Worst Editorial Ever?

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.

Comments

The Laffer Curve keeps changin' shape. What a flexible little idea.

Posted by Jeffrey Davis | July 13, 2007 1:52 PM

Hmm, look how close France and the U.S. are on that graph.

Posted by TDE | July 13, 2007 1:52 PM

I don't understand why anyone would think that tax revenues/GDP is a function of corporate tax rates. Maybe corporate tax revenue/GDP might be a function of corporate tax rates - but "tax revenues" includes VAT, which differs greatly among different countries and doesn't exist in the US, and income tax, which also varies greatly. So the whole thing is nonsense.

Posted by Bloix | July 13, 2007 2:15 PM

I don't understand why anyone would think that tax revenues/GDP is a function of corporate tax rates. - Bloix

And if that was the point they were trying to make, why plot the corporate tax rates on the y-axis?

The curve they drew doesn't even make any sense as a "partial" Laffer-type (assuming the Laffer phenomenon is "decomposable" by tax type) curve (which would go through the origin, reach a maximum point -- i.e. be the most to the right at some level of corporate income tax -- and then go back to the left again). Certainly Norway is on the "lowering taxes will increase revenue side", but the USA certainly is not -- if anything, we are on the side of the curve where raising taxes would increase revenues.

Posted by DAS | July 13, 2007 2:22 PM

Oops ... never mind ... they did draw the axes right. I just got confused by the labeling.

Sheeze ... if I showed a plot that was that unclear like that to my boss, I'd never hear the end of it.

I wonder if the lack of clarity is done that way on purpose ... hmmm ....

Posted by DAS | July 13, 2007 2:24 PM

It's a Laugher Curve.

What?

Posted by John Rosevear | July 13, 2007 2:27 PM

Statistics don't lie, but statiticians...

Posted by Maddog51 | July 13, 2007 2:45 PM

I love the UAE datapoint - Zero percent corporate tax leads to zero corporate tax revenue. Who would have thought?!!

Posted by Robert | July 13, 2007 2:49 PM

http://www.aei.org/scholars/scholarID.26/scholar.asp

Professional Experience
-Senior economist, Division of Research and Statistics, Board of Governors of the Federal Reserve System, 1995-1997
-Economist, Division of Research and Statistics, Board of Governors of the Federal Reserve System, 1992-1995
-Associate professor, Graduate School of Business, Columbia University, 1993-1994
-Assistant professor, Graduate School of Business, Columbia University, 1989-1993

Education
Ph.D., economics, University of Pennsylvania
B.A., with honors, economics, Swarthmore College

Posted by Robert | July 13, 2007 2:54 PM

I'm pretty sure the left axis does, in fact, show corporate tax revenues as a % of GDP. Federal corporate tax revenues in the US were 1.2 percent of GDP in 2003; add in corporate taxes at the state level and you could get the 2% we see on the graph. Of course MY is right that calling it a Laffer curve is absurd, but they are at least comparing apples to apples.

Posted by lemuel pitkin | July 13, 2007 3:01 PM

Here, courtesy of the link provided by Robert, is his email address:

Now we can all write him and make fun of his obvious stupidity.

Posted by Joe Klein's conscience | July 13, 2007 3:04 PM

It's an insult to everyone's intelligence.

Only if you assume that they're producing this sort of thing to bolster the claims of people who are both a) intelligent, and b) intellectually honest. Since there's no evidence of that...

Posted by latts | July 13, 2007 3:22 PM

Klein's conscience,

What, you didn't place a direct call to his office? :-)

If AEI had an integrity, Hassett would be summarily fired, but I answered my own question with my premise...

Posted by Robert | July 13, 2007 3:22 PM

A question.

Does the graph in question refer to nominal corporate tax rates, or real tax rates?

Because leaving out the various tax deductions and exemptions enjoyed by big corporations would obviously have a major effect on the revenue side.

Posted by anonymous | July 13, 2007 3:37 PM

This from a non-economist, science-type:

What is the functional form of a Laffer curve? I ask this because the faults of the Hassett curve above include not only the inclusion of an obvious flier, but also in not making clear the assumptions in the plot. When fitting a curve to data, one is applying a set of assumptions encapsulated in a mathematical model. As one reader noted, of course one expects that, as the abcissa approaches a 100% tax rate, then the revenues will approach zero. This is why the best-fit line (san fliers) is also not valid?

Unfortunately, you can't spell out a complicated math model in WSJ editorial (too science-y for the creationists), so how to explain? In any case, it would be nice to see an attempt at a rational explanantion of assumptions in a model before one shows off a curve and makes grand pronouncements based on ditto.

Posted by rlgordonma | July 13, 2007 3:57 PM

Robert, you forgot 'author of Dow 36,000'.

Posted by Barry | July 13, 2007 4:05 PM

As a Norwegian 'flier' I suggest our outlier is from the fact that the general business tax rate 28%, although our (huge) oil industry is taxed at 78% (yes, seventyeight), and hence we get a very high number with a 'modest' general rate.

Posted by Morten | July 13, 2007 4:11 PM

Let's all remember that, in the end, individuals pay tax and corporations don't. They simply pass it through as a cost of doing business, which it is. To the individual, it's a hidden tax that they ultimately pay.

Posted by Fred Jones | July 13, 2007 4:38 PM

First, the data points seem to support their argument. If revenues increase directly with tax rates, why do the points to Norway's right top out at Australia and drop off from there?

Second, even if the curve is intended as a best fit (the editorial doesn't say so), doesn't the reach up to Norway argue for higher tax rates up to that point? Excluding Norway would decrease the curve's slope before that point. The curve as drawn suggests that rate increases to Norway's level generates more revenue than a curve that excluded Norway. But once the curve is up there, it has to get down somehow, and you get the cliff.

Take your pick: either a curve that exaggerates marginal revenue increases from rate increases to Norway _and_ marginal decreases after, or a curve that ignores Norway and has less dramatic marginal effects on both sides. I'd prefer to exclude the outlier but it isn't an obvious choice.

I'd say choosing the most dramatic of two plausible curves has more to do with the editorial penchant for drama than it does malice. Or should we be reading more into the operatic language on the Times' editorial page?

Posted by chernevik | July 13, 2007 5:01 PM

Oh I don't know, a good graph summarizes a large volume of information in a dramatic, easily understood way.

Supply-side economic theory really IS as laughably ridiculous as this graph makes it look.

Posted by Jalmari | July 13, 2007 5:10 PM

Second, even if the curve is intended as a best fit (the editorial doesn't say so), doesn't the reach up to Norway argue for higher tax rates up to that point?

That the reach up to Norway argues for higher tax rates up to that point is irrelevant to whether the curve was intended as a best fit. Anytime a curve is drawn over data points it should be a best fit curve (or at least as close as possible). To draw a curve any other way is to intentionally misrepresent the data. Or maybe it is just incompetence. But I'm sure Hassett knows how a curve is fit to data points, so my money is on intentional misrepresentation.

And why in the world would the editorial need to actually state that the curve was intended as a best fit? That should be a given unless some explanation is given for some other drawing of the curve.

Posted by Brian | July 13, 2007 5:29 PM

FYI -- here is somebody who claims to have done the regression on the data (excluding Norway as an outlier):

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

Posted by Brian | July 13, 2007 5:34 PM

Not exactly the first time the WSJ editorial page has been called out for its misuse of graphs, isn't it?

Posted by Reality Man | July 13, 2007 5:52 PM

rlgordonma, I don't think the Laffer curve has a well defined parametric form but I guess most people use a skewed bell curve and regress on the data to determine the mode. Hassett seems to have determined this mode by using only the Norway data point, thus neglecting the rest of the data. At any rate, the Laffer curve is oversimplistic BS that assumes concavity when there is no good reason to believe there is.

Posted by michesmith | July 13, 2007 5:57 PM

Fred Jones,

If you really think that by cutting corporate tax rates corporations are going to reduce their prices by a corresponding amount, I've got a Patent for a Left handed smoke shifter I'd like to sell ya.

Posted by Robert | July 13, 2007 6:03 PM

The scary thing is this graph alone would be enough to convince Ronald Reagan of his rectitude. I don't think Bush even cares enough to justify his positions with skewed graphs-- not that that stops his flunkies at the WSJ from trying.

Posted by Gregorio | July 13, 2007 6:08 PM

Matt--

Could you get Sullivan to cite Mark Thoma too? He did the work, and it would be nice if the people who are being informed by his work at least know his name...

Brad DeLong

Posted by Brad DeLong | July 13, 2007 6:24 PM

The sad thing is, there really is some decent empirical support for the Laffer curve. But this abortion from the Wall St. Journal is not it. They clearly don't know what a best-fit curve is -- or else they do know and they lied about it. BTW, the guy who did the regression on his spreadsheet appears to have done a linear regression, which is not a Laffer curve either -- the formulations I've seen are usually quadratic.

Posted by Barton Paul Levenson | July 13, 2007 6:35 PM

What is the functional form of a Laffer curve?

I wondered about this, too. Here's a different, related question: The most common way to fit a function to data is to minimize the sum of the squared residuals (vertical distances from the data to the prediction--the points to the line, in the case of a linear regression). Just what is being minimized by this function, aside from Kevin Hassett's credibility?

Posted by RSA | July 13, 2007 7:31 PM

Let's all remember that, in the end, individuals pay tax and corporations don't. They simply pass it through as a cost of doing business, which it is. To the individual, it's a hidden tax that they ultimately pay.

Let us all remember that corporations don't actually pay their employees' wages, they simply pass them through as a cost of doing business...

Posted by Pooh | July 13, 2007 8:16 PM

there really is some decent empirical support for the Laffer curve.

Here on the internet, Barton, it's traditional to accompany a statement like this with a link.

Posted by lemuel pitkin | July 13, 2007 8:40 PM

So Norway stumbled upon the optimal corporate tax rate! All of our problems (at least of a financial nature) are solved. Deficit-schmeficit. If the US simply reduced the corporate tax rate to 27% we too would optimize tax at 10% of GDP, thereby increasing tax revenue by $1.033 trillion. Why didn't we think of this earlier -- we could have funded the costs of two additional and simulataneous "wars in Iraq" and still had $148 billion for the domestic agenda. Good for both the left and the right.

Posted by cahay | July 13, 2007 9:04 PM

michesmith, thanks for the insight.

RSA, I am a fan of Gershenfeld's book on mathematical modeling:

http://www.amazon.com/Nature-Mathematical-Modeling-Neil-Gershenfeld/dp/0521570956#ggviewer-offsite-nav-9122208

Thanks for reminding me that it is not only the model that dictates our assumptions, but also the metric we use to choose the parameters for said model. An interesting point from the book is that least squares minimization makes sense when the deviations form the model are truly Gaussian distributed, i.e. random. However, that makes some rather drastic assumptions, i.e., the likelihood of getting an outlier like Norway is, well, extremely small. As Morten explains, there is a reason for the outlier behavior.

In such cases, it may be more reasonable to assume that the error is not Gaussian distributed, but has more slowly decaying tails. The math gets more complicated, and matrix algebra doesn't do the trick anymore, but it leads to more robust estimation. This is something that I would hope the Hassetts of the world would be privy to.

Anyway, I am probably getting way too technical for this post, but my point is that there are way too many ways to interpret data for definitive conclusions to be drawn without proper explanations of their context. I'm no economist, so I don't know what the right procedure for the data analysis is, but someone like Hassett should be more forthcoming about this.

Posted by rlgordonma | July 14, 2007 12:14 AM

Let x0 = 0, and x1 = the corporate tax rate of Norway. Let y1 = the yield of Norway.

It's then obvious that the solid curve f(x) is a minimum of the functional L[f]:

L[f] = \int dx\, (\delta(f(x)) + \delta(f(x) - y1))

Posted by matt | July 14, 2007 3:05 AM

Oops, stupid mistake. The correct functional is

L[f] = \int dx\, (f(x)^2\delta(x) + (f(x) - y1)^2\delta(x1 - x))

Whew. Glad I caught that before the Journal.

Posted by matt | July 14, 2007 3:10 AM

The weird thing is, it isn't even really a Laffer curve. The Laffer curve is supposed to represent the total tax collected as a function of tax rates, not "percent of GDP collected as tax" for a given tax rate. The graph may be a representation of "here's what we pretend the tax rate is, and here's what it actually is", but there's no way to draw any Laffer-esque cause-and-effect type conclusion from this data since the Laffer argument is based on the supposition that GDP (not GDP/Tax revenues) is a function of tax rates (i.e. if you have to pay 100% in taxes, GDP will fall to 0 since nobody will work)

What about Norway? Poor Norway, they pay a lot in taxes. What about us? We should stop bitching, notwithstanding a relatively high nominal corporate tax rate, the actual tax we pay on GDP is lower then everybody but Iceland, Germany and the UAE.

Posted by Josh | July 14, 2007 4:04 AM

Also, in many ways "tax revenues as a percentage of GDP" are a very different statistic from "tax revenues" or a per capita variant which a study of the Laffer Curve is supposed to maximize. Such a curve (a Hassett Curve), could conceptually be shaped like the letter N.
The UAE zero-point, then Norway (and in fairness to the AEI and WSJ Australlia and Luxembourg), then other countries until the corporate taxes get so high that they take up almost all of the GDP (which the high taxes lower), leading to another peak at, say, Cuba, as far as I can tell. Actually Norway and Luxembourg are such wealthy countries, that a real Laffer Curve may actually result from such a plot.

Posted by Wido INcognitus | July 14, 2007 5:22 AM

Thanks for the reference, rlgordonma. I'll have to take a look.

Posted by RSA | July 14, 2007 8:37 AM

cahay seems to have nailed the lesson here. We should mimic Norway's tax code. 10%! Whew. So, France or Germany's health care and Norway's tax code and we'd have so much money we could reduce our CO2, too! And take vacations in Thailand to relax.

Posted by Jeffrey Davis | July 14, 2007 10:27 AM

Whkat should be done is to challenge the WSJ to post the R^2 values for their curve and a linear regression. I will bet the ranch that their curve has a significantly lower value then a linear regression.

Posted by SLC | July 14, 2007 3:24 PM

SLC,

A lower R^2 value is not the point. The question is, what's the model, and what metric is used to evaluate the fit to that model? The model contains all of the assumptions. The R^2 value gives the sum of the squares of the errors of the data w.r.t. the model.

I think linear regression is a poor choice for the model, for reasons mentioned above, so a lower R^2 value is a red herring.

Posted by rlgordonma | July 14, 2007 6:45 PM

Morten already explained Norway case. With coordinates defined so-sloppily, it is possible to have quite healthy revenus from zero tax rate (I would not find it strage if Emirates taxed oil industry like Norway does and actually report revenue), and flurishing enterprises in the presence of 100% tax rate (and a slew of loopholes).

Otherwise, Ireland taxes profits at 13%, collects 3.7% of GNP that way, so profits can be extrapolated to 30% of GNP. USA collects 2.1% at 35%, so the profit can be extrapolated to 6% of GNP. The latter is surely totally false.

Posted by piotr | July 15, 2007 12:56 AM