It is Outlandish

[Note: There are five updates - including additional rebuttals to the posts at The Economist's blog from others - at the end of this post.]

The Economist blog, Free Exchange, criticizes me in a post called "Outlandish," but it's based on a false assumption about who said what.

We need to go back a bit. On June 27, I posted something from Stephen Gordon's site, a graph along with a sentence of two he had written about it. No comments of my own, just the graph, Stephen's comments, a link to the original, and the title "Social Spending and GDP per Capita" which is fairly neutral.

Free Exchange responded by implying the comments were mine, and though I could have been clearer about who said what than simply using the standard blockquoting to indicate the comments weren't mine, just a minute or two of investigation back to the original would have revealed that the implication in their post "Trompe l'oeil" that those are my comments is wrong. Stephen Gordon made it very clear in comments on the post at Free Exchange that he was the author of the graph as he defended it.

Today, in their post "Outlandish" they say I should be embarrassed for:

what seemed like an embarassing error on his part*

* To wit, illustrating a point with a graph which on not-particularly-careful examination shows exactly the opposite of what he was claiming

But, of course, I never claimed anything at all, and as I noted Stephen has defended the graph so whether it shows the opposite or not is open for debate. They also say I implied "that since we are a classical liberal paper, we endorse the Wall Street Journal's supply-sidism" when I can find nothing in my post that does anything of the sort. My point was about outliers.

As I said in comments over there, if anyone ought to be embarrassed by all of this, it is Free Exchange, not me. I think in the rest of their post they defend throwing out Norway as an outlier to support their contention about the graph, but to be honest I just skimmed it since there was no reason to believe that the quality of the post would be any better and hence no good reason to read carefully any further.

I was already planning to say a little more about outliers, the rest of the post at Free Exchange is responding to Kieran Healy's comments at Crooked Timber on this topic, so hopefully I can get some time to do that at some point. This paper of mine with Jo Anna Gray, "Financial Market Variables Do Not Predict Real Activity," though a bit dated, shows how outliers can influence estimates and conclusions, so it might be of interest. (Prior to publication there were a few emails back and forth with Ben Bernanke since part of this paper comments on work he did with Alan Blinder that brought the federal funds rate into prominence as a policy variable, e.g. see footnote 2. At the time the paper was written there was a "horse-race" between monetary aggregates, interest rate spreads, and the federal funds rate, with the federal funds rate eventually coming out on top, and part of the paper comments on how a particular outlier might influence the race).

Update:  In comments, Economist Blogger responds. Instead of taking responsibility for mis-attributing comments to me, then taking me to task for them, they have decided to take a cue from the Bush administration and remain aggressive, throw out falsehoods to support their case, and blame me for their mistake. The argument is, as far as I can tell, that because I didn't explicitly say I disagree with something, I must agree, therefore it's my fault they said I agree. My reply is in brackets:

1) Since you posted the link with no comment, I took it to be an endorsement of said argument. I'm not familiar with any blogger who posts links and graphics they think are wrong without pointing out that they're, eh, wrong. Have you posted, say, links to supply side arguments without comment?

[Response: Yes, I post lots of things for discussion, often with just a "so and so says" or very little comment. I didn't take a position whether it was right or wrong, as I said, it was for discussion. However, when people make obviously false claims, or distort things as in the Laffer curve graph, that is pointed out. But sure, things on the supply-side and other work are presented without much comment, rebuttal, etc., routinely, that's almost always true of academic work, e.g. NBER papers.

You seem to continue to want to hold me responsible for what Stephen said. If I didn't explicitly say I disagree, then I must agree, is that the argument? Sorry, but I can't take responsibility for his writing and his graph. And just so you know, I will continue to bring things up for discussion and, with any luck, we will all learn more as a result. For you, it's pretty simple from here on - if I don't actually say something, then don't say that I did, and don't attribute the position of the author of the piece to me. Attribute the opinions of the author to the author. Pretty simple.]

2) The words "feels like a strange jab at us" were intended to cover the possibility that you did not mean it that way; but that is certainly how it came across to us. Apologies if we were mistaken.

[Thanks. Again there's nothing I said that implies you endorsed the Laffer curve, nothing at all. If that's how it came across, maybe you are too sensitive on this point (being lumped together with the WSJ editorial page). I will leave it to you to figure out why that might be.]

3) Your commenters seem to be under the misimpression that we defended the Wall Street Journal, probably because you didn't excerpt any of the offending post. Allow me: ...

While we did not join Mr Thoma and others in branding Kevin Hassett a liar--we try to refrain from that sort of debate--we made it abundantly clear, I hope, that the chart is nonsense.

[I never called Kevin Hassett a liar. In fact, I thanked Max in comments for pointing out he wasn't sure Hassett was even the one who drew the line. Again, quit putting words in my mouth I didn't say. I don't generally call people liars here, even when they falsely accuse me of saying things I didn't say as you have. As for quoting you, why would I quote a source that makes false attributions or allegations?]. ...

Finally, on Norway: in this case, commenters who share Mr Thoma's general ideological cast, you <i>want</i> Norway gone. Norway doesn't help your case.

[Please quit implying my ideology drives my economics, it doesn't. You were the ones who threw Norway out and made a big deal out of something that disagreed with your ideology and continue to do so, not me. The point is to learn and move forward, not defend a position, and part of that is being able to discuss things without ideological blinders.]

And in this case, as almost ever when talking about the relationship between GDP and public policy, Norway <i>should</i> be gone. ...

I want to throw Norway out in both cases, because it isn't relevant. ...

[We have actual evidence and research papers on this - that was Stephen's point in his first comment on your first post. You can say what you think about this all day, fine, but at some point you might consider actually addressing the work Stephen has cited to support his case instead of eyeballing charts - something I was trying to make fun of in the post that got this all started - including the linear line I drew. There's also a whole lot more than just a chart in Stephen's post, but that has all been conveniently ignored with your continued strange refusal to actually engage the author of the graph and comments.]

Update: Brad Delong comments:

...Megan is trying to take a middle position between Mark Thoma's sensible criticism and Donald Luskin's idiotic defense of the clown show that is the Journal editorial page.

I see three misrepresentations by the Economist here:

1. The WSJ line is not "draw[n] through noisy data." It is drawn above noisy data.
2. To say that the WSJ line is "not... the obvious" one to draw implies that there might be some non-obvious reason to draw it. There isn't.
3. The claim that the WSJ line is "not the only... one to draw" is a statement that it is one of the lines that one might draw with some justification. It isn't.

All I can say is:

Questo misero modo/ tegnon l'anime triste di coloro/ che visser sanza 'nfamia e sanza lodo./ Mischiate sono a quel cattivo coro/ de li angeli che non furon ribelli/ né fur fedeli a Dio, ma per sé fuoro./ Caccianli i ciel per non esser men belli,/ né lo profondo inferno li riceve...

This is indeed the behavior of the banner-chasers of Dante's Inferno: those who did not have the morals to be worthy of heaven but also lacked the guts to sin enough to be worthy of hell, and who were thus rejected by both.

One more point, with respect to "omitting Norway": Personally I see no need to omit Norway. I do see a need to plot the Norway point on the graph correctly. The revenues plotted on the vertical scale include oil excise taxes levied on corporations. The tax rates plotted on the horizontal scale do not--hence the Norway "tax rate" of 28% rather than the correct 52%. Move Norway out to its proper position--with the same tax concept on both axes--and everything is fine.

Update: Stephen Gordon responds in "The social spending and GDP per capita graph redux:

For reasons unknown to me or to Mark Thoma, The Economist blogger doesn't seem to be willing to engage me directly in the interpretation of the following graph from this post...

My original comment on this graph was: 'These are countries whose per-capita incomes are greater than the OECD average. The point here is that there's no trade-off between high levels of national income and high levels of social spending.'

Now, it seems obvious enough - to me, at least - that

• The Nordic countries are rich, and
• The Nordic countries spend lots more on social programs than do Canada or the US.

I could have left it at that, I suppose, but I also wanted to illustrate the range of variation of the horizontal axis. To my mind, the variation in the GDP per capita numbers says more about the vagaries of cross-country GDP data than it does about comparing living standards: the countries plotted in the graph are rich, and that's pretty much all you need to know.

The interesting question is *how* the Nordic countries accomplish this feat, and that's what the post was about.

Update: Kieran Healy at Crooked Timber also responds. He's not happy either:

Dept of Being Savaged by a Dead Sheep, by Kieran Healy: Someone I believe to be Megan McArdle weighs in at the Economist blog on the laughable graphic run by the WSJ the other day. Brad DeLong is not impressed, nor is Mark Thoma (in part because comments are misattributed to him in the post), and nor am I. She singles me out for membership in “a special category of wrong,” I think mostly because my Ph.D is in sociology and not economics.

In his own special category of wrong is Kieran Healy, who vindicates an old professor’s precept that “sociologists rush in where angels fear to tread.” In this case, he lectures a bunch of economists (and economics journalists) on their temerity for throwing Norway out of the set: … [...]

Megan doesn’t like me much in part because she’s has a long history of making an ass of herself on the topic of economics and its status as a social science. Again, just bear in mind that the original post was prompted by people saying “I want to see this graph/regression estimate but with Norway (or some other country) removed as an outlier.” After the inevitable snipe at sociology, she concedes most of the point I was making, which was just that when working with small cross-national datasets—e.g., of broadly comparable rich industrial democracies—deleting some countries as outliers in order to make the quantitative analysis go more smoothly is generally a bad idea. It typically doesn’t make much theoretical sense and it also usually isn’t technically necessary because you can use more robust estimators than OLS if you need to. This is just a corollary of a more general principle that one shouldn’t throw away observations to make your life easier. The post was intended as a corrective to the several commenters on various blogs who seemed to think that picking which cases to drop was the starting point of a quantitative analysis.

This simple point applies to the debate about the WSJ graph, but the fact that the figure was basically a dishonest piece of rubbish sort of drowns it out. The original graph included Norway, but probably for the dishonest reason that you could draw the ridiculous curve around it as a result. When discussion started about what the “right” line should be, some people began with the assumption that Norway should be left out just because it was well off the main cluster of points. My post gave some reasons why that wasn’t really necessary, theoretically or methodologically, although—as I said—the basic stupidity of the graph meant that this was something of a subsidiary point.

Megan asserts that Norway’s status as an oil producer means that “economists, on left and right, are often willing to agree to ‘Omit Norway’ as a first order of business when discussing developed economy GDP” because “’find oil’ and ‘experience a rapid appreciation in fossil fuel prices’” are not replicable policy recommendations.” But what exactly is the claim here? If we’re talking about discussing the determinants of GDP in general in developed economies, then every rich industrial democracy has a raft of important, specific, non-replicable features of this sort. A policy of discounting them would quickly mean your “first order of business” would leave you with no cases to discuss. Such features might include, “Have No Feudal Past,” “Have a very large proportion of your population under 25 at the right time,” “Be an island with big coal deposits and start growing rapidly in 1780,” “Have a lot of nice ports and alluvial plains,” “Have a lot of Protestants,” “Be in the middle of Europe,” “Have three to twenty times as many people as the other guys,” or what have you.

Norway certainly isn’t ruled out from consideration for its oil any more than the particular features of other countries rule them out of play. (Besides, Norway is not Nigeria—clearly its government is managing its oil endowment in ways that are within the domain of policy, as opposed to gifts of nature.) It’s facts like this that give rise to what Stephen Gordon, in a related post calls the “vagaries of cross-country GDP data.” My original post, which Megan quoted at length but seems not to have understood, made that point. When generalizing across countries you will want to be clear about your scope conditions, which can include stuff like Norway’s oil industry if it’s relevant. This is different from dispensing with the case altogether. Moreover, demanding a restriction of discussion on GDP growth to only those variables that are strictly within the reach of policy might make it easier to believe that policy-makers can control the world, but you will certainly be ignoring a lot of information about the much more important question of why certain countries are rich—especially those that stubbornly do not fall into line with your favored policies. Of course, such considerations are far from the minds of believers in the Laffer Curve As Economic Law. Nor are they much to the fore in Megan’s mind, her oft-stated belief in the scientific content of economics notwithstanding.

If we’re just talking about the Laffer Curve as policy in particular, rather than GDP growth in general, I presume Norway was included in the original figure precisely because of its apparently high corporate tax rates, which allowed a curve to be dishonestly drawn around it. But as subsequent discussion has shown, in the end there’s really no reason to “throw out” or “drop” Norway in order to say why the curve is stupid, or to calculate a sensible and robust regression estimate for the relationship at hand, in contrast to what some people were originally saying when they saw this ridiculous figure.

I also believe Megan wrote this, but do not know for sure as they don't have bylines on their blog posts. That generally indicates the opinions are those of the publication itself, not the individual, so it is The Economist (or at least the Free Exchange staff) that is saying these things, not any one person. For example, Megan (I think) has stated that she talked over the first post with others so as to deflect responsibility for her post and statements onto the whole group.

Update: One more (and with any luck the last) thought on all of this. Stephen Gordon has not been given the respect he deserves as all of this has unfolded. One of the complaints from The Economist is that I didn't take exception to Professor Gordon's graph when I posted it.

I would no more trash him or his work when presenting it here, or simply dismiss it before investigating further, any more than I would trash someone or dismiss their work when introducing them at an academic seminar. There are many who have not earned or have given up that sort of a priori respect, but he is not one of them.

The graph I posted came from a presentation he did at the Canadian Economics Association meetings (by invitation) based upon work he had presented on his blog in the past. It's not just a graph, it's a whole set of material, including comments from a discussant and the sources he used for the analysis. I read everything he posts and my experience is that the work is analytical, often data based, and employs the tools and techniques of a professional economist. His work is worthy of asking questions and giving him a chance to respond before simply jumping to conclusions one way or another.

While not everyone will get this respect, Professor Gordon has it, and I will continue to post his work, as I have in the past, when it addresses interesting and relevant questions for the visitors here. I may raise questions when I post something, or I may simply let him speak for himself as I did last time, but I won't automatically dismiss anything.

I encourage all of you to follow his blog, or to go take a look around. He doesn't post often, but it's more than "Worthwhile" when he does (the name of the blog is Worthwhile Canadian Initiative).