VoxEU Column Politics and economics

Does the mass-media have political influence?

High quality empirical evidence from the shows that mass media influences voters but it is not clear that the media imparts a bias. It could be that improving access to any media informs voters and prompts them to turn against an embattled incumbent.

During the French presidential elections, presidential hopeful François Bayrou severely denounced the « media kill » that the mass-media and his opponents had supposedly orchestrated against him. And indeed, the French media watchdog (CSA1) had revealed an « excessive bipolarisation in favour of two candidates» on « certain (television) channels» in December (Libération, 10 January 2007). Bayrou was proposing, if he was elected, to introduce several new laws to guarantee media independence, with particular regard to controlling the grip of large industrial groups on the mass media. However, in speeches where he denounced the mass-media’s power, Bayrou never failed to praise the voters, who, he said, don’t let themselves get caught up in it («But we French, we are civic people: we don’t give into your onslaught»). Indeed, we would expect that: confronted with systematically biased messages, voters should be able to redress the balance mentally.

Who is right? Bayrou the pessimist who accuses the mass media of manipulating elections? Or Bayrou the optimist who points out the futility of their efforts?

Two recent studies help to clarify this question. The first, from Stefano Della Vigna and Ethan Kaplan2, studies the impact of Fox New, Rupert Murdoch’s ultraconservative cable television channel. Fox news was launched in the USA in 1996. In 2000, it was available to 20% of American homes. The study’s authors take advantage of the fact that not all cities had access to Fox News at the same time for technical reasons having nothing to do with their political orientation. They compare the changes in the proportion of Republican votes between the presidential elections of 1996 and 2000 in the cities receiving Fox News before 2000, and the cities that hadn’t received it yet. They conclude that Fox News caused an increase in the Republican vote of about a half percentage point, or 200,000 votes, enough to change the result of the 2000 presidential election.

What about the printed press? A group of Yale3 researchers (Daniel Bergan, Alan Gerber and Dean Karlan) conducted an interesting experiment to find out. Three weeks before the 2005 gubernatorial election, the researchers contacted about 1,800 people living in the Washington area, and offered half of them (chosen at random) a free subscription to a daily newspaper. Half of those lucky chosen ones (also chosen at random) received a subscription to the Washington Post, a center-left newspaper (favouring the Democratic candidate). The other half received a subscription to the Washington Times, a newspaper known for its conservative opinions (favouring the Republican candidate). The authors were then able to obtain electoral data to study the effects of the newspapers on participation in the election (the vote itself is anonymous, but not whether one has voted), and they carried out a complementary telephone survey. It was not a good time for President Bush and the Republicans. The bad news of the war in Iraq, in particular, dominated both newspapers. Nevertheless, the researchers showed that the Washington Times had the tendency to present bad news in a more favourable light than the Washington Post. It could therefore be expected that the two newspapers had a different effect.

Receiving a free subscription to one or the other of these newspapers upped the likelihood of voting in elections a year later by about 3.5% (the average rate of participation is 65%). According to the opinion survey, voters who had received the Post were 11% more likely to vote Democratic than the group that hadn’t received any newspaper subscription. But strikingly, even the voters who had received the Right newspaper (the Times) were even more likely than the control group (7%) to vote Democratic. And the political opinions of the two groups, in both cases, shifted towards the left (in comparison to the control group). This study therefore concluded that readers don’t let themselves be influenced by the given stance of a newspaper, also receiving a newspaper does increase the chance that they are active in politics.

The two studies covered two different periods. The years 1996-2000 were marked by the Clinton scandals. It’s therefore possible that the effect of Fox News, like that of newspapers, reflects better access to information. In the 1996-2000 period, more information meant more bad news for Clinton. In 2005, more information means more bad news for Bush and the Republicans. There could thus be that there was no impact of the “media slant”. Improving access to any media just improves the voters’ access to information and gets them to turn against the embattled incumbent.

Another more worrying possibility is that newspaper readers know how to decipher messages, while television viewers let themselves be manipulated. To find out, the ideal would be to replicate the “Fox News” experiment in the run up to the upcoming US presidential election or even, ideally, to try and contrast voters’ access to different slant (say, Fox News on the right, CNN in the middle, or the Daily show on the left). It may be difficult, as these cable channels have greatly increased their reach already. In the mean time, the principle of prudence seems to justify a «Bayrou law» in the audio-visual realm.

Editors’ note: This column is based on a essay that appear in French in Libération.





1 Conseil supérieur de l'audiovisuel (in English: The Supreme Audio-Visual Council) is an independent administrative authority that was created in 1989 to guarantee broadcasting freedom.
2 “The Fox news effect, Media Bias and Voting” Quarterly Journal of Economics, August 2007, Vol. 122, pp. 1187-1234
3 “Does the media matter: A field experiment measuring the effect of newspaper on voting and political opinions” Yale, Working Paper, 2007


12,388 Reads