tax cuts, most of which had little or no feedback effects. The report makes this clear right here.

"The central component of the 2001 and 2003 tax relief was lower individual income tax rates. For example, the 15 percent rate was lowered to 10 percent for low-income taxpayers and the top tax rate faced by high-income taxpayers was lowered from 39.6 percent to 35 percent. The 10-percent bracket accounted for nearly one quarter of the individual income tax relief. This tax change, combined with the reduction in the marriage penalty and the expansion of the child tax credit, provided a short-term economic stimulus by letting taxpayers keep more of their income. Nevertheless, even though these provisions increased taxpayers' after-tax incomes, they did little to improve economic incentives. Indeed, to the extent these tax provisions were financed with additional government borrowing, they may well have detracted from economic growth in the longer term by adding to the deficit and increasing long-term interest rates."

This blog is what is not a reliable source!

Don Pedro said...

Dear Anonymous,

I've gone back and forth on whether or not to allow anonymous comments on the blog. On the one hand, I like to encourage wide discussion. On the other hand, I feel like an idiot writing "Dear Anonymous."

The Tax Foundation document is not a "report." It is a brief based on forthcoming research that is not public.

Yes, the paper is about reductions in tax rates, not about overall generic tax reductions. I didn't explain this in my post because I wasn't trying to summarize everything in the paper.

From a policy perspective, the question is not "what were the effects of the overall 2001 and 2003 tax cuts" but rather "what would be the effects of specific tax cuts going forward?"

Anyway, my main objection is to the Tax Foundation putting out a brief when the paper on which it is based is not yet available, not even in a working paper version (at least not that shows up on Google Scholar.)

When the paper is published, I'll give it a close read, and revise my evaluation accordingly.

Anonymous said...

But that Treasury figure study you cite (10%) is out of ALL THE BUSH TAX CUTS!

The Tax Foundation report/study/whatever that quotes the 40 percent figure for just the cutting of the top two rates is therefore not comparable to the 10 percent figure.

The two are estimating different things (different denominator).

Therefore, your statement of "much bigger effect" is incorrect.

Don Pedro said...

Dear Anonymous,
Please no shouting on this blog.

Your post gets to the essence of my problem with what little I know of the paper. How did they isolate the effects of the rate changes from all the other tax changes? This could only be done with some strong modeling assumptions, and I can't tell what those assumptions are (implicit or explicit) from the brief.

Anonymous said...

I am a bit hesitant to post to this blog as I have found that it is rare for there to be a genuine discussion. More often than not, it is just shots going back and forth.

Also, I find the headline questioning the integrity of the Tax Foundation to be offensive, highly unprofessional, and disinterested in any type of honest discourse. Clearly, my piece touched a nerve.

In any case, I'll try to respond to some of the points raised in the initial posting on my Tax Foundation Fiscal Fact.

First, I'll just mention that I have gotten critical emails from both people on the left and on the right. Those on the left take issue with the notion that the revenue offset related to changes in the top two rates can be as much as 40 percent. Those on the right take issue with it only being 40 percent; that is, tax cuts don't pay for themselves. Given that I am getting criticism from both sides, maybe I have this close to right.

Also, I should just point out that I have never claimed that tax cuts pay for themselves. Just the opposite. My standard line has been that, as a general issue, I don't think tax cuts pay for themselves, but they can provide important economic benefits that, in part, may lead to an increase in the size of the tax base and offset part of their static revenue effect.

As for the blog posting. Well, the paper is available. It was presented at the Spring Symposium of the National Tax Association held here in Washington, DC. The final version for publication in the National Tax Journal was submitted last month. I'd be happy to send a copy to anyone who would like one. Just let me know.

Second, the Fiscal Fact (and the paper) states that the reduction in the top two individual tax rates likely resulted in an expansion in the size of the tax base to such an extent that about 25 to 40 percent of the static revenue cost was offset. So, our simulation only pertains to the top two tax rates.

How does this compare to what Treasury actually assumes? In a 2005 American Economic Proceedings paper I (and Warren Hrung, another former Treasury economist) explain some of the thinking that underlies the Treasury estimates of tax rate changes. In that paper we indicate that Treasury's revenue offset is about 25 percent.

So, how does one get to a 40 percent revenue offset? Well, the 40 percent revenue offset is estimated by simply applying the taxable income elasticity for taxapayers' in the top two tax rates. By construction, this only considers changes in the individual
income tax base. As the Fiscal Fact and the underlying paper indicate (as well as the earlier 2005 paper I mentioned above) to estimate the revenue offset, you also need to account for changes in the corporate tax base. This works in the opposite direction and might take the form of firms or new investment occurring in the noncorporate form rather than the corporate form in response to the change in the relative taxation between the corporate and noncorporate sectors. This shift can be thought of as self-help integration.

If you use the experience of changes in organizational form from around the 1986 Act (based on one paper by Austan Goolsbee and another paper coauthored by myself and a Treasury economist), you might expect the revenue offset to be closer to 25 percent. But, I'd point out that it is a bit unclear to me that you would expect as much disincorporation as we saw after the 1986 Act.

Another point to make is that the official Treasury and Joint Committee on Taxation estimates hold GDP fixed. That is, although they are micro-dynamic, they are not macro-dynamic. If the reduction in the top two tax rates caused GDP to increase at all, the revenue offset would likely rise above the Treasury 25 percent. In 2006, Treasury released a detailed analysis of the dynamic effects of the 2001 and 2003 tax relief. The estimates were derived from a standard dynamic model used also by the Congressional Budget Office, the Joint Committee, and a number of academics. This model estimated that the reduction in the top four rates would increase the level of GDP in the long-run by about 0.7 percent. I'll let others calculate the feedback effect. As an aside, the 2006 report also indicated that the lower tax rates on dividends and capital gains increased the level of GDP in the long-run by 0.4 percent and the remaining tax cuts, the child tax credit, the marriage penalty relief, and the new 10 percent tax rate, actually lowered the level of GDP in the long-run by 0.4 percent. On net, the total package increased GDP by about 0.7 percent.

Another point on the estimates is that we used the taxable income elasticity for the overall taxpaying population when simulating the effect of lowering the top two tax rates. One should expect, however, that the response for those in the top two tax rate brackets would be larger than for the general population. I thought this was a reasonable approach, although not quite right, because it allowed us to be conservative.

Finally, the taxable income elasticity that we estimate, 0.4, is pretty much the central tendency estimate from this literature. It is below the 0.6 elasticity that a coauthor and myself estimated when looking at the 1986 Act. It is below the elasticity I estimated when looking at the 1993 Act (a Treasury Office of Tax Analysis Working Paper). It is pretty much identical to the elasticity obtained by Jonathan Gruber and Emmanuel Saez in their 2002 paper examining the tax changes throughout the 1980s. Also, the estimation procedure is generally similar to this other research as well. So, frankly, I am a bit puzzled about the strident concerns on the size of our estimated response or the methodology.

There is a lot more that I could say about the posting, but my response is already quite long.

Don Pedro said...

Thanks very much for visiting. Can you please send a copy of the paper to I looked for the working paper version and I couldn't find it. I'll be happy to post again after seeing the full paper.
Don Pedro

Don Pedro said...

One more thought: I entirely stand by my indictment of the Tax Foundation. Both the past brief I looked at on the past experience with the past windfall oil profits tax, and the organization's materials which argue that corporate tax rates are high in international terms are essentially campaigns of misinformation, designed to confuse the public policy arena rather than to help policymakers make informed decisions.

I resented reading that material because I was trying to form my own position on these issues, and it took me some time to look at them closely enough to realize that the authors were not trying to help me understand the issues, but were twisting the facts to sell me on their view.

