Most Dishonest Wall Street Journal Editorial Ever

Yes, it's the Wall Street Journal editorial page reporting more American Enterprise Institute-quality research from Kevin Hassett. This is the most mendacious ever.

Mark Thoma is on the case:

http://economistsview.typepad.com/economistsview/2007/07/yet-again-tax-c.htmlMark Thoma: Yet Again, Tax Cuts Do Not Pay for Themselves: The Wall Street Journal says Kevin Hassett has discovered the Laffer curve, but I think these data might say something else. Here's the picture from the editorial where they are making their usual plea for more tax cuts:

The blue line is supposed to be the Laffer curve, but this is far from compelling. Since it looks like all that's been done here is to draw a line through an outlier, Norway (an outlier that in other contexts we are told to ignore because it is an outlier, e.g. see below), and since this is clearly not the best fitting line to these data, here's another possibility:

...[R]evenues rise with tax rates, and the fit also looks better than in the first graph. Toss out Norway, and the fit looks even better (and to quote The Economist blog on this point, "Throwing out Norway...")

Comments

[Ponzi Q. Globalization]

Brad's line does fit better. But even better than the line would be a 30th order polynomial.

We shouldn't expect simple relationships like the one here between revenues and tax rates frozen in time to be too helpful when trying to understand complex reality.

Much of economics would benefit from this statement of the obvious.

[theo]

Not even the scientific journal in my field with the lowest standards and an unmeasurable impact factor would allow such a manifestly fraudulent fit to slip through the review process.

And this is in a paper with nearly a million readers.

Simply deluded.

I wonder what Edward Tufte would say. Or Darrell Huff.

Hey, that's a laugher of a curve.

Okay, sorry about that.

[Justin X]

Or, in other words, showing up the "con" in econometrics.

But then, what is the value of a cross-country relationship in proving or disproving a relation that is supposed to hold as a dynamic response function within country?

[Bernard Yomtov]

"Thoma's is even more egregious than the WSJ. His linear relationship implies that at a 800% tax rate the revenues would equal 100% of the GDP. This is utterly absurd."

No. It's your interpretation that's absurd.

Thoma's line describes the relationship between tax rates and revenues in the range of values observed in the industrialized world. It says nothing about 800% tax rates or other such idiocies.

[Richard Green]

The best fit is the one that minimizes the chosen loss function. And the loss for the Wall Street Journal is that any outcome that refutes the Laffer Curve is infinite.

More seriously, to response to fairhavenhorn, all we can infer from these data is the impact of corporate tax rates between zero and (it looks like) 35 percent--beyond 35 percent, there is the sort of identification problem that is the foundation of Manski's text: because there is no support of X beyond 35 percent, we cannot infer E(Y|x)beyond 35 percent.

For X < 35 percent, however, my strong suspicion is that the sum of squared residuals (and certainly the sum of absolute residuals) from Thoma is much smaller than from Hassett, and so based on my own loss function, Thoma wins.

[PaulC]

Isn't the Laffer curve supposed to show actual revenue rather than percent of GDP?

I thought the point was that excessive taxation would eventually reduce GDP (which is trivially true, just not very meaningful until you plug in actual numbers). Taking all taxes into account, you should collect a larger and larger percentage of GDP monotonically (if people cheat, that's a de facto lower tax rate). The problem is that GDP tends to 0 on the far side of the curve.

Or am I totally wrong about this?

All I get out of this data is that the data points mostly cluster around a bullseye target. There isn't much of an obvious trend.

[Knecht Ruprecht]

Mind you, the dishonesty begins long before the fitting of the curve. On the Y-axis you have "tax revenues [all taxes] as % of GDP, and on the X-axis you have the corporate tax rate. So a country like Sweden that has low corporate tax rates and high personal income taxes (arguably a good way to structure the tax system) shows up as a data point with a low X and a high Y. No wonder the authors didn't label Sweden on the graph.

[PaulC]

"I wonder what Edward Tufte would say. Or Darrell Huff."

...or the authority on all things supply-side, H. R. Pufnstuf.

[anne]

http://www.washingtonmonthly.com/archives/individual/2007_07/011682.php

"The Wall Street Journal editorial page is really getting desperate. Even for them. In an editorial today they present data on corporate tax rates around the world and, like those people who find an outline of the Virgin...."

Ah, I should have known, I should have seen, I must practice seeing and knowing, me, being a woman of belief.

[Bloix]

I suspect that Norway is an outlier due to its relatively large petroleum production sector. Norway imposes a 50% surtax on companies involved in oil and natural gas. As companies wishing to exploit Norway's North Sea oil reserves have no choice but to pay, the tax would not be expected to reduce economic activity significantly.

[Bloix]

I didn't put that very well. The point is that corporate tax may be a high percentage of total tax revenues in Norway because it has a corporate sector that does not exist in most countries - one that can be taxed at a very high rate without destroying its profitability, driving it out of business or leading it to relocate to a low-tax country.

[Collin]

Bliox, are you implying that there are no other countries with exploitable natural resources?

[anne]

Bloix is essentially correct, and Norway is using its oil and gas resources with remarkable forward looking skill, providing for ample saving against future resource limits.

[Charles]

Colin, c'mon.

The correct question to ask is "Are there no countries from which as large a part of the GDP as Norway's is based on exploitable resources?"

There aren't many. If my calculation is correct, Norway's oil revenues amount to well over 15% of its entire GDP.

[Mike Liveright]

I agree the WSJ plot is bad. What I would like to see is:

1) Exclude Norway as a special case.

2) A curve fit that use a quadratic curve,or similar "humped" curve, a(x) = b + c*x + d*x**2 so that we could see the how good a fit would occur if a function that permitted the possibility of a Laffer curve if it was supported by the data.

3) As was pointed out, the KEY question is what is the impact on growth of the economy, or perhaps government revenue? of the TOTAL corporate and individual capital taxes. The chart I would like to see, with appropriate curve fitting is perhaps this TOTAL tax rate vs. the growth 5 years later. -- Obviously Norway and similar outliers should be excluded.

---

Finally: My own personal proposal would be: to eliminate the Corporate taxes entirely, but then have the dividends and capital gains (inflation adjusted) treated as normal income, taxed at the standard rate, and require that all investments be re-valued at least every 5 years and taxed appropriately.

[Gerard MacDonell]

Values matter but are not the dominant consideration in every application.

The line drawn by the WSJ is an indication of stupidity or mendacity, not just conservatism. (I assert no relation between stupidity and conservativism either way.)

I recall Donald Trump once attacked Pat Buchan(n)an as a right-wing nutbar and then corrected himself, fearing that he had insulted right wingers. PB was just a nutbar.

This puts me in mind of that. It may well be that lower taxes are better,
either because redistribution is theft or because there are efficiency gains from low taxes that trump the distributional considerations. I side with the left on these issues, but humbly because I could be wrong.

But all that aside, this really does take the cake. The author of that WSJ piece is an idiot. Dow 36000 got me thinking and this proves it.

[Geoff]

Mike: Funny that you ask for that. In an exercise to avoid doing some real research, I tried to read the data points off the graph, plot them in Excel and fit a second order curve to them. The result is a parabola (duh) that peaks at about (28%, 3.3%) and then starts a downward trend.

It's obviously not a great fit, and one could question whether a second order polynomial fit makes sense. But had Hasset been more mathematically rigorous, he probably could have made a better case for himself.

[Kieran]

Obviously Norway and similar outliers should be excluded.

Obviously.

[Hackticus]

They must have used John Lott's regression package.

[Dave L]

Kieran - look, even if you DON'T exclude Norway, you don't get a fitted curve that looks anything like the WSJ monstrosity. They obviously just drew it free-hand - show me any kind of statistical regression that would reproduce it.

[Matthew Saroff]

This is why I'm not concerned about Murdock buying the WSJ. Whatever the value of the news coverage, it is outweighed by the editorial page.

When Murdock takes over, the news coverage will get worse, and hence, people will take the paper less seriously, including the editorial page.

[derek]

Geoff, I suspect a lot of people would be grateful if you posted the data points you got off the graph somewhere, to save us all the effort of doing the same thing. I thought about it, but quailed at the work.

[anne]

Mark Thoma and Brad DeLong of course have it right, as any reasoned look will show us, but there are those who see otherwise and I really do find the Virgin twinkling at me from the corner so who knows. Say what?

[Neil B.]

Sure, the curve fit is crap. However, note that the highest rates are about 36%, which may mean that nations realize it really isn't worthwhile to levy "high tax rates" (like 40% and up.)

[Neil B.]

Geoff: Of course you get a parabola at 28% with Norway there, the whole point is that Norway is an outlier. But even without that being the point, note that what really matters is how many points lie roughly equal along the range 12-36%, with about 30% being near optimum.

[Neil B.]

Geoff: Of course you get a parabola at 28% with Norway there, the whole point is that Norway is an outlier. But even without that being the point, note that what really matters is how many points lie roughly equal along the range 12-36%, with about 30% being near optimum.

[Geoff]

Neil:

By definition, you get a parabola with a second order polynomial fit. The question in this case is the broadness of the curve the location of the maximum.

Even if you take out Norway, you see a parabola with similar features. It's broader and has a lower maximum but it's there (and actually fits the data better).

I'll try to make the numbers I compiled accessible later this evening. Obviously my numbers are estimates, so there are limits to their accuracy.

[andres]

Just out of curiosity, the outlier in the graphs that interests me is UAE, not Norway. How do the United Arab Emirates manage to provide basic government services (education, law enforcement, etc) with 0% tax revenues? Or does the data merely measure the tax revenues of the central government as opposed to the individual emirates? I'd appreciate being enlightened by anyone with experience in the UAE.

[anne]

Andres, the UAE is just an administrative designation. The is Kevin Hassett selective rubbish and only of significance for the absurdity and dishonesty.

[anne]

What I like is your way of noticing what takes a little imagination to notice. I always play with extremes in thinking. Nice.

[anne]

http://maxspeak.org/mt/archives/003184.html

July 13, 2007

UP AND DOWN THE LAFFER CURVE
By Max Sawicky

This (the chart) is the dumbest thing I've ever seen, and I've seen a lot. Check out that curve; it's a gotdam roller coaster! (Original WSJ editorial here.)

For fun I did a regression line for similar data (same year, same countries). I threw out those annoying communist outliers at the top (Luxembourg and Norway) -- countries with high tax rates and revenues. (By the way, when you include them without forcing a zero intercept, there is no significant relationship between the corporate tax rate (combined, central and sub-central govs) and corp rev/GDP. Very slight negative relationship. Throw out the two outliers and the same non-result (insignificant positive relationship).

[bigTom]

If I remove UAE and Norway, which both look like outliers, all my eye see's is scatter. That should mean that the X axis value is of very little untility in predicting the Y acis value.

Of course as (relative) noise increases, the amount of bogosity that a higher order fit generates goes up substantially.

[anne]

http://matthewyglesias.theatlantic.com/archives/2007/07/worst_editorial_ever.php

July 13, 2007

Worst Editorial Ever?
Eited by Matthew Yglesias

Via Brad DeLong, Marc Thoma finds a winner. The culprits, naturally, are at The Wall Street Journal editorial page, specifically Kevin Hassett.

That doesn't even remotely resemble a best fit curve. They've drawn the line straight-through an outlier. And look how steep it is at the right hand side. They're asking us to believe that the marginal impact of increasing corporate income tax rates above the Norwegian level is not only negative, but massively negative in a way that none of the non-Norway data bears out. It's an insult to everyone's intelligence. At this point, one needs to think that letting Rupert Murdoch destroy the WSJ news pages might be better for the world than letting the WSJ news pages' credibility continue to provide a "halo effect" to the editorial page.

[anne]

Big Tom has a nice sense of humor:

"Of course as (relative) noise increases, the amount of bogosity that a higher order fit generates goes up substantially."

[Jared Bernstein]

What's interesting to me is that the US has a relatively high statuatory rate yet a low rev/GDP ratio. That implies a low effective rate (liability as share of income).

The reason is that so much corporate income escapes taxation. There are loopholes like the one which let the Blackstone group actually collect $200 mil from the IRS after their $4.8b IPO (see today's NYT story by David Cay Johnston). There's devilish transfer pricing schemes, and there's the ~$20b gap between book profits and IRS profits.

I think we need another '86 reform moment: broaden the base, lower the rate, simplify, stop treating everything so differently (tax equity and debt financing at the same rates), etc...

[Ponzi Q. Globalization]

For heaven's sake, bigTom hit the nail on the head. If all we have are these rate-revenue points, then there seems to be little correlation between the rate and the revenue. You can mess around with parabolas, but your fit will suck. You can just look at the plot and see this.

Here's a game you can play. Fit the parabola and then use it to predict where other nations will be in one variable when you know the other. Use historical data or nations not plotted here. Hell, use the nations plotted here. The fit sucks anyway.

As I said before, rate alone is a crappy predictor of revenue and vice-versa. There are other factors involved. Maybe some of these can be quantified and we can have a nice n-dimensional plot that actually reveals something useful.

[James Killus]

"What did you expect? 'Welcome, sonny.' 'Make yourself at home.' 'Marry my daughter.' You've got to remember that these are just simple farmers. These are people of the land...the common clay of the [Wall Street Journal Editorial Page]. Y'know: morons."
--ht to "Blazing Saddles"

[allan_in_upstate]

But the Dow's almost 14,000.
So only 22,000 more to go.
See, Kevin Hassett isn't so stupid after all.

[ogmb]

Does this work?

0.0 0.0 UAE
8.0 2.5 [1]
12.5 3.6 Ireland
17.0 2.2 [2]
18.0 1.4 Iceland
19.0 2.0 [3]
19.0 2.5 [4]
22.0 3.4 Canada
23.0 5.8 Luxembourg
25.0 2.9 [5]
26.0 1.6 Germany
27.0 3.5 [6]
28.0 3.2 UK
28.0 4.8 [7]
28.0 10.0 Norway
29.0 3.6 [8]
30.0 2.9 [9]
30.0 3.2 [10]
30.0 3.8 [11]
30.0 5.7 Australia
33.0 2.3 [12]
33.0 2.8 [13]
33.0 5.5 [14]
34.0 2.3 [15]
34.0 3.6 [16]
34.5 3.1 [17]
35.0 2.2 USA
35.0 2.8 France
35.0 3.4 [18]

[ogmb]

Datapoints with second-order and third-order trendlines (via Excel):

http://i8.tinypic.com/66twex2.jpg

[Ragout]

Is there actually any reason to think that Kevin Hassett drew the graph? I assume that the footnote indicates that he supplied the tax rate data, not that he drew the "fitted" line.

And by the way, my impression is that most public finance economists would find the idea that the US corporate tax rate is past the peak of the Laffer curve pretty plausible. Jared Bernstein, a few comments back, might not agree, but he's given some good reasons to think that raising the corporate tax rate could decrease revenue. It's really easy for corporations to avoid the corporate income tax: they can shift their profits to lower-tax jurisdictions via transfer pricing, increase their use of tax shelters, tilt their financing towards debt rather than equity, etc.

[anne]

Jared Bernstein's reference:

http://www.nytimes.com/2007/07/13/business/13tax.html

July 13, 2007

Tax Loopholes Sweeten a Deal for Blackstone
By DAVID CAY JOHNSTON

The Blackstone Group, the big buyout firm, has devised a way for its partners to effectively avoid paying taxes on $3.7 billion, the bulk of what it raised last month from selling shares to the public.

Although they will initially pay $553 million in taxes, the partners will get that back, and about $200 million more, from the government over the long term.

The plan, laid out in the fine print of Blackstone’s financial documents, comes as Congress debates how much managers at private equity firms like Blackstone and hedge funds should pay in taxes on their compensation.

Lee Sheppard, a tax lawyer who critiques deals for Tax Notes magazine and has studied the Blackstone arrangement, said it was a reminder of the disconnect between the tax debate in Congress and how the tax system actually operates at the highest levels of the economy.

“These guys have figured out how to turn paying taxes into an annuity,” Ms. Sheppard said. “What people don’t realize is that the private equity managers, the investment bankers, all the financial intermediaries, are in control of their own taxation and so the debate in Washington about what tax rate to pay misses the big picture.”

The debate in Congress is about whether most of the compensation that fund managers earn should be taxed at the 35 percent rate that applies to other highly paid Americans, or at the 15 percent rate for capital gains.

Questions in Congress about possibly raising taxes on such compensation were prompted in part by publicity about the rich rewards for people who run these firms. Stephen A. Schwarzman, the co-founder of the Blackstone Group, made nearly $400 million last year, for example.

The Blackstone partners’ tax deal, however, is for the sale of part of their stake in the management firm, which is why their profits were taxed at the usual 15 percent tax rate for capital gains. Over all, the company raised $4.75 billion in the initial public offering, but the benefits of the tax structure involve just $3.7 billion of that.

Other private equity firms and hedge funds that have gone public, or plan to, make use of similar techniques, their documents show....

[anne]

http://economistsview.typepad.com/economistsview/2007/07/paul-krugman-an.html

July 13, 2007

Paul Krugman: An Unjustified Privilege
Edited by Mark Thoma

Paul Krugman urges Democrats to close the loophole that allows Warren Buffet to be taxed at 17.7 percent while his receptionist is taxed at around 30 percent:

NY Times: During the 2000 presidential campaign, Ralph Nader mocked politicians of both parties as “Republicrats,” equally subservient to corporations and the wealthy. It was nonsense, of course: the modern G.O.P. is so devoted to the cause of making the rich richer that it makes even the most business-friendly Democrats look like F.D.R.

But right now, as I watch Senate Democrats waffle over what should be a clear issue of justice and sound tax policy..., I’m starting to feel that Mr. Nader wasn’t all wrong.

What’s at stake here is a proposal by House Democrats to tax “carried interest” as regular income. This would close a tax loophole that ... basically lets fund managers take a large part of the fees they earn ... and redefine those fees, for tax purposes, as capital gains.

The effect of this redefinition is that income that should be considered ... ordinary income taxed at a 35 percent rate is treated as capital gains, taxed at only 15 percent instead. So fund managers get to pay a low tax rate that is supposed to provide incentives to risk-taking investors, even though they aren’t investors and they aren’t taking risks.

For example, the typical hedge-fund manager ... gets a fee equal to 2 percent of the funds under management, plus 20 percent of whatever his fund earns. It’s not exactly straight salary, but none of this income comes from putting his own wealth at risk. Except for the fact that he might make a billion dollars a year, he resembles a waitress whose income depends on a mix of wages and tips, or a salesman who lives on a mix of salary and commissions, more than he resembles an entrepreneur who sinks his life savings into a new business. ...

There’s a larger question one could ask: should we even be giving preferential tax treatment to true capital gains? I’d say no, because there’s very little evidence that taxing capital gains as ordinary income would actually hurt the economy. ...

But even those who ... think the special treatment of capital gains is justified, should be able to agree that treating the income of fund managers differently from ... everyone else who works for a living makes no sense. And that’s why it’s very disheartening to read that prominent Democratic senators are taking seriously the claims of fund managers that making them pay taxes like regular people would discourage risk-taking.

The immediate response should be: what risk-taking? ...

Look, this isn’t about envy, about punishing success. ... [C]losing the carried interest loophole should be a simple question of fairness: other Americans also earn their pay, but they don’t get special tax breaks. Plus, we’re talking about a lot of lost revenue due to that loophole — revenue that could, for example, be paying for the health care of tens if not hundreds of thousands of children.

And since we’re living in the real world of politics, there’s also the Republicrat issue: the hesitation of the Senate Democrats is terrible for the party’s image. It conveys the impression that they’re as beholden to hedge funds, one of the few types of businesses whose campaign contributions strongly favor Democrats, as Republicans are to the oil and drug industries.

So here’s a plea to Democratic senators on the fence: do the right thing and close this unjustified tax loophole.

[anne]

http://economistsview.typepad.com/economistsview/2007/07/yet-again-tax-c.html

July 13, 2007

For the record, his numerous sins notwithstanding, I don't think it's been established that Kevin Hassett drew the infamous curve on that graph. It's in the WSJ editorial, not in anything that he has done that I have seen. The data points in the graph are unexceptionable. I couldn't reproduce them exactly, but my own diagram doesn't look all that different [from Mark Thoma's].

-- Max Sawicky

http://maxspeak.org/mt/archives/003184.html

[anne]

Also, by the way, in this absurd dishonest presentation and analysis by the Wall Street Journal and Kevin Hassett, the United Arab Emirates is an administrative entity and levies no taxes and collects no tax revenue and should not have been included by Hassett and the WSJ.

Thanks for noticing to Andres....

[Matt]

I went back and read Kevin Hasset's complete research article:

http://www.aei.org/publications/pubID.14866,filter.all/pub_detail.asp

He structures the theme around international competition for attracting corporate clients. Hence, he phrases the issue in terms of market competition for government services, which is what government generally sells.

So, why this laffer stuff, why not use the standard supply/demand analysis?

Mankiw suggested the capital tax should be zero, implying (if he ever passed econ 101) that government should not sell services to corporations at all.

What is the optimum amount of government services that corporations need? The answer, emperically, should be a guassian distribution of amount vs price, to a first order, and I would apply a normal distribution.

Since we are talking about industrial economies, I wouold throw out those dominated by raw materials. In this case, he best result, so far, is about a 30% effective tax rate with 4.5% of government devoted to corporate services.

If you believe in existing markets, then this is the best we can do.

Another conclusion is that the laffer curve is a guassian distribution for those who cannot compute well with exponentials.

[D. F. Linton]

I entered the underlying data points by eyeball into Excel and fit linear, 2nd-, and 3rd-order polynomials. The R2 for the linear fit was -0.0183, for the quadratic 0.1666, and for the cubic 0.2371. Both of the higher order curves were downward sloping at the end. None of the curves go thru Norway (perhaps showing good taste?). But the linear is clearly a poorer fit that an upside down quadratic. Peak revenues for the quadratic occur in a gently sloped region around 4% of GDP at a 24% tax rate while the cubic implies just under 5% of GDP could be extracted at at 26% rate.

I anchored at curves at 0% of GDP for 0% tax rates. Snarky comments about the UAE not withstanding I think that we all would agree on that end of the curve.

Here are the curve equations:
y = 0.1246x
y = -0.6343x2 + 0.3193x
y = -5.6103x3 + 2.4059x2 - 0.0708x

Here are the data points I used:
10% 28%
5.90% 22%
5.80% 31%
5.70% 34%
4.70% 27%
3.70% 13%
3.60% 23%
3.90% 30%
3.80% 29%
3.70% 27%
2.50% 8%
0% 0%
2.10% 16%
2.20% 18%
2% 18%
1.30% 17%
3% 25%
1.50% 26%
3.20% 27%
3.20% 30%
2.80% 30%
2.80% 34%
2.20% 34%
2.20% 35%
2% 35%
2% 35%
2.60% 35%
2.80% 34%
3.50% 35%
3.60% 35%
3.80% 34%
10% 28%

Share and enjoy

[Neil B.]

Aside from what taxation percentages are optimal as such, how about trying a graduated corporate income tax? I mean, the percent levied on the profit would rise with profit margin (after a deduction, to avoid punishing little companies that have to markup higher to make enough money at all.) Then we reward the industries that serve the public by selling by volume at low profit margin, and penalize the “gougers” without having to directly regulate prices.

[John M 307]

I would say that the WSJ editorial plot, along with most crap produced by the propaganda right (such as the CEI's "They call it pollution, we call it life" ad), is aimed at those who can't think straight and who know next to nothing. In other words, it's BS or [bull] -- the people who drew the curve had no concern about the accuracy of the fit, merely its propaganda value.

They also don't care if some of us see through their BS, because we aren't their target. In order for ridicule and denunciations to be effective, they must be public and widespread.

As for the data itself, and fitting anything to it -- in situations where the data itself isn't immediately obviously linear, best-fit lines and curves are only useful if the parameters are accompanied by standard deviations and goodness-of-fit estimates. In this case, I think that it would be most instructive to do the following:

1. Fit a centroid to the data, excluding the Norway outlier -- call it A=(Ax,Ay). (Of course, you want its standard deviation as well.)

2. Fit a line through A: y - Ay = S(x-Ax), and present the best-fit slope S and its standard deviation.

I predict that the S would be small, and its standard deviation huge -- both compared with Ay/Ax. In other words, there's little relationship between y and x in the plot.

[Ponzi Q. Globalization]

I know economists would rather play mathematical games with simplified models only tenuously linked to reality than try to analyze the real world but this is getting ridiculous! No one is even talking about a model here anymore. You get some 2D data points and try to fit polynomials to them and hope for what?

Even I can come up with some justification for a second order polynomial but third order? Again try a thirtieth order (or thereabouts) and you get perfection.

But why stop at polynomials? Think outside the box and a whole world of cool functions opens up. Throw off the constraints of differentiability and, yes, even continuity! You will be amazed at what you can do.

Go further. Start using non-measurable functions. You will see just how much distance you can put between your analysis and the planet we all live on.

Again, again, adding more variables is the solution here. I imagine that this is very difficult but that's why many of you economists are paid so much. Well, the real reason many of you economists are paid so much is because many of you economists promote the interests of those who are currently wealthy and powerful. However, we can pretend it's because of your knowledge of real world economics.

[anne]

There is no snarky comment about the United Arab Emirates, simply the need to understand as was poointed out several times, that the UAE is an administrative designation and the UAE does not tax or collect taxes and of course should not have been included in the data.

Also, Norway is an amazingly successful economy and as successful a market in which to invest.

Plot what you will, and argue with Brad DeLong as you will, but the proper plot is the straight line drawn as a loose estimate of data of little meaning. The point of the Wall Street Journal is distortion and deception.

[anne]

PQG is completely right. I noticed the absurdly selectively data spread and fitted curve and laughed. The point of the Wall Street Journal editorial page is always mere distortion.

[George Smiley]

Jeebus. If they're going to make Luxembourg a labeled point, they may as well do the same for Toledo. Dipshits.

[andres]

Hmm. It's time to dust off my rusty knowledge of econometrics and make some points on D.F. Linton's hard work in actually turning the WSJ into regression specificiations.

(1) "I anchored at curves at 0% of GDP for 0% tax rates. Snarky comments about the UAE not withstanding I think that we all would agree on that end of the curve."

Sorry, can't agree with you there. Unless the WSJ made a misprint, the vertical axis measures tax revenues from all sources, not just corporate taxes.

[Not a safe assumption to make: the vertical axis looks like revenue from corporate income taxes. Except for Norway, where it looks like the vertical axis is the sum of corporate income taxes and oil excise taxes. But the horizontal axis is still corporate income taxes--if it were the sum of corporate income and oil excise taxes, Norway's tax rate would be roughly 52%.

Don't assume these people are even barely competent at anything.]

A government which had 0% corporate profit taxes would presumably still get revenue from sales taxes, tariffs, and whatnot, or even from direct transfers from state-owned enterprises which are not labeled "corporate taxes" in an accounting sense. ie there is no theoretical reason why the regression should restrict the y-intercept to the origin.

(2) Before running regressions, one should try to explain the nature of the outliers if at all possible. As anne mentioned, I suspect the UAE is a quasi-fictional central government which exists for foreign policy purposes only, and that the real governing and revenue collecting is done by the local emirates, which do not show up in the data. The UAE data point can thus be eliminated as an accounting error.

Similarly, the Norway outlier probably has something to do with a direct link between Norway's petroleum and natural gas earnings and its central government. A regression specification should try to include this factor as a regressor.

(3) Beware of data mining. One should choose a regression specification not because it gets the highest adjusted R2 but because the mathematical model behind it makes sense. For instance, one needs to have an economic model explaining why tax revenues are negatively related to the square of the corporate tax rate before one can confidently adopt a quadratic specificiation. Similarly other variables besides the corporate tax rate should be taken into account, eg the % of GDP produced by state-owned or semi-public enterprises. Without such forethough, one is more likely to obtain a spurious correlation which is a product of a single data sample rather than reflecting the actual economics relationships involved.

(4) If outliers such as Norway and the UAE don't have an explanation that results either in their elimination or in a respecification of the regression, then some loss function other than least squares should be used in the regression in order to reduce the influence of the outliers. In such a case, a least absolute residual or a least median of residuals regression might be more appropriate. But in this case, I think the outliers can be easily explained.

(5) Again, specification is critical. It is also quite possible that there is two-way feedback, eg a high percentage of non-corporate tax revenues might encourage policy makers to reduce corporate tax rates. In such a case, two-stage least squares or orthogonal regression might be more appropriate. More likely however, the relationship is one-way and outliers like Norway can be explained by respecifying the regression, eg by adding energy exports as a % of total exports or even accounting transparency measures as independent variables.

Once the data is cleaned up and the relationship is respecified, I'm pretty confident that practically all countries will be shown to be on the upward-sloping portion of the Laffer curve. The WSJ's inverted curve is simply bad econometrics.

[Ponzi Q. Globalization]

"PQG is completely right."

Thanks Anne. It's good not to be arguing about 'free-trade' or immigration with you for once (although I do silently agree with your posts on most other topics).

The WSJ editorial board has truly let it's ideology lead it into stupidity.

[anne]

Nice, Ponzi.

[Count Iblis]

I agree with Ponzi Q's comnment about the ad hoc fitting functions. And if we play that game, then it would be far more effective to take into account other variables.

We can see that for fixed corporate tax rate there is a range of tax revenues possible and that scatter is pretty large. So, you could just identify other relevant factors and do a fit with multiple variables.