A Balanced News Diet After All?

by John Sides on April 30, 2012 · 10 comments

in Media

This is a guest post from Michael LaCour, a Ph.D. candidate in political science at UCLA.  The paper on which this post is based is here.


The reemergence of a prominent partisan press has led many scholars to investigate partisan self-selection of news outlets. Political scientists and journalists have concluded that individuals are motivated to select media sources that match their own political views and avoid media sources that challenge their political views. However, an analysis of individuals’ actual media exposure patterns lead to conclusions about selective exposure quite different from previous research based on self-reported media exposure.

I use data collected by Integrated Media Measurement Incorporated (IMMI) – to measure individuals’ actual media exposure. Nine hundred and twenty panelists from the New York and Chicago media markets were given smartphones equipped with audio recognition software that continuously and  automatically captured their exposure to all radio and television media – over the course of the 2006 midterm campaign.

Results indicate that a majority of viewers consume little or no news and the remainder consume very high levels of local news  as well as an ideologically diverse set of partisan news programs.   The figure below measures panelists’ news diet among partisan sources on television and radio.   This is a net measure of how “balanced” individuals partisan news diet is (weighted by volume).  The figure displays the distribution of panelists’ partisan news diet in the form of separate kernel density estimates for Democrats and Republicans.

On this scale a high positive score represents a panelist watching and listening to media that is conservative. Conversely, a negative score indicates a panelist watching or listening to media that is liberal – and all the more so if the exposure occurs frequently at high volume. A detailed description of how media outlets are coded is here.

The unimodal distributions centered around zero, indicate that Democrats and Republicans have similar news consumption  patterns. Only a small unrepresentative subset of the public is sorting themselves into ideologically like-minded enclaves.

To summarize, most individuals do not refuse to hear the other side. In fact, most people consume predominately non-partisan  local TV newscasts, while tuning out news from partisan  sources altogether. Of those who do turn to partisan sources, most Republicans and Democrats have virtually indistinguishable news diets. Contrary to recent claims, there is little evidence that the electorate is self-sorting into “ideologically like-minded information cocoons” at the level being described by scholars and political commentators.


Joel April 30, 2012 at 2:51 pm

I’d like to know a little bit more about the decision to code local news as having “no ideological slant.” I think it’s reasonable to expect the various local news broadcasts that viewers can choose from in major markets to be distributed across an ideological spectrum. If local news constitutes the majority of our “news diet,” then isn’t its ideological substance a more definitive test of the self-selection theory?

Michael LaCour April 30, 2012 at 7:02 pm

Thank you for your question. Most local news focuses on sports, crime and justice, and weather, with a little bit of local politics and sometimes national politics. It is typically not ideological…but if you have evidence to suggest otherwise or come up with a way to detect partisan slant in local news — let me know.

Jack C. April 30, 2012 at 3:09 pm

Joel, see p. 13 of the linked article:

“Local News- local television news programs broadcast in the New York and Chicago media markets. To my knowledge there is no transcript archive that includes all local news from New York and Chicago media markets, therefore I am unable to apply the slant method to the local news programs. All local news programs were coded a priori as “mainstream news” indicating no ideological slant. Anticipating the criticism this decision could create, the second section of results omits local news from the analysis and focuses solely on partisan news consumption.”

My question regards Internet news consumption. I’m curious to know whether there is a difference between TV/radio news choices and Internet news choices. A basic Internet connection affords the web surfer the opportunity to choose from a massive variety of ideological sources. While I’m sure more people have access to cable channels than do not, there should still be a slight discrepancy in access, thus favoring “mainstream” or local options, shouldn’t there?

A larger issue may be the lack of a discussion of treatment effects in the paper. If people are made aware that their media consumption is being monitored, what are the chances that some of the people are changing their habits? Checking the individual data against aggregate information (e.g., Nielsen ratings) does not seem to address this issue fully.

This is really cool stuff, but I’d like to see LaCour look at the Internet sources (which he says he intends to do in the conclusion) and at least address the possibility of a treatment effect. If the results hold up to further inspection, it would be a fairly counterintuitive and controversial finding, given the widespread, conventional wisdom that the proliferation of news sources has led to ideological reinforcement.

Michael LaCour April 30, 2012 at 7:26 pm

Thank you for your detailed feedback. Addressing your questions.


Data including TV, Radio and Internet tracking is the “gold standard” of media measurement…however, we are not there yet. Three points.

1) TV and Radio still account for most of the action. As you probably know, people like us that spend time reading about politics on the Internet are not representative of the population. Contrary to conventional wisdom, research using behavioral measures of media exposure demonstrate that viewers are supplementing their media diet with new media, not substituting it: 90% of all time spent with media is still with TV, while 9% of all media time is spent on the Internet and 1% of all media time is spent on mobile Internet (Edwards 2012). As long as the public remains relatively apathetic about politics, the rise of new media should not significantly increase the prevalence of selective exposure.

2) I would expect to find that people’s media consumption behavior is fairly consistent across platforms. i.e. people who watch a lot of sports on TV probably look at a lot of sports online. Indeed, I find that increased exposure to like-minded news is positively correlated with an increase in exposure to cross-cutting news. Garrett, Carnahan and Lynch (2011) find the same positive correlation using self-reported Internet behavior.

For more information on Internet blog behavior I recommend:
Lawrence, E., J. Sides and H. Farrell. 2010. “Self-segregation or deliberation? Blog readership, participation, and polarization in American politics.” Perspectives on Politics 8(01):141–157.

Treatment Effects:

Two points. First, panelists’ media habits were observed 24/7 over the course of three months…which is a long time to continually alter your habits! Second, the panelists were not told that the study was political in nature. Treatment effects are a problem that confronts all media measurement methods. People must consent to the fact that their media consumption is being tracked, I do not know of a direct way to address this…above and beyond using passive electronic measurement. If you have any ideas let me know.

According to Nielsen 88% of American households have cable…which I control for in my analysis.

Jack C. April 30, 2012 at 7:38 pm

It was not clear in the paper that they were unaware that the study was political in nature. That clears a lot up about treatment effects — I encourage you to include this in a future version to defuse the “treatment effects” bomb reviewers like to toss around.

Michael LaCour April 30, 2012 at 7:42 pm

Great suggestion. Thank you.

Adria Tinnin April 30, 2012 at 4:52 pm

Wow, what an innovative method of analyis! As a Political Science graduate student I have to say these are interesting findings amd seem to contradict some of the current discourse on the topic, such as Groseclose’s book “Left Turn: How Media Bias distorts the American Mind.” LaCour’s results show that despite the presence of media bias, most people receive relatively balanced information (as evidenced by the spike around zero for both groups). The information on internet sources would be helpful but after some research on my own, the data does not seem to be readily available…would make a great NSF project though!

Emrys April 30, 2012 at 5:52 pm

Excellent – strong study, well designed. But consumption patterns don’t reveal attitudinal response / effect. I’d be interested to see a study correlating consumption patterns with attitudinal measurement. Most media effects research suggests that the opinions people form, and the reactions to content consumed, is shaped as much by prior attitude, social context and group affiliation.

Michael LaCour April 30, 2012 at 7:34 pm

I agree! But these data do not permit an analysis of attitudinal response/effect and are probably beyond the scope of this study.

“Although other factors, such as attention during exposure (Chaffee and Schleuder 1986; Chang and Krosnick 2002), may condition the effect of political messages, the causal chain starts with exposure, and exposure appears to be consequential even when media users pay little attention (Krugman and Hartley 1970; Zukin and Snyder 1984)”

Adam April 30, 2012 at 7:54 pm

Interesting, well-executed study. I think you’re on solid ground in arguing that local TV news has no ideological slant, as so little of its content can be mapped onto the national partisan/ideological spectrum (see Rosenstiel, et. al., “We Interrupt This Newscast. . .”).

However, for this *exact* reason, I worry that your key finding is too reliant on the fact that local TV news accounts for a hefty proportion of news viewership. After all, a person consuming large amounts of local TV news, and then a little bit of Fox, is not getting a “balanced diet” anymore than if they watched CSI, American Idol, and then some MSNBC. Either local news has ideological content — in which case your assumption of balance is problematic — or it doesn’t, in which case it’s not particularly relevant to your normative inquiry about the “dire” democratic implications of information cocooning. It might be good to highlight the aspects of your findings that show surprising amounts of cross-cutting partisan viewership (as you smartly did by showcasing Figure 10 in this blog post, which is an amazing graph) and ease up a bit on the claims in the paper that rely on local TV news.

Comments on this entry are closed.

Previous post:

Next post: