Do you like your newsprint right-justified or left-justified?

My AFR column this month is on the economics of media bias. Full text over the fold.

Economics of Media Bias, Australian Financial Review, 17 May 2007

Two years ago, a team of economists from Yale University decided to answer a question that many a journalist has pondered in the pub: does the newspaper you read affect the way you vote? To test the theory, Alan Gerber, Dean Karlan and Daniel Bergan randomly chose one thousand households in the Washington DC area, and gave them a free subscription to either the left-leaning Washington Post or the conservative Washington Times. A few months later, they surveyed both groups – plus a control sample – to see how they voted. The finding? Those who received the Post were 8 percentage points more likely to vote for the Democratic Party. Getting the Times had no impact on how you voted.

In recent times, economists have expanded their research well beyond topics such as macroeconomics and labour markets that have traditionally constituted the bread-and-butter of the discipline. One of the subjects that the new economic imperialism has laid its tentacles upon has been media bias. What impact does the media you consume have on your political behaviour? And how and why do media outlets shape the news?

From a theoretical standpoint, one of the most important contributions has been a paper published last year by two young researchers at the University of Chicago, Matthew Gentzkow and Jesse Shapiro. Contrary to the simple explanation that media bias is driven by the personal predilections of proprietors or journalists, they argue that media slant emerges mainly as a result of outlets trying to tailor their news reporting to consumers’ prior beliefs. For example, people who buy a progressive broadsheet will be more likely to think that it provides accurate news if the newspaper gives a positive spin to Labor’s policy ideas. Similarly, those who listen to a conservative talkshow host will be more likely to tune in the next morning if the host praises the Prime Minister.

In a separate study, Gentzkow and Shapiro set about testing their theory, focusing on media bias in local newspapers. To estimate each newspaper’s bias, they develop a unique index. Searching the 2005 Congressional Record, they identify the phrases most commonly used by Democratic and Republican politicians. Among the top phrases used by Democrats are “change the rules”, “budget deficit”, “American workers” and “veterans’ health care”. Among the top phrases used by Republicans are “death tax”, “personal accounts”, “private property” and “stem cell”. Some of these reflect a concerted effort by political strategists, such as the campaign by Republicans to re-label estate taxes as death taxes, and to refer to personal social security accounts instead of private accounts.

How much of this subtle difference in language was picked up by the media? The duo then look to see which newspapers opt for the Democrats’ favourite phrases, and which outlets tend to use Republican phrases. When writing about tax reform, will the economics correspondent use the Republican phrase “tax relief”, or the Democratic phrase “tax cuts for the wealthy”? Will the international reporter talk about the “global war on terror”, or the “war in Iraq”? Using these subtle differences, they place all local US newspapers on a left-right spectrum. According to this index, the Tri-Valley Herald (circulating in the San Francisco Bay area) is the most left-wing newspaper, while the Houston Chronicle is the most right-wing.

Gentzkow and Shapiro then turn to explaining media slant. Consistent with their theory that media bias is mainly driven by customer tastes, they find that reader ideology explains significantly more of the variation in media bias than the identity of its owner. This finding holds up even when they take account of the possibility of reverse causality. Places with more churchgoers (a trait unlikely to be affected by newspaper bias) tend to have more right-wing newspapers. Conversely, cities with fewer churchgoers tend to have more left-wing newspapers.

So while Gerber, Karlan and Bergan have shown that left-wing newspapers make people more likely to vote for left-wing candidates; Gentzkow and Shapiro demonstrate that in areas with more left-wing people, newspapers are also likely to be more left-wing. These mutually reinforcing results raise the prospect of a vicious cycle, in which the media feeds and reinforces voters’ prejudices. How can we stop it?

Reassuringly to economists, the answer is a familiar one: more media competition. In the 2000 US election, places with a larger number of local television stations tended to give more equal airtime to the two presidential candidates. In the Middle East, countries with more competition between media outlets tend to provide more balanced reporting on politics. The more journalists are reporting on a topic, the less biased each can be.

In the Australian context, this finding has a straightforward implication: policymakers who want to reduce bias should focus on boosting the number of independently owned media outlets, rather than worrying too much about foreign ownership. All the news that’s fit to print doesn’t come from one source.

Dr Andrew Leigh is an economist in the Research School of Social Sciences at the Australian National University.

I didn’t have room in the op-ed to note that Gentkow, Karlan and Shapiro were all mentioned in January on the NYT’s list of 13 promising young economists to watch. Gentkow and Karlan have made recent visits to Australia – I’m still working on getting Shapiro out to do some talks.

This entry was posted in Economics Generally. Bookmark the permalink.

21 Responses to Do you like your newsprint right-justified or left-justified?

  1. Slim says:

    Just to be pedantic – in typographic terms ‘justified’ means each line is wordspaced so that it lines up against both the left and right margin. Otherwise text is center, left or right aligned. It’s now commonly misused, but there is no left or right-justified.

  2. David says:

    I’m not sure why more competition should lead to less bias (rather than, say, more extreme positions in an attempt to stand out from the crowd or more investment in understanding your potential readers’ prejudices so you can pander to them). Any thoughts?

  3. Ed says:

    I’d like to see more of the specifics of that DC-area study. Seeing as how the greater Washington area overwhelmingly votes Democratic (and, indeed, DC itself usually votes about 90-95% Democratic), I’m not really sure that newspaper readership has anything to do with it.

    The Washington Post’s circulation is 10 times that of the Times’ … in short, nobody reads the Times. But it’s not post hoc ergo propter hoc here … more people probably read the Post because of their political leanings, not the other way round.

    Or, more people probably read the Post because it’s a MUCH better newspaper. The Times is a lame excuse for a newspaper, and has virtually no credibility in Washington.

    BTW, the Post isn’t terribly “left-leaning” anyway.

  4. David says:

    Ed, it sounds like a pretty solid study. The people who read the Post in this study didn’t read the Post because they were left-leaning: they read it because they were randomly assigned to a group that was given a free subscription to it. Another group was given the Times and another group was given neither. Under those conditions it seems pretty safe to infer causality – it’s not just a correlation study, it’s an experimental one.

    Anyway, Andrew’s provided a link to the paper so check it out for yourself.

  5. ChrisPer says:

    Great line of inquiry!

    Media bias is not a simple thing to nail. This work is wonderful because its measurable and reproducible, compared with previously reported work which relies on a more subjective assessment of the articles.

    The problem is, its hard to assess a ‘framing’ which is unjust. Phrasing is probably a good proxy for that, but you need to identify the source clearly on a social grouping spectrum, and a simple left-right divide is only a first approximation of something more diverse and interesting.

    For instance, the whole ‘culture war’ iidea framework is infused with ideas from the left, while there is a large liberal centre that might qualify as ‘latte-set’ but are committed to intellectual integrity. They are badly undermined by more activist ways of thinking that do not admit of contradiction.

    The media bias in that framework vastly overplays the sensational framing of the activists because conflict and blood sell so easily. And the errors can be corrected on page 43 under an airconditioning advertisement, whicle people take bad political decisions and even die as a result of the frontpage sensationalism.

    The use of left-right is damaging; the more important questions are better solved without validating lunatic viewpoints. Even the framing of ‘left-right’ is problematic, because a centre-left gets labelled right-wing and the whole story re-aligns as though the rational moderate is some kind of fascist.

  6. Andrew Leigh says:

    David, the G&S model has competition reducing bias because it provides more signals of the truth. Their JPE paper has some nice examples of how this works, especially in Middle Eastern countries that move from zero competition to some competition.

    Slim, left-justified and right-justified aren’t oxymorons, though I admit that they are a tad archaic. See here, for example:

    vb. To justify, as text, along the left. See also justify (definition 1), rag. Compare right-justify.

  7. Peter Brent says:

    Nice piece, Andrew.

    Regarding the three whiz-kids’ paper. Without reading the whole thing, is it a fair to note that if each group numbered about a third of 1,000, ie 333 on average, that gives a rather large error margin?. A little over 5 percent, perhaps?

    So their 8 percent difference becomes what you guys call “insignificant”?

    In any event, the findings of the second study, of a kind of a self-sustaining, seems to make more sense than the first.


  8. Andrew Leigh says:

    Peter, thanks for the comment. The GKB result is significant at around the 5-10% level, depending on which controls they include (click on the link to their paper – and see Table 4).

    BTW, did your comment post prematurely?

  9. Panadawn says:

    David, more competition leads to less biased reporting because outlets need to cover a wider range of views in order to expand their customer base. With more outlets competing for the same amount of viewers, there is less discretion.

    Although I do wonder if this is true because it does seem some progressive voters intentionally read conservative-leaning rags and watch Bill O’Oreilly because they enjoy getting angry? I mean, Bill O’Reilly is incredibly polarising, yet he has the highest rating cable news show in the USA.

    Andrew, I am curious if the size of the customer base has any effect? I mean wire services like Reuters need to cater for media outlets world-wide, so is a Reuters peice less likely to be biased than a local one? Or do media outlets tend to cherry-pick which pieces they buy from Reuters based on which Reuters correspondent produced it?

  10. ChrisPer says:

    I think the idea of competition expanding the range of opinion has a major flaw. If the entire class of journalists is drawn from the uni-educated ‘new class’ (Katharine Betts ‘The Great Debate’) there is no pressure to balance opinions that are received values of that class.

    It is evident that until Fox News in the US the 50% conservative opinion was seriously under-represented in major news markets, and the (fairly subjective) IPA news bias studies around elections in Australia found private and public broadcast journalists were generally all biased the same way.

    There is plenty of media competition in Australia, but little recognition that framing of stories is terribly uniform. Look for instance at how the Wolfowitz story is being played – reflecting the gloating of the anti-neocons worldwide and with no indication that the situation was basically set up by political enemies, or that Wolfowitz alerted the board to the problem and tried to recuse himself BEFORE HE EVEN TOOK THE JOB. It says a lot about Australia’s media.

    It is IMHO not a matter of deliberate bias, but a convergence of trends that produces an inability to present a story with depth and balance. The international services are particularly bad.

  11. Andrew Leigh says:

    ChrisPer, I also thought that Wolfowitz was getting a bad rap from the media – but was persuaded otherwise after talking to an insider who’s been following it very closely. It was a mistake to ask him to take care of her new job, but he should’ve been smart enough to refuse to have anything to do with it.

  12. Jacques Chester says:

    David, more competition leads to less biased reporting because outlets need to cover a wider range of views in order to expand their customer base. With more outlets competing for the same amount of viewers, there is less discretion.

    This reminds me of median voter theory.

  13. Sacha says:

    “David, more competition leads to less biased reporting because outlets need to cover a wider range of views in order to expand their customer base. With more outlets competing for the same amount of viewers, there is less discretion.”

    How is this affected by the observation that it appears as if, in competitive marketplaces, there are often outlets that appear to various worldviews – eg my guess is that fox news appeals to people with a certain world view, the different newspapers in Australia appeal to different world-view segments (think of the SMH and the Australian) ? Just throwing the idea out there.

  14. Dan says:

    These papers are very interesting. I only had a chance to have a very quick look so far, but none of them seem to mention the point that newspapers make almost all of their money from advertising rather than sales. Does this affect the conclusions? Are there any studies on this? I know that the idea has been much talked about, but I don’t know of anything quantitative.

  15. Panadawn says:

    Thats actually a pretty good point Dan, one article in the SMH today made the point that The Oz panders to business interests because thats where the advertising dollars are, as a result The Australian website is the most profitable news website in Australia, according the head honcho at The Oz himself.

  16. Andrew Leigh says:

    Dan, it’s a topic that GS look at, though I didn’t have time to mention it. And there’s a bit of a literature on this (here’s one of my faves). The simple answer is that the effects are there, but they’re small.

  17. Pingback: Andrew Leigh » Blog Archive » Half full or half empty?

  18. ChrisPer says:

    Thanks Andrew, but my Wolfowitz comment was not to claim Wolfowitz was misjudged, but that the story was almost uniformly framed to present the worst possible picture – in OUR media.

    (Stupidity appears more harshly punished than cupidity, but thats just the way it is.)

  19. Dan says:

    Andrew – thanks for the reference, I’ll take a look at that.

  20. Richard Innes says:

    Prof. Leigh,

    I apologize for using your blog for essentially an e-mail, but I would like to get a copy of your report on the impact of the best teachers in Australia. It it available electronically?

    Richard Innes
    Education Analyst
    Bluegrass Institute for Public Policy Solutions (Kentucky, USA)

  21. Pingback: Tim Worstall

Comments are closed.