You can get your panties in your wad if you like, but asserting that the Tax Foundation is not credible was exactly my point.

Researchers at the Tax Foundation, AEI, and all the machinery of right wing think tanks present themselves as disinterested analysts, but often they are pushing a point of view which is bought and paid for by their funders.

So, as I said, they don't deserve the benefit of the doubt when publicizing unpublished analysis of crucial public policy issues.

That said, it's possible that this particular work is an exception. I'll be able to form a view on this after seeing the paper.

Jonah B. Gelbach said...

In the interests of discussion, rather than "shots going back and forth", let me add one thing concerning Robert Carroll's use of an elasticity of 0.4. The most convincing paper I've read on this topic, in the sense that it's relatively unplagued by data or endogeneity problems (due to the fact that the MTR is a function of taxable income, which is the LHS variable) is Austan Goolsbee's JPE paper on the 1993 tax increase and highly paid executives. Here's the money quote from the abstract:

The short-run elasticity of taxable income with respect to the net-of-tax share exceeds one in this sample, but the elasticity after one year is at most 0.4 and probably closer to zero.

I'm also familiar with Carroll's own ReStat paper, coauthored with Auten (I believe this is the paper to which Carroll refers in his comment). Like much of the rest of the literature on this issue, I think that paper suffers from a mis-specification problem due to the use of an instrumental variable that includes actual tax rates in the pre-reform period. That is, the method of instrumenting the "tax price" in Auten & Carroll doesn't solve the endogeneity problem it's meant to. This problem is not specific to Auten and Carroll's paper -- almost every paper on this topic (though arguably not Goolsbee's, since his sample is almost all people in the top bracket) has this problem. It's not a moral failing on Auten and Carroll's part -- it's a difficult econometric problem to solve, plain and simple. But I don't put a lot of weight on much of this literature, which I used to teach down to the details in PhD public econ courses, because I think the econometric implementation is basically indefensible.

So if it were up to me, I would probably rely almost exclusively on Goolsbee's paper, at least for folks at the top of the pre-tax income distribution, and so I would use an elasticity of less than 0.4.

Anonymous said...


Thanks for your thoughtful comment. Maybe there is some hope that at some will actually read a paper and reflect upon it before commenting.

Just a quick comment on the Goolsbee JPE paper. One problem with just examining taxpayers in the top tax bracket or two is that the identification of the tax response is just within group, and this is a particularly week source of identification.

Also, Goolsbee just looks at company executives of publicly traded companies and just looks at their compensation, not their taxable income. Indeed, this is one of the novelties of his paper -- he uses Compustat's ExecComp data that captures the compensation of the top five executives of public traded companies. So, I was never really surprised with the paper's small estimated response.

In one version of the taxable income paper I wrote on the 1993 Act (presented at the NTA annual meeting with Austan Goolsbee as a copanelist) I ran a specification that just looked at company executives and used the change in wages as the dependent variable. The tax data I was using included an occupation code that allowed me to identify company executives. I thought this specification was pretty close to the one used by Goolsbee -- just executives and just compensation. This specification resulted in an elasticity that was close to zero (elas.=0.08, s.e.=0.39), similar to what Goolsbee reported.

But, because I had the tax data, I could, of course, run the same specication for a taxable income equation and include non-executives.

Indeed, when the same model was run for the entire taxpaying population (roughly speaking) and I used taxable income rather than wages as the dependent variable, I got a taxable income elasticity of 0.5. When I ran the full model, but used wages as the dependent variable, the estimated elasticity was 0.6.

These results tell us a few things. First, it is very likely that a lot of the action was coming from non-wage income. Second, and perhaps more important, the fact that I got an elasticity of 0.6 for the general taxpaying population for the wage equation, but ostensibly a zero elasitcity for executives, told me that the Goolsbee results are probably only useful when thinking about how tax rate changes affect the wages of executives, not the size of the tax base (i.e., taxable income) nor the broader taxpaying population.

