I definitely have to read this Groseclose book

Yesterday we discussed Tim Groseclose’s Fox TV interview in which there seemed to be some confusion about who gets to be a tax-exempt 501( c)(3) group.

This got me curious about the content of Groseclose’s book. Here’s what I found:

Media bias aids Democratic candidates by about 8 to 10 percentage points in a typical election. I [Groseclose] find, for instance, that if media bias didn’t exist, John McCain would have defeated Barack Obama 56 percent to 42 percent, instead of losing 53-46.


The above quote is from a columnist named Paul Bedard. I’ve never heard of him before—I found the link through Google—but given that, according to the source, “the Drudge Report is approximately the most fair, balanced, and centrist news outlet in the United States,” I’d guess that Bedard is probably slightly to the left of center.

I was amused to see that, in the quote above, the word “John McCain” was given a hyperlink to a page with this photo:

The following profile begins, “John McCain, a two-time presidential candidate and the senior senator from Arizona, was born in the Panama Canal Zone to a well-known military family. Both his father and grandfather were decorated admirals.”

In contrast, Barack Obama received no link, but appended to the quote was a hyperlink to “[See editorial cartoons about Barack Obama.]”, a page that begins with this clever bit of political satire:


This seems like liberal bias to me. I’d much rather be portrayed shooting baskets than posing with a tie.

OK, let’s try to apply some statistical methodology here. As we know, the best way to measure the bias of any news organization is to count who they cite. Here’s a quick check of the sources cited by Bedard:

Drudge Report
Fox News
National Public Radio
New York Times
Washington Times

Now let’s code these. I think the most objective approach is to use Groseclose’s estimates. This gives us the following:

left: ABC, CBS, CNN, NBC, National Public Radio, New York Times
right: Fox News, Washington Times
exact center: Drudge Report

And I don’t think anyone will object if I place UCLA on the left.

Bedard mentioned 7 sources on the left, 1 on the right, and 1 in the exact center. Seems pretty liberal to me. Now I don’t know whether to trust him at all!

In all seriousness, I really do want to see the book now to see how the concept of “absolute bias” is defined. As I wrote a few years ago, I thought the Groseclose and Milyo paper was interesting but I couldn’t see how they could claim to get any absolute, rather than relative, measures.

8 Responses to I definitely have to read this Groseclose book

  1. Jon July 22, 2011 at 1:40 pm #

    Hi Andrew, you can read actual excerpts from the book here:


  2. Andrew Gelman July 22, 2011 at 2:12 pm #


    Thanks for the pointer. Unfortunately these excerpts have stories but nothing on the methods or estimates. The guy’s a lively writer. I can see why he turned down offers from Yale and Chicago. LA has much better weather and much better food.

    It was a bold move for Groseclose to write this book because I think it will be hard for a lot of people to take his work seriously now. I admire that, rather than just sitting in a comfortable academic job, he wants to make a difference in the world and be more of a political activist.

  3. Doctor with only a thermometer July 22, 2011 at 2:26 pm #

    . . .and I’m not at all interested in reading it. As a scholar of bias (whose work even acknowledges that I’m not the first person to ever think about these things), I can’t imagine I’m in his target audience. I’ll hold out until I see some credible reviews; but from all the hype, and the excerpts on PowerLine, it sounds like a typical “how” book, of the sort that pollute the “current affairs” shelves at Barnes and Noble: “[short catchy title][colon][How (bad person or group) (does something bad to America)].”

    I guess I’m sort of interested in how he determines that McCain would have won in a landslide with a balanced press. I agree with the most common critiques of his bias measure: while innovative and sophisticated, it simply doesn’t measure absolute bias in any remotely content-valid regard. So I’m wondering if his electoral findings involve taking similar liberties with our measures of how elections work. And if so, will election scholars — who tend to have more visibility than media scholars within political science — challenge it in any serious way.

  4. Justin H. Gross July 22, 2011 at 3:20 pm #

    I’m reading the book now. I, too, was interested in seeing how “absolute” bias might be measured; seems like quite a trick. In a nutshell, he assigns legislators PQ scores based on their ADA ratings, projects think tanks onto this same scale based on how frequently the legislators cite them, then projects media sources onto the scale based on how frequently they mention the think tanks.

    The “center” is defined as the mean of all legislators’ PQ (adjusted ADA) scores in a given year. Also, the center of public opinion among voters is calculated by assuming legislators to represent the ideological center of their district and weighting voters in inverse proportion to voting power (to undo the influence of unequal representation) and adding a phantom Washington, D.C. district that is assumed to be the most liberal. Groseclose uses NOMINATE scores as the gold standard, saying “NOMINATE is the most impressive research achievement ever in political science,” and takes comfort in the fact that his PQ scores correlate highly with those from NOMINATE.

    I’m especially curious to understand the basis for his causal argument that links media bias to specific changes in citizen ideology, measured on a point scale. It seems that his counterfactual that McCain would have won if it weren’t for leftward media bias and, moreover, that the Washington Times and Fox News with Brit Hume would appear slightly left of center if it weren’t for media bias turning the average voter from a Ben Stein into an Arlen Specter (the GOP version).

  5. Andrew Gelman July 22, 2011 at 8:59 pm #


    So the story seems to go as follows. The center is defined as the center of Congress as of 2004 or so (or whenever the analysis was done). The median voter is near the median congressmember, which makes sense. But most media outlets are about 20 points to the left (on a 0-100 scale) of the median voter. If Congress is at 50 and the median voter is at 50, the media are at 70.

    The assumption is, if the media were at 50, then the median voter would be at 30 and John McCain would be president. But if the median voter were at 30, then the median congressmember would eventually be at 30 also. But then the media, sitting at 50, would be biased again! So the media, to become unbiased, would have to go from 50 to 30, which in turn would allow the voters to move to 10, but then the media would be biased again, . . . .

    Did I get that right?

  6. frankcross July 22, 2011 at 10:06 pm #

    Very good Andrew. But the weakest part of the case is on the effect of media bias that may exist. He relies on very remote experimental evidence with little parallel to this context and, on top of that, uses some pretty questionable assumptions to extrapolate it.

    The effect of media on votes has been directly studied (i.e., when Fox enters a market) and the effect is much smaller than Groseclose estimates.

  7. anonymous coward July 24, 2011 at 9:20 am #

    Also, the center of public opinion among voters is calculated by assuming legislators to represent the ideological center of their district

    Thank goodness that assumption isn’t at all dodgy and there’s no evidence at all that MCs are more extreme than their districts. Obviously far better to estimate public opinion that way than to use, say, public opinion data.

  8. crstephens August 15, 2011 at 11:20 am #

    Readers might be interested in this critique of the methods that Groseclose used:
    And to people how are touting this book as “proof” of a liberal media bias I also refer them to this: