A test of Marginal Revolution political bias

Here is an email from Daniel Stone at Bowdoin, I am not imposing a double-formatting on it for ease of reading and formatting:

“Dear Tyler (if I may),

I’m a big fan of your work in general, and MR in particular, and think that you do as good a job as anyone at exploring a variety of political perspectives, and sharing related (diverse) research.

Still, you’re human after all J. I’ve always been curious if there are systematic patterns in your writing or links you post.

It occurred to me a couple weeks ago that you sometimes describe research as speculative or imply this by adding a question mark to the end of the link (the example that made me notice this was: “Minimum wage effects and monopsony?”https://marginalrevolution.com/marginalrevolution/2019/07/thursday-assorted-links-215.html). At other times your link simply states the main research finding or directly quotes from the paper or its title.

So, while it might be hard to identify a general bias in your links – even if the majority were, say, “pro-liberal”, this wouldn’t necessarily mean *you* were biased, since the majority of good research out there could be pro-liberal, using the added “?”s provides an identification strategy: if you were more likely to add a ? for research that leans one political direction or the other, that would suggest a bias on your part.

As a fun side project, that I thought might also have some value given the importance of MR and understanding bias more generally, I had my RA (Maggie Hanson, cc’d) grab all your links from Assorted Links posts to social science research this year (as of a few days ago). Together we coded the ‘slant’ of each as L, R or N (neutral) – depending on whether the research supports regulation, indicates market failure, etc (admittedly our process here was not extremely scientific). She also recorded whether your link text is phrased as a question (or notes that the finding is speculative, which you did a couple times and seems similar). In addition, for link text phrased as a question, we also noted whether this text is a direct reference to the research paper’s title, as this means you didn’t actually add the “?”.

We did a bit of very basic analysis, here are results:

The distribution of slant across links is quite balanced, but leans left:

. tab sla

Slant |

(L/N/R) |      Freq.     Percent        Cum.


L |         35       29.17       29.17

N |         58       48.33       77.50

R |         27       22.50      100.00


Total |        120      100.00


But you were slightly more likely to phrase your link as a question for “L” links vs for Rs (9/35 for Ls vs 5/27 for Rs):

.   tab slant endswith

Slant |      Ends with ?

(L/N/R) |         n          y |     Total


L |        26          9 |        35

N |        48         10 |        58

R |        22          5 |        27


Total |        96         24 |       120


And you were a bit more likely to do this for links that were not direct quotes of article titles that were questions (7/33 = 0.21 for Ls vs 2/24=0.083 for Rs):

tab slant endswith if linktex==”n”

Slant |      Ends with ?

(L/N/R) |         n          y |     Total


L |        26          7 |        33

N |        48          8 |        56

R |        22          2 |        24


Total |        96         17 |       113


But the magnitude of this difference is not large (and I bet not statistically significant), and the large majority of both L and R links were presented by you without questions marks.

Bottom line: you do present a quite balanced set of research findings, the general distribution leans left but it is hard to interpret this (without knowing the slants of research in general or the slant of research you post elsewhere, aside from Assorted Links). And there is suggestive evidence of a small tendency for you to be more questioning of research supportive of liberal/leftist policies.

Here’s a link to the data:


This includes a sheet with all the links that end in ?, that aren’t quotes of article titles, and their slants.

I wanted to share this with you before sharing with others. Please feel free to let me know any questions or comments!

Thanks, and thanks again for all your work. All the best – Dan”


'I wanted to share this with you before sharing with others. '

Makes one wonder whether Prof. Cowen returned the courtesy.

Speculative, obviously.

Additionally, it seems impossible to raise a public policy topic without spawning a hate tornado.

Whose alternate facts is one to believe? The decisions turns on ideology.

And to think I am old enough to remember the age before there were 'alternate' facts. As is everybody who recently finished third grade, for that matter

Particularly as the existence of facts is utterly indifferent as to whether anyone believes them or not.

Seems to be an old legal term, when the one side presents the facts that might not match those the opposing counsel is presenting. Likely you are not older than it.

Sort of (the facts remain facts in and of themselves), but it is interesting to see how such a defense of 'alternative facts' actually supports the idea that 'ideology' is the basis, and not the facts themselves. 'Before we move on to some nonlegal examples of alternative facts, here is but one of many ways the concept can be employed in an actual case. Of particular note is the clause “provided there was plausible evidence to support both alternatives.” In other words, two points of view may differ and still both be truthful.' https://www.americanthinker.com/blog/2017/02/alternative_facts_a_common_legal_term.html Of course, the sleight of hand involved in going from facts to 'truth' may be just a bit much for some people to easily accept.

I think the point is that your lawyer will cherry pick the facts that are beneficial to you, and leave out the ones that are not, and vice versa. All the facts presents are facts, but not necessarily all the facts of the case. In the case that made this news, Kellyanne Conway's use of it, Conway was using it as a way of saying she's providing additional facts that were left out.

No, but if Chuck Todd says they are false, then likely they are not. I don't even remember what the convo was about, but that's the way alternative facts is used

Well, the conversation, when talking about crowd size, was about Spicer telling an explicit falsehood on his first day - https://www.youtube.com/watch?v=PKzHXelQi_A

Facts do not change due to belief, and the fact remains, Spicer's claim about crowd size was false, regardless of Conway's attempt to use 'alternative facts.

Oddly, now that Spicer is not employed by Donald Trump, this is what he says about his falsehood that Conway attempted to defend - 'Spicer, who served as President Trump’s first press secretary, got off to a rough start with reporters when he falsely claimed the crowd gathered at the 2017 inauguration was “the largest audience to ever witness an inauguration, period, both in person and around the globe.”

Even as reporters pushed back and photographs showed that Trump’s crowd appeared much smaller than the one that had gathered from Barack Obama, the new president and Spicer stood by their claim. Since his resignation from the White House last summer, Spicer has tried to rehabilitate his image and told the New York Times in September that he regrets attacking journalists over the crowd size issue.' And again, you can watch him say that at the link - https://time.com/5088900/sean-spicer-screwed-up-inauguration-hitler

I am really delighted to glance at this blog
posts which contains lots of valuable information, thanks for providing
these kinds of information.

However they themselves were the ones, who were judging whether the links were left/right/neutral. Since the neutral position can change it place with changes in politics/time/place/country and so on and so forth, what I see here is not 0 bias, but that your 2 biases are close to each other.

Your mistake is confusing this cultural performance with research. These academics simply wanted to make sure Cowen was labelled a "left of center" academic like them, and to reaffirm his value as an "objective" source of ideas. It was done to flatter him and reassure themselves, all under the guise of a scientific process. They dressed it up in statistical sampling to add legitimacy, but there is no explanatory power to their work, and thus no value.

What is the meaning of political "bias"?

Is it systematically off the center of political opinion?

Or is it systematically off political wisdom?

Can the center of political opinion be biased?

Hi Dan (Prof Klein?)
Thx for this. I think your questions relate to the interpretation of the general distribution of L/R in links (as potentially reflecting bias), but are less relevant to the relative distribution of ?s in L vs R links. For this, I think the definition of bias is more clear (if ? is systematically added more often to either L or R links, that would reflect bias in presentation of the links by TC - but could probably still be rationalized by a Gentzkow & Shapiro 06 model, if that's what you're getting at). Re your 1st Q on the definition of bias, I like discussion in Gentzkow et al Handbook of Media Econ chapter :), https://www.nber.org/papers/w19880.pdf

Bayesian persuasion (?)

Prof. Stone's analysis omits an extremely important variable: the quality of the article (or research) that Tyler is linking to. Surely an article with questionable conclusions is more likely to draw a "?" from Tyler, regardless of its or his political leanings.

In addition, sometimes Tyler makes his opinion clear, with statements such as "self recommending" or "I do not agree with their conclusions but interesting nonetheless" or the like. In those cases, Tyler is making his evaluation more explicit and detailed than when he merely adds a "?".

But we run into the same problem; in some cases Tyler is clearly saluting the research that someone has done, while not totally agreeing with it or its conclusions. Or vice versa, he might agree with someone's political stance but find flaws in their research.

So, many of Tyler's comments have considerable nuance, even the very short ones (including, I would argue, the "?" ones).

The research might be on more solid ground if it looked at the articles that Tyler judged to be "self recommending". I'd guess there's less ambiguity and nuance there, compared to his "?" comments.

I was thinking along the same lines, but Tyler did pass this on without correction.

True, but he didn't append a "self recommending" to it.


Good pt, thx. I do implicitly assume quality is not a confounding variable. I think this is a reasonable starting pt but should be examined if analysis was taken further (which I don't plan to do).

But can look at outlets (listed in spreadsheet) to get a sense of this issue. Of the 7 L links w ? not quoting paper title, 2 were published in top PS journals (APSR, BPSR), and 1 is NBER. Neither of the 2 analogous R links were published in a top journal. And there are both L and R links without a ? that were published in less reputable journals (eg Applied Economics Letters). So, at first glance at least, ? does not seem to correlate highly w quality.

'Prof. Stone's analysis omits ....'

Pretty much anything that one could consider interesting, as a libertarian should not actually have much of a right/left framework. That being part of the true appeal of libertarianism to anyone that finds binary distinctions of no use except to those who feel they can take advantage of simplistic framing to advance their own goals.

The L/R slant word-clouds of summaries:

L http://i63.tinypic.com/1z2ftjo.png

R http://i65.tinypic.com/1253j14.png

Question: why did Dan spend all this time trying to prove if Cowen leans left or leans right in his blog posts? If I were to undertake a similar exercise I would make a distinction between orthodoxy and heterodoxy. There are scholarly papers about the difference (the schism?) between orthodox economists and heterodox economists. To quote one such paper: ""Heterodox economists are the dissenters among economists. The field . . . is dominated by thinking that solidified in the nineteenth century and became orthodoxy. . . . Mainstream economics . . . uses its models to predict the future. Heterodox economics is not about predictions per se . . .; it is about explaining the reality of the economic aspect of human behavior within the social context."

I'd like to think that Cowen reads all those books and papers not to reinforce what the already knows, but to learn something new. Indeed, he has described his well-known speed reading ability that way: He has read so much that he can quickly and easily identify something new in what he is reading; if there's nothing new, he becomes the economists Johnny Appleseed.

The real question is why Prof. Cowen would care in the least what someone else considers his political leanings.

Unless, of course, one views this web site from a PR perspective where perception is the primary criteria.

Not that any public choice economist would ever be so cynical about motivations, apparently. Or at least not openly.

Wait, is TC a libertarian or a public choice economist? I’d say neither, but then I’m not an economist. But c_p apparently views him as a cynical empty suit. Or sweater.

Fame is so apparently fleeting, regardless of whether one directed the James M. Buchanan Center for Political Economy, or is the current director of The Center for Study of Public Choice.

What, you were unaware of the fact that Prof. Cowen and Prof. Tabarrok are leading institutional lights of the Virginia School?

Admittedly, one of them does seem to have finally stopped playing at being a libertarian, though opinions continue to vary on the subject.

Why have we become obsessed about left and right? So obsessed that we find real or imagined leanings in everything we do, everything we say, everything we write, even the clothes that we wear. One my age can remember a time when it wasn't so, when Republicans and Democrats were identified by stereotypes that weren't threatening but amusing, amusing in a way that revealed how superficial most differences actually are.

Today, "[s]ocial media drive us, incessantly, into opposing corners." That's a quote from an op/ed in today's WP written by Ted Koppel. If only we could return to a time without social media. But think about it: who is Ted Koppel? He was the host of Nightline, the show that began in 1980 and obsessed about Iran and what became known as the "hostage crisis". Every night Koppel obsessed about the hostages, and the country learned to obsess along with him. We have been obsessing ever since. No, it's not social media, it's media generally and its influence on us, its obsessions that become that nation's obsessions. Koppel is unaware of his own contribution to our obsessions. Maybe it's just as well, for he ends his op/ed by quoting Edmund Burke: "The only thing necessary for the triumph of evil is for good men to do nothing." Stop obsessing.

Shouldn't the headline be: "A test of Marginal Revolution political bias?"

ha - maybe! though maybe that would have just proven the point...

"if you were more likely to add a ? for research that leans one political direction or the other, that would suggest a bias on your part."

so now the "bias experts" have determined that punctuation is a
valid metric for bias?
isn't this is more likely another pile of postmodern bowdoin1 bullshit
from the sociologists
stop the bias- avoid punctuation


On the matter of "?" shouldn't one consult The Rt Hon Jacob Rees-Mogg, Leader of the House of Commons?

"Whadya say, Jake old boy?" "Whaddevah!"


There was a bloc of years when the moderators' favorite quote mine among Republican 'economists' was... Bruce Bartlett. And, like every other econblogger in America, they treat Paul Krugman as if he were a decent and trustworthy human being. Professor-from-Bowdoin, you need to work smarter.

| "The distribution of slant across links is quite balanced, but leans left..." |

it cannot be "quite balanced" if it "leans left".

whatever TC may have been in the past, his output in recent years is indeed left of center....

Obviously this is a suspiciously perfect result and ..

lol, jk

Good work Tyler 👍👍👍

I'm going to post this here because it is related to the concept of centrism .. and I have to wait another couple hours on the west coast before I can start using loud power tools.

RIP, Rational Debate About the Federal Budget by Tyler Cowen

In a way this piece is right, and in a way it is terribly wrong. It is true that this was a bad budget, but not really for reasons of "polarization." It was a bad budget for reasons of attention. Neither side really cares, because they have bigger, and more real concerns. So Congress produced this budget on auto-pilot.

Democrats are wholly occupied with whether they should in fact impeach this president. And this president is wholly concerned with whether he's going to get impeached. But that sure as shit is not "polarization" and "both sides do it." Not after the Mueller report. It is not, as one cloying line states, because "Trump is considered too odious to be worthy of commendation on such a significant issue" like its' some taste or style issue.

And Dear God man, to drop "Trump will look like a genius, and the fiscal conservatives will continue their slide into irrelevance" this week?

Trump is busy trashing Baltimore and Rep. Cummings, and not talking the budget for the reason I gave up top. He's trying to squirm out of impeachment. He's trying to keep your eye off what Cummings' committee learned this week.

President Trump's Attacks on Baltimore Came as Elijah Cummings Quietly Built Effective Investigations Into the Administration

Don't let a commitment to "both sides" centrism blind you to a more moderate and pragmatic centrism, committed to the values and laws of the United States and its Constitution.

.. and that doesn't even touch on the fact that in so doing, Trump has made his racism overt, and any defenses of his strategy as well.

It's a hell of a week for "both sides do it."

"He's trying to squirm out of impeachment. " You think there's a change of impeachment after last week's sh*t show. LOL.
The chance went from 10% for the past year, to .1% Trump would likely welcome it as most of his tweets are about getting the Democrats to self identify as complete morons - for which they seem eager to do.

The chance of successful impeachment is low, but here's the deal:

I think that whenever a chief executive does obstruct justice to protect himself, impeachment charges must be brought, each and every time.

Because with out that, what even is a constitutional chief executive?

What if the underlying charge is bogus, the evidence clearly points in that direction, and yet the entirety of the federal apparatus spends two years investigating the POTUS only to declare there was no underlying conspiracy...

And then the opposition party says, sure there was no conspiracy, but obstruction of justice doesn’t require an actual crime !

We’re discussing whether a factually innocent president should be impeached for firing an FBI director for not curtailing an investigation into a factually innocent person.

I don't think anyone can read the Mueller report, or watch the testimony, and come to that conclusion.

The "excuses" offered, that someone thought they were innocent, or that they didn't get away with it, are recognized under the law. It is legally obstruction even so.

You or I certainly would not be given thoar special benefits or considerations.

Oops, important typo. The excuses are *not* recognized under the law.

And good news everyone, I didn't shoot myself with my nail gun. New toy. Neat.

(What you want obviously, is a President with the strength of character to go right down and testify to the FBI about what exactly happened, and how it *was* in keeping with his inauguration pledge "to be bound by oath or affirmation, to support this Constitution." On a bible no less.)

Are you saying that the fact that Baltimore is infested with rats is racist? http://www.animalplanet.com/wild-animals/3-baltimore-md/
What makes you say that? And if a fact is racist, what should be done?

Could have saved a lot of time just asking Tyler, "What are your political leanings these days? In what ways do you align with libertarian thought, in what ways you do not?"
He would probably answer in his own convoluted kind of way.

First of all: Dear God, that poor RA. Is there any doubt that the world has too many RAs, if their time is squandered like this? If you are comfortable with this setup, you have no soul.

Secondly, countless hours could have been saved by simply asking Tyler who he voted for in the last, say, five Presidential elections. I am very certain I know the answer. And if you had to put money on it, you know which way you would bet. That's all you need to know.

Someone that habitually pontificates on blog posts without bothering to present any actual data is commenting that RAs doing with data is a waste of time. Hahahaha. K buddy.

Dan - I like where your mind is going with trying to find out political leanings, but you have to realize that you can't identify "bias" this way.

A more productive project would do a similar analysis but compare posting of the same paper by different blogs.

This way you could situate Cowen relative to others -- and that's really the best you can do to answer the question you are asking.

Thx - not sure what your concern is re identification - the slant coding or identification assumption that the ?s are unconfounded? Agree latter is potentially issue but still thought this was worth checking out for pretty low cost mini-project. See also here, https://marginalrevolution.com/marginalrevolution/2019/07/a-test-of-marginal-revolution-political-bias.html#blog-comment-159966750, from quick look, added ? is not highly correlated w quality of outlet.

Anyway your idea of comparing blog posts is a good 1 but not sure if there are enough econ blogs posting lots of links for that to work. Not something I'm likely to pursue anyway. 1 advantage of my 'design' is it (tries) to get at bias 'within' a blog rather than just comparison across blogs which is a bit less interesting I think

I think they'd do better looking at a given issue, eg What Tyler posts on minimum wage.

Ah, but was there bias in the labeling of liberal and conservative viewpoints?

Wherein two academics living tell a culturally identical academic that he fits a "slightly left" norm using a qualitative and un-falsifiable methodology. If you wanted someone in your cultural group to affirm your sense of "rationality" and tell you that you were really very objective, you could achieve that very easily, but this little game these two played provides literally no information. Their (admittedly) unscientific way of coding your posts means that all conclusions reached can be dismissed unscientifically. Garbage in, garbage out. What did anyone in this situation hope to gain? MR should be about more than cultural signalling.

With respect to the VaderSentiment reference, the attributes of the L/N/R slant summaries as classified by Dan:

L: {'neg': 0.096, 'neu': 0.748, 'pos': 0.156, 'compound': 0.9872}
N: {'neg': 0.083, 'neu': 0.769, 'pos': 0.148, 'compound': 0.9953}
R: {'neg': 0.101, 'neu': 0.808, 'pos': 0.090, 'compound': -0.5093}

All: {'neg': 0.102, 'neu': 0.755, 'pos': 0.144, 'compound': 0.9973}

Excluding the neutral scores, so wrt to Dan and VaderSentiment, TC is more neg on the R and more pos on the left.

With VaderSentiment alone for all the summaries, TC is more pos. No reference for political leaning.

But what do sentiment analysis know about Straussianism??

Comments for this post are closed