David R. Henderson  

My Review of Groseclose's Left Turn

PRINT
Confidently Uncertain... Are Political Labels Uniquely ...
Virtually all of us who identify ourselves as libertarians or conservatives (I'm the former) have believed, for as long as we have been paying attention, that the mainstream media, whether print or electronic, have a left-wing bias. The late columnist Edith Efron, in her 1971 book The News Twisters, documented that bias among the three major television networks of the time--ABC, CBS, and NBC. Now, University of California, Los Angeles political scientist Tim Groseclose has actually measured the bias, not just of the three traditional networks, but also their present-day network competitors and major newspapers.

Most of his findings will probably not surprise most readers of this publication. Groseclose concludes that, indeed, the mainstream media do tilt left. Why then do I review a book that tells us what we already "know"? There are four reasons: First, most of us don't know it to the extent Groseclose knows it--his argument is an empirical tour de force. Second, he is so numerate that he makes clear with the data just how extreme the left-wing bias is. Third, there are some surprises in the data, particularly about the Fox News Channel and the Wall Street Journal. Fourth and finally, Groseclose shows that the biased information people get causes them to vote to the left of their true positions.


This is from my review of Tim Groseclose's Left Turn: How Liberal Media Bias Distorts the American Mind. The review appears in the current issue of Regulation. Some further excerpts:
The bias shouldn't be surprising given the political views of reporters. Surveys show that Washington correspondents vote for the Democratic candidate at a rate of 85 percent or more, Groseclose notes. Studies of contributions to presidential campaigns have found that more than 90 percent, and as many as 98.9 percent, of journalists who contribute to a presidential campaign give to the Democratic candidate. These overwhelming numbers mean, Groseclose says, that residents of left-wing academic communities like Cambridge, Mass. and Berkeley, Calif. are, on average, much more conservative than Washington media correspondents.

Groseclose examines a few issues to show the bias at work. The first item he discusses is a Los Angeles Times article on the number of black students at UCLA. Groseclose dissects the story to show that the reporter, Rebecca Trounson, presents the data and reports interviews in a biased way. For instance, to buttress her case that the UCLA admissions process discriminates against black people, she cites six people, five of whom are on the political left, and only one of whom is conservative. Moreover, she pulls a favorite trick of left-wing reporters: identifying the ideology only of the conservative. Trounson's L.A. Times colleague, Ralph Vartabedian pulled the same trick on me--although, unlike Vartabedian, Trounson at least got the ideology right. (Vartabedian described me as a conservative. See my August 18, 2010 blog post, "Media Bias and the L.A. Times" for more.)

I should note that the UCLA admissions process is racist. As Groseclose notes, UCLA discriminates, probably illegally, in favor of black applicants. One problem he identifies with Trounson's approach is that she missed the big story: the rising percentage of Asians at UCLA and the falling percentage of whites.


And finally:
In an interview with the Hoover Institution's Peter Robinson, Groseclose explained it another way: Currently, the average U.S. voter has the same political quotient as the average Iowa voter. But with no media bias, the average U.S. voter would, instead, be like the average voter in Kentucky or Texas.

I do have one criticism, early in the review, of Groseclose's misstating an important bill before Congress and a couple of criticisms at the end. But the bottom line is that Left Turn is an outstanding book.


Comments and Sharing





COMMENTS (29 to date)
Daniel Kuehn writes:

Obviously I'd have to read the book for this particular case, but two things regularly raise flags for me in these discussions:

1. False equivalences: for example, the SQ is calculated using think tanks. Certainly some think tanks lean certain ways, but that shouldn't be the only factor in assessing why a news organization cites a particular think tank. There is a dramatic variation in the quality of think tanks. I care a lot less about how many liberal and how many conservative think tanks got cited, and a lot more about the quality of the think tanks that got cited. That is what we care about when we talk about bias, right - objectivity and rigor? Does he get into that? If not, I think that's not quite getting at the right issue when we think about "bias".

2. The politicization of truth: I guess this is kind of related to the first, now that I think about it. When people talk about "bias in the media", they often refer to a balance of political views. To me, that can be a very misleading way of talking about "bias". If the question is "is free trade welfare-enhancing?", my understanding of what the "unbiased" answer would be is to universally answer in the affirmative. I would not consider it "unbiased" for there to be a difference of opinion on this point. If the question is "is climate change man-made and a threat", I would consider the "unbiased" answer to be universally answering in the affirmative with some clear statements about (1.) uncertainty in forecasting, and (2.) the adaptability of society to those changes. My suspicion is that a lot of people would consider these responses that I consider "unbiased" to be "biased", because they're doing what Groseclose is doing and identifying "politically balanced" with "unbiased".

It wouldn't surprise me at all if a fully comprehensive assessment of an "unbiased"/"objective" reporting posture would look left of center.

Perhaps we should consider the prospect that the American population is biased, and the media does a more or less decent job of being unbiased. I doubt that's perfectly true - I'm sure there's some bias in the media - but I think taking the political distribution of the American population as a benchmark is likely to dramatically overstate the degree of bias.

AngryKrugman writes:

As a former journalist, I get a bit of a chuckle out of the claims of media bias. You have a large group of people--some smart, many more not-so-smart, like any profession--trying to get out thousands and thousands of words in some coherent form everyday. They can barely accomplish that task, much less orchestrate some more sinister plan. Many of the institutional structures of the profession may lead to bias--the search for "objectivity", wanting to write a compelling story, deadline pressure, etc.--but most of it is unintended. Chuck Klosterman makes this point in Sex, Drugs, and Cocoa Puffs--the "media bias" readers often interpret actually often results from rather innocuous set of events, such as a reporter missing a phone call because he left his desk to go to the vending machine. It's the nature of the industry that the first person to call you back can shape a story simply because you're on deadline.

All that said, it doesn't necessarily invalidate the claim that, as an empirical matter, news coverage skews "liberal" (although I think it varies greatly based on the subject matter--that's not the case for national security reporting, for instance, where unidentified gov't officials often have opportunity to spout their views without much pushback). But it makes it much less invidious.

Ken B writes:

Invidious is the point is it?

I'm not sure I want to actually thank you David, as the last thing I need is another book to read, but I got the Kindle edition. I found a link to his site to get your own PQ

The PQ is, to answer Daniel, is based on issues selected by Americans for Democratic Action.

Daniel Kuehn writes:

Ken B - SQ is based on citation of think tanks somehow.

But this is a problem with PQ, right? How is your place on a political spectrum indicative at all of bias? Couldn't the objective position (or one's best stab at it) be firmly on one side of the political spectrum or the other?

Ken B writes:

@Daniel Kuehn: Well I am reluctant to discuss the book before reading it -- I gather you're a Keynesian? -- but meaningful statistical measures will exhibit certain consistency properties. One assumes there will be a discussion in the book.

My PQ came out quite low, 15.

Daniel Kuehn writes:

I have no doubt at all that the measures are consistent. The question is, what does it consistently measure? I'm afraid - from just reading the summary - that he's getting at whether the media reflects the political views of society and not whether the media is biased in its reporting.

Unless by "unbiased" all we mean is "reflects the views of the average American"... but that doesn't strike me as a particularly useful definition of "bias".

Daniel Kuehn writes:

I'm 67.7 - right about at Arlen Specter level. That's a long test! I'm not sure about its one dimensionality. In a typical Nolan test (which is nice for its multidimensionality), I pretty consistently fall at the adjoining corners of "centrist", "liberal", and "libertarian".

A lot of these questions have to do with stimulus-related legislation too, and I don't think my views on stimulus have very little to do with my politics. The only sense in which they have to do with my politics is insofar as I'm not inherently opposed to public spending for public purposes. It's not some kind of indication that I'm a big-spending liberal (which I generally don't think of myself as).

Daniel Kuehn writes:

*I don't think my views on stimulus have much to do with my politics.

To elaborate... a climate scientist who believed the consensus and was in all other respects a conservative would be predisposed to support climate legislation, I'm guessing. An economist who believed the consensus and in all other respects wasn't especially liberal would be predisposed to support stimulus legislation.

Daniel Kuehn writes:

Sorry to be commenting so much, but this genuinely interests me on many dimensions. I'm not trying to be cantankerous - I am sincerely intrigued by all this.

One of the dimensions on which it interests me is this think tank score as another method of getting at the SQ. I've been reviewing the QJE article that discusses that score in detail, and I don't buy it at all.

The Urban Institute gets the same "average score of legislators who cite it" as PETA, for example, and Cato and AEI are both closer to a score of 50 than the Urban Institute. I don't buy it. Both of those organizations explicitly talk about an ideological mission, with Cato going as far as calling itself "libertarian". I've never seen any ideological priorities sway output from The Urban Institute or change any of the results produced.

The reason why more liberal Congressmen might cite it are: (1.) they research social and economic policies liberal Congressmen like to talk about, and (2.) maybe liberal Congressmen are more evidence based in their assessment of the issues, at least when it comes to these issues.

This gets back to the whole problem I noted above. Being the political median has nothing whatsoever to do with "bias". Nothing. We can argue over what constitutes an objective, unbiased position, but one thing that doesn't constitute an objective, unbiased position is scoring a 50 on the PQ scale.

But that's kind of where this fails the smell test for me... if PETA and the Urban Institute are on par and somehow Cato is less biased than the Urban Institute. Sorry for Cato fans. They do excellent work. But they do work that is explicitly informed by an ideological perspective, which is not what I would call "unbiased".

Daniel Kuehn writes:

Groseclose says as much about the definition of "bias" on page 1204 of the QJE and I am aware he says that. I don't see how this definition of "bias" matters all that much.

Vipul Naik writes:

Agree with the general thrust of Daniel Kuehn's first comment. I don't see any reason to prima facie privilege the political beliefs of the median US voter.

Experts do differ systematically from ordinary people in their views on a number of issues. Whether this difference stems from better information or from bias is an empirical question. Your co-blogger Bryan Caplan wrote a whole book demonstrating that the median voter holds systematically biased beliefs on economic matters. Voter beliefs on matters such as toxicology and evolution also systematically differ from those of experts. Experts who address the general public (like economists blogging) are, in a very limited form, influencing the general public to move closer in the expert direction. Pace Groseclose, would you suggest that economists start catering more to the median voter's beliefs about economic matters?

The real question is whether the systemic differences between journalists and the public they seek to inform are due to some expertise possessed by the journalists, or "bias" of some sort stemming from plain ignorance or lack of understanding of opposing perspectives. Since the questions at hand are essentially contestable, one way of trying to measure this would be to figure out how clearly the journalists can explain the ideas of the "other side" -- do they represent or portray such ideas accurately, in a manner that the adherents of the ideas would recognize? One way of tackling the question might be using what your co-blogger Bryan Caplan calls an ideological Turing test.

c141nav writes:

I used to think that I was a conservitive. I was born in Arizona and came of age politically during Barry Goldwater's presidential bid. Later I thought that I was libertarian. Finally I learned that I am a classical liberal.

Hume writes:

Daniel,

I suggest the work of Dan Kahan on culturally-motivated cognition (as well as the Cultura Cognition Project at Yale Law School).

http://www.culturalcognition.net/kahan/

Kahan claims that how one actually perceives "facts" is importantly influenced by one's values and world views. For example, one of his studies found the following: "Subjects of opposing cultural outlooks who were assigned to the same experimental condition (and thus had the same belief about the nature of the protest) disagreed sharply on key “facts”—including whether the protestors obstructed and threatened pedestrians." [Kahan, et al., "'They Saw a Protest': Cognitive Illiberalism and the Speech-Conduct Distinction" (2012)].

From the abstract. Read the whole thing:

http://www.culturalcognition.net/browse-papers/they-saw-a-protest-cognitive-illiberalism-and-the-speech-con.html

Costard writes:

Daniel - "Perhaps we should consider the prospect that the American population is biased."

Then we should also consider the prospect that the community of experts is biased. You cannot dismiss the broad consensus without questioning the lesser one, and if it is faith to say that the public knows best, then it is also faith to trust in an expertise you do not personally understand. But there is a more relevant point to be made here. The experts just as well as the public rely upon the media for their perspective upon the world, and even upon their own profession. Consensus and paradigm might be the result of a well-proven argument. But they can also be a symptom of the self-reinforcement that occurs when communication channels - whether scientific journals or the television networks - aren't neutral.

Vipul - "I don't see any reason to prima facie privilege the political beliefs of the median US voter."

Then is there any reason - prima facie - to privilege his vote? If an unrepresentative media is a good thing, then perhaps an unrepresentative government is also a good thing.

Hume writes:

"Then is there any reason - prima facie - to privilege his vote? If an unrepresentative media is a good thing, then perhaps an unrepresentative government is also a good thing."

This does not work, unless you want to say that the only value associated with a form of governance is instrumental value, and the only instrumental value of democratic representative governance is epistemic value. I find this line of thought implausible (as I find Bryan's normative suggestions in The Myth implausible as well).

Ken B writes:

Vipul Naik: "I don't see any reason to prima facie privilege the political beliefs of the median US voter."

'Privilege'? Well, let that pass. There certainly does seem to be some sense in defining bias in relative terms. I'm not sure what absolute bias would look like. Can you define it for me?

He also gives an interesting argument on why the median voter is important in discussions of bias: he is the elusive quarry of policy advocates of all stripes, the man or woman whom all want to influence. So it's not just prima facie.

Vipul Naik writes:

Costard:

Vipul - "I don't see any reason to prima facie privilege the political beliefs of the median US voter."

Then is there any reason - prima facie - to privilege his vote? If an unrepresentative media is a good thing, then perhaps an unrepresentative government is also a good thing.

Perhaps you're right, but your analogy is weak.

Teaching students a subject:Test taking::Media reporting on political issues:Voting

You may strongly believe that, at test time, students should take the tests on their own without help or interference from teachers. This does not mean that when teaching a subject to students, teachers should avoid trying to deal with or eliminate student misconceptions, and should stick to telling students only what they want to hear.

In the same way, one may believe that voters should, at voting time, cast their own votes and majoritarian systems should be used to determine the winning politicians or policies. This does not imply that the media, whose job is to inform people's political knowledge and opinions, should reflect people's prior political bent.

Vipul Naik writes:

Ken B:

Vipul Naik: "I don't see any reason to prima facie privilege the political beliefs of the median US voter."

'Privilege'? Well, let that pass. There certainly does seem to be some sense in defining bias in relative terms. I'm not sure what absolute bias would look like. Can you define it for me?

By "prima facie privilege" I simply meant the presumption that, pending evidence to the contrary, the median voter's belief is to be taken as the standard of reference for measuring bias. I question this presumption.

I cannot define "absolute bias" because of the essential contestability of many of the underlying values being debated -- but I offered some alternatives in the final paragraph of the original comment you quote from.

He also gives an interesting argument on why the median voter is important in discussions of bias: he is the elusive quarry of policy advocates of all stripes, the man or woman whom all want to influence. So it's not just prima facie.

Thanks for bringing this up. I hadn't been aware that he had made this argument.

Nonetheless, the argument doesn't do much to change my view. First, it may be important to *know* what the median voter thinks, but that doesn't mean it is necessary to treat the views of the median voter as a standard of reference. Again, I don't see how this is different from median skills at mathematics or median economic or historical literacy.

Second, I don't even think that influencing the median voter is the target of influence of policy advocates. At least, not directly, considering that most policy issues aren't decided by referendum, and there are many stages to making a view percolate to the general public.

Third, the fact that the media don't *express* the views of the median voter doesn't mean they aren't *targeting* the median voter. If your goal is to push the median voter in your direction, you cannot do this by saying only the things the median voter already believes in.

Daniel Kuehn writes:

Costard -
re: "Then we should also consider the prospect that the community of experts is biased."

Certainly. I thought we were all in agreement on considering that one.

re: "Then is there any reason - prima facie - to privilege his vote?"

Of course! A free society shouldn't be dictated to by a minority. Reality has no such strictures - its plausible for a minority to rule on that.

Ken B writes:

@Vipul Naik:
You seem to be conflating biased or slanted with incorrect. Groseclose does not make this identification. Again, I don't see how you could logically do so without some absolute, not relative, measure of bias. His claim is that the media consistently presents a 'left' perspective on most issues. That is a meaningful 'positive' claim because 'left' and 'right' are relative terms. His claim is to have found meaningful statistics to quantify and demonstrate the effect.

Consider this assertion: In most cities in the nothern hemisphere housing prices are higher in the western (left) part than in the eastern (right) part. That is a meaningful assertion. It is a positive assertion not a normative one. Eastern and western are relative terms. One can imagine statisitical tests of this assertion. Imagine i present a study claiming just that. You wouldn't make any of your objections listed above to this study I think.

Daniel Kuehn writes:

Ken B -
Bias means slanted or incorrect or off of the right position in some way. It means that there is something tugging you in a different direction. That's what's so strange about the idea that this book demonstrates the media has a liberal bias.

It makes sense to say "the media is more liberal than the Congress", it doesn't make much sense to say the media has a liberal bias, based on this evidence. That's true regardless, but given how people read "liberal bias in the media" it's even more important to recognize in this case.

Ken B writes:
Bias means slanted or incorrect or off of the right position in some way. It means that there is something tugging you in a different direction. That's what's so strange about the idea that this book demonstrates the media has a liberal bias.

Are you arguing that a liberal bias is impossible because the liberal approach is the correct one, and so there could be a conservative bias but not a liberal one?

Here is what Groseclose is actually arguing (I have now read enough of the book to say this): SQ and PQ are sound statistical measures and SQ is a strongly biased estimator of societal PQ.

(I am using math speak.)

He also argues that this has an effect on policy and election results. he does not claim that a PQ of 50 (the recalibrarted median) is 'correct'. These are positive not normative claims.

Ken B writes:

While we're discusing what counts as bias, and I am annoying people with statistical pedantry, let me criticize David R Henderson over a case where I think he mixed up definitions.

In his book Joy of Freedom (which I recommend) he discusses a question on a scholastic aptitude test. The question was dropped from the test because all the inner city black kids got it right, and few others did. David made a remark like 'Remember this next time you hear such tests are not biased against blacks.'

This reflects a confusion on what bias in a predictive test is. Say the goal of a test is to predict some future measure of acadamic performance. The test is biased against black kids if they consistently outperform its prediction, and biased for them if they consistently underperform its prediction. It is entirely possible the bias of the test could be reduced by dropping that question.

I think David was mixing up this kind of bias with cases where someone like David Duke would drop a question and then use the results not as a predictor but as a conclusion.

Ken B writes:

I read a lot of history, and I hate books that use history to grind axes. After tossing aside just one too many irksome books I developed a simple predictive test. I looked in the index for Reagan, Thatcher, Clinton, and Bush v Gore. If I found any I passed on the book. I found at least one of these surprisingly often. I am sure my bias predictor is biased against left-wing bias, with three flash points rather than one, but it seemed to work pretty well.

Daniel Kuehn writes:

Ken B -
All I'm saying is that one could also say "SQ and PQ are sound statistical measures and societal PQ is a strongly biased estimator of SQ."

Why anyone would assume PQ can even be talked about as in anyway grounded is beyond me.

And if Groseclose is just talking about bias in a statistical sense - with reference to an estimator, that is - he settled on a strange subtitle.

re: "He also argues that this has an effect on policy and election results. he does not claim that a PQ of 50 (the recalibrarted median) is 'correct'. These are positive not normative claims."

What I'm wondering is why I shouldn't believe that politics and policy are pulling the media more conservatively than it otherwise would be. I'm no claiming a PQ of 70 is "right". It's a positive question I have, not a normative one.

re: "While we're discusing what counts as bias, and I am annoying people with statistical pedantry"

I'm not concerned about "math speak" as you call it. What I'm a little concerned about is that you seem to be selectively applying it, and I'm not sure I'm understanding why.

Daniel Kuehn writes:

So the last line of that last comment sounded a little cattier than I intended it.

But I really think there is a problem with your logic here. He seems to have fine measures here (the SQ bothers me from my personal experience with many of those think tanks, but whatever - its not a think tank evaluation tool, it's a media characterization tool and it probably does pretty well at that).

My concern is entirely interpretation.

I have no idea why he titled it "left turn" rather than "right turn". I guess I just don't know how he parsed that one out. If you get to that point in the book perhaps you can share.

Identification problems are critical to nail down. Otherwise you're just a pundit, not a social scientist. I'm scratching my head over how he nails this one down.

Ken B writes:

DK:
"Why anyone would assume PQ can even be talked about as in anyway grounded is beyond me."

The distribution defines the median, no? Would you object to measures that try to quntify other sqishy stuff, like religiosity or social levels of trust? There are issues (is it interval data, etc) but surely there is somethiong there to be measured.

In the latter part of the book, where I am now, he quantifies the effect on the populace of the bias, ie measures how media SQ moves population PQ. Hence Left Turn.
I am less confident of his stuff here, as it looks like you need interval data for that, but I haven't checked the formalism (and won't).

My comment about math speak was for those who don't know that biased estimator etc are precisely defined terms.

Ken B writes:

Havingow read it...
A good book overall, with an interesting discussion of using stats in the social sciences. Unlike some here I think that on some contentious issues people do disagree in a way that can be described as right vs left. On these issues they get information from the media. Again unlike some here I think that sources of information can be biased, and in a way similarly describable as right vs left. Groseclose has a defensible measure of both of these. He makes a convincing case for what I think is obvious: the media strongly favors one side ('left').

He is on shakier ground when he measures the influence of the media and claims it has a strong effect. However he does make a case it has some effect.

Ken B writes:

Daniel Kuehn: "And if Groseclose is just talking about bias in a statistical sense - with reference to an estimator, that is - he settled on a strange subtitle."

The subtitle comes from his conclusion that the bias matters.

Let me sketch the argument.
-- begin sketch

There is a range of opinion on a subset of particularly contentious issues which can be characterized as R-L difference. It is possible to measure this polarization. PQ.

News reports can be seen to have a similar R-L polarization. (I assume not even DK denies this). It is possible to measure this. SQ, corresponding in some strong sense to the PQ of the media authors.

The measures show the media is sharply to the L of the populace. That is SQ, correlated to media PQ, is L of population PQ.

There is a media effect: bias in media reports affect population PQ. Thus the persistent difference between the media PQ as estimated by the SQ pulls the population PQ to the L of where it would otherwise be. This is a strong effect.

-- end sketch

That is the skeleton of the argument. The individual claims are all argued for in the book.


More prosaically: reporters are more likely to be liberal democrats, this biases their coverage (mostly by affecting what they choose to report), that coverage drags the populace left of where they would otherwise be, boosting the democrats. Can we come up with robust stats to measure all this stuff? Yes we can!


Comments for this entry have been closed
Return to top