Also, one should note that another plausible explanation is that there simply may not be sufficient exogenous variation in the tax rates when running the model on just executives as the vast majority of executives tend to be in the upper most brackets; that is, the elastcity may not be zero at all, its just that we cannot estimate it from these data.

Bob Carroll

Jonah B. Gelbach said...


Thanks for your thoughtful reply. I didn't know about your other paper (which I think you just emailed me, and which I look forward to reading!).

You make an interesting point about Goolsbee's paper vis-a-vis the compensation/taxable income distinction. If I understand you correctly, you're suggesting that compensation responds less elastically to the tax price than does TI, since by definition TI is what is taxed, whereas some forms of compensation aren't taxed.

Fair enough as a general point. But to be convinced that Goolsbee's estimates miss something important for the purposes of measuring TI, I'd need to be convinced that there are important components of the compensation measure he uses that are not taxable. You no doubt know the guts of the tax code better than I, so all I can say is that in the long run (which is what Goolsbee's focusing on), I'm not sure which of the components of his compensation measure would fit this argument. I'd be glad to learn more about this issue.

I'd also say that I disagree with this argument:

another plausible explanation is that there simply may not be sufficient exogenous variation in the tax rates when running the model on just executives as the vast majority of executives tend to be in the upper most brackets; that is, the elastcity may not be zero at all, its just that we cannot estimate it from these data.

Lack of variation would imply imprecise estimates, not small but relatively precise ones, which I believe is the case in Goolsbee's paper. Also, the identifying variation in his paper strikes me as much more plausibly exogenous than that used in most of the literature, most notably those papers using data from TRA86 (for ex your ReStat paper, which again I have nothing against in particular -- I think and have always thought that in the context of that literature, it's very well done).

In any case, I look forward to reading the papers you emailed me, and I do hope you'll feel welcome to offer your thoughts here at E4O.



Richard H. Serlin said...

So the the right wing propaganda tank Tax Foundation claims that tax cuts recover up to 40% of their costs through so-called dynamic effects, while Bush's own Treasury Department estimated less than 10%.

Even if it actually were 40%, are tax cuts a good idea, especially tax cuts going predominantly to the rich and extremely rich? They're still costing the government 60% that can't go to many extremely high social return projects that the free market won't undertake due to market imperfections that are well established and proven in economics (real scientific, academic economics, not screaming talk show host, propaganda tank economics), like externalities, asymmetric information, impracticalities of patenting, large economies of scale and monopoly issues, the zero marginal cost of information and ideas, the inability to price discriminate well, and many more available in any university introductory and intermediate economics texts.

Suppose we consider continuing Republican policies and spending another 1 trillion on tax cuts for the rich. Even if 40% were recovered (and in the long run, as opposed to just looking at short run effects, the dynamic effects go in the opposite direction -- a dollar in tax cuts ends up costing a lot more than a dollar in government revenue if that means a dollar, or even 60 cents, less in investment in high return government projects.).

The vast majority of the tax cuts, it has been shown, will eventually be spent on consumption items of little long run investment value -- leaving little to show or to grow. If instead, even just 60% of that 1 trillion were spent by the government on high social return investments like infrastructure, education, basic scientific and medical research, alternative energy, etc., then 10 or 20 years from now that 600 billion could result in many trillions, or even tens of trillions more in national wealth, as opposed to having the whole 1 trillion spent on rapidly depreciation Ferraris and yachts and ultra luxury vacations, and other things for the rich that have little or no productive value.

In the long run, a dollar spent on tax cuts for the rich, instead of badly needed social investment puts us one more step closer to losing our status as the most wealthy and modern nation, and over the long run, like any other decision to increase frivolous consumption at the expense of high return investment, it costs us a lot more than a dollar, not less.

Richard H. Serlin said...

Please also see my post, "More on tax cuts (pun intended)" at: