SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

This study uses multiwave panel data from the 2008 presidential election to investigate the impact of partisan news exposure on changes in vote preferences over time. Overcoming key limitations of prior research, the analysis distinguishes among the potential effects originally delineated by Lazarsfeld and colleagues (1948): (1) activation—motivating partisans who initially say they are undecided or planning to defect to shift their vote back to their own party's candidate; (2) conversion—motivating partisans to shift their vote to the opposing party's candidate; and (3) reinforcement—strengthening partisans’ preference for their initial vote choice. The results reveal only modest evidence that partisan news reinforces existing vote preferences. Surprisingly, partisan news plays a more robust role motivating changes in vote choice: news slanted toward citizens’ own partisanship increased the odds of activation and decreased the odds of conversion, while news slanted away from citizens’ own partisanship proved a strong counterforce working in the opposite direction.

Do the news media influence Americans’ electoral choices? This question is far from new. It was a focal point of early empirical scholarship, driven by concerns about the mass media's influence over an unsuspecting public. Following the underwhelming evidence of media effects reported in the early Columbia studies (Berelson, Lazarsfeld, and McPhee 1954; Lazarsfeld, Berelson, and Gaudet 1948), subsequent studies likewise concluded that news media have limited capacity to influence individuals’ electoral preferences (Beck et al. 2002; Dalton, Beck, and Huckfeldt 1998; McGuire 1986; Patterson and McClure 1976). In part, these lackluster findings of media influence on vote choice reflected major U.S. news organizations’ general adherence to balanced coverage, which conveyed competing messages that cancelled out one another's effects (Zaller 1996).

But starting with talk radio and accelerating with cable/satellite TV and the Internet, the proliferation of partisan news outlets has sparked renewed debate over whether and how media exposure influences American voters. Discourse in the media and among political actors tends toward the assumption that partisan media can sway voters. For example, during the 2008 presidential campaign, Barack Obama commented, “I am convinced that if there were no Fox News, I might be two or three points higher in the polls. If I were watching Fox News, I wouldn't vote for me, right? Because the way I'm portrayed 24/7 is as a freak! I am the latte-sipping, New York Times-reading, Volvo-driving, no-gun-owning, effete, politically correct, arrogant liberal. Who wants somebody like that?” (Bai 2008). Indeed, much of the finger pointing about partisan bias—e.g., charges that MSNBC is “an organ of the Democratic National Committee” and Fox News is “a wing of the Republican Party”—reflects an underlying belief that Americans will be influenced by these media sources (CNN 2009; Kurtz 2008).

Yet scholars are less certain about the capacity of partisan news to exert effects on vote choice. Some have argued that the rise of partisan media heralds “a new era of minimal consequences” (Bennett and Iyengar 2008, 725), citing selective exposure as a key constraint on the ability of partisan media to do much more than reinforce existing attitudes. In this view, partisan media are unlikely to change people's opinions because individuals will gravitate toward politically like-minded news sources and avoid sources with a counterattitudinal slant. Even if counterattitudinal exposure does occur, people will be motivated to counterargue, thus blocking any persuasion effect.

Meanwhile, the empirical evidence has yielded few clear conclusions. Different studies have yielded contradictory findings, and methodological limitations have shackled claims of causality. Moreover, prior research has often lumped together different types of effects, making it difficult to determine the nature of partisan media influence. For example, reinforcing individuals’ initial decision to vote for their own party's candidate is different from motivating individuals to change their vote choice toward their own party's candidate—yet no study appears to have distinguished between these two types of effects. In short, it remains an open question whether the influence of partisan media is limited to reinforcement of existing preferences or encompasses the ability to change the electoral choices that citizens make.

Overcoming key limitations of prior research, this study uses multiwave panel data from the 2008 presidential election to test hypotheses about the impact of partisan news exposure on changes in vote preferences over time. Drawing on the delineations originally laid out by Lazarsfeld, Berelson, and Gaudet (1948), three potential effects are investigated: (1) activation—motivating partisans who initially say they are undecided or defecting to shift their vote back to their own party's candidate; (2) conversion—motivating partisans to shift their vote to the opposing party's candidate; and (3) reinforcement—strengthening partisans’ preference for their initial vote choice.

The Theoretical Basis for Partisan Media Effects

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

Although Lazarsfeld and colleagues’ conceptualization of activation, conversion, and reinforcement originated in a far different context—even before the spread of TV—the logic underlying these media effects can readily be extended to today's media environment. For example, news slanted toward citizens’ own political views would be expected to encourage activation as well as reinforcement. This proposition is consistent with cognitive theories of opinion formation, which identify the relative intensity of competing messages conveyed in the media as a key factor explaining change in opinion over time (Chong and Druckman 2007; Zaller 1992). The “louder” message will disproportionately determine which considerations are salient in voters’ minds, which then influences the political preferences they express. Prior theorizing has also emphasized that congruence between a message and one's prior beliefs increases the likelihood of message acceptance (Petty and Cacioppo 1996; Zaller 1992).

Applying these theories to a partisan media environment, the plausibility of activation and reinforcement effects of like-minded news rests with both the imbalance in message volume (i.e., the greater intensity of one side's message) and the congruity between the slant of the news and individuals’ own political views. For example, by providing disproportionately more like-minded arguments, news slanted toward citizens’ own party would be expected to raise the salience of positive considerations about their party's candidate and negative considerations about the opposing party's candidate and thus increase the likelihood that previously undecided partisans (or even those initially intending to defect) will shift their vote choice toward their party's candidate—i.e., an activation effect.

In a similar way, exposure to news slanted toward the candidate whom individuals already prefer would be expected to strengthen that vote preference over time—i.e., a reinforcement effect. An unbalanced pool of information, biased in favor of arguments supporting citizens’ initial preferences, has been argued to produce attitude polarization (Sunstein 2007). Thus, it makes sense to expect that exposure to news slanted toward an individual's initial vote choice would enhance the intensity of preference for that candidate over the opposing candidate. Indeed, it has been argued that, due to selective exposure, reinforcement is the most likely outcome of partisan news exposure (Bennett and Iyengar 2008).

Beyond these cognitive pathways, the effects of politically like-minded news may also have an affective component. By providing coverage that disproportionately favors individuals’ own party, like-minded news could encourage feelings of enthusiasm for the party, which fortify the link between partisanship and vote choice (Marcus, Neuman, and MacKuen 2000). This idea harks back to the hypothesized role of the press in the nineteenth century, when party papers created “a political, intensely partisan world for their readers” and “reinforced the sense of party as the basic guide to men and events” (McGerr 1986, 17; Schudson 1995).

What about conversion? Is it plausible to expect that exposure to partisan news favoring the opposing  party produces conversion? Theories of opinion formation allow ample room for a conversion effect, emphasizing that resistance to counterattitudinal messages requires the motivation and ability to discern a discrepancy between a message and one's own political values (Zaller 1992). As many have noted (e.g., Converse 1964; Delli Carpini and Keeter 1996; Zaller 1992), most Americans are neither ideologically consistent nor politically knowledgeable, potentially paving the way for acceptance of arguments that conflict with their own predispositions. Individuals may also accept counterattitudinal arguments from partisan media if they do not perceive the information as biased—a scenario perhaps most likely among moderates, who tend to perceive less media bias (Baum and Gussin 2007)—or if they hold ambivalent political views. Consideration of more competing criteria can, in turn, weaken candidate preferences and reduce consistency between party identification and vote choice (Barker and Hansen 2005).

Even among those who are alert to the counterattitudinal slant of a given news source, influence may still be exerted through more subtle aspects of partisan news coverage. For example, subtle textual or visual characteristics of partisan messages can influence political preferences by cuing certain beliefs without individuals even realizing it (e.g., Barker 2002; Valentino, Hutchings, and White 2002). The arousal of anxiety—for example, if a Democrat becomes anxious after watching Fox News’ coverage of Obama's religion (“is he a Muslim?”)—can decrease reliance on party loyalty and enhance openness to conversion (Brader 2005; Marcus, Neuman, and MacKuen 2000).

Of course, not all individuals are equally likely to be swayed by counterattitudinal news. Theories of motivated reasoning suggest that citizens with greater political sophistication and stronger political priors will be especially likely to defend their prior conclusions (Hart et al. 2009; Taber and Lodge 2006; Zaller 1992). Individuals engaging in motivated reasoning not only gravitate toward congenial messages (thus limiting exposure to counterattitudinal messages in the first place), but they also defensively process information, evaluating arguments that support their prior attitudes as stronger than opposing arguments (Eagly et al. 2000; Taber and Lodge 2006). By preventing acceptance of arguments in support of the opposite view, defensive processing inhibits attitude change. In fact, an individual may find the counterarguments he self-generates in response to an attitude-discrepant message so much more persuasive than the arguments presented by the message that he may do the opposite of what is advocated—a boomerang effect (Petty and Cacioppo 1996). Based on this line of theorizing, one might expect exposure to news with a counterattitudinal slant to trigger counterarguing among political sophisticates and individuals with stronger political priors, thus blocking conversion or even producing a boomerang effect.

Inconclusive Empirical Findings

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

Despite widespread interest in partisan media effects on vote choice, the existing body of evidence is inconclusive. Broadly speaking, prior research has yielded contradictory conclusions, with some finding that exposure to news with a partisan slant significantly influences candidate evaluations and vote choice (Barker 2002; Druckman and Parkin 2005; Kahn and Kenney 2002), and others finding weak or null effects (Gerber, Karlan, and Bergan 2009; Yanovitzky and Cappella 1999).1

A more detailed look at the literature reveals that research on partisan media often fails to distinguish among activation, conversion, and reinforcement. For example, relatively few studies examine individuals by political orientation, which is necessary in order to distinguish conversion (which would result from counterattitudinal news exposure) from activation or reinforcement (which would result from like-minded news exposure). A causal relationship between Fox News exposure and voting Republican would represent conversion for a Democratic viewer but would be activation or reinforcement for a Republican viewer.

Among the studies that do analyze individuals by political orientation, the results are mixed. Some evidence suggests that exposure to media slanted against individuals’ party affiliation leads to greater support for the opposing party's candidate, implying a conversion effect (DellaVigna and Kaplan 2007; Lee and Cappella 2001). Yet other research has found only weak conversion effects or even boomerang effects of such exposure (Yanovitzky and Cappella 1999). Meanwhile, exposure to like-minded news has been found to increase the gap in favorability toward one's own party's candidate compared to the opposing party's candidate (Stroud 2010). However, no study appears to have examined whether the impact of like-minded news reflects activation or reinforcement (or both).2

Findings from related domains of research have further complicated the picture. For example, some evidence suggests that people gravitate toward politically like-minded news (Iyengar and Hahn 2009; Stroud 2011), thus limiting the opportunity for counterattitudinal news to exert conversion effects. Moreover, increased media choice has lured less politically involved citizens toward entertainment content, leaving strong partisans—who are most likely to engage in selective exposure and resist counterattitudinal arguments—to constitute most of the news audience (Prior 2007). However, other evidence suggests that liberals and conservatives do not segregate themselves into like-minded information bubbles; rather, they overlap considerably in their media selections (Gentzkow and Shapiro 2010). In addition, many individuals appear to lack the strong preference for consistency that theoretically motivates aversion to dissonance and thus selective exposure (Cialdini, Trost, and Newsom 1995; Garrett 2009). Indeed, many partisans say they disagree with their party on personally important policy issues (which, interestingly, has been shown to increase their susceptibility to appeals from the opposing party) (Hillygus and Shields 2008).

Methodological limitations have compounded the inconclusiveness of prior research. Much of the literature has relied on cross-sectional surveys, which are not only quite vulnerable to spuriousness but also provide few definitive insights into changes in vote preferences. To assess whether activation, conversion, or reinforcement has occurred, one needs knowledge of preferences at an earlier point in time. With evidence of a cross-sectional relationship between partisan news exposure and vote choice, one would have to assume, rather than demonstrate, that changes in vote preferences occurred over time. Further, a single cross-section cannot distinguish between reinforcement among those who supported their party's candidate from the get-go and activation among initially undecided or unfaithful partisans.

Panel data have enormous potential for parsing apart the three types of effects. Unfortunately, the few existing panel studies of partisan media and vote choice have fallen short of fulfilling this potential. For example, panel studies of politically like-minded media (TV, Web, radio, and newspapers) (Stroud 2010) and of political talk radio in particular (Barker 2002; Jones 2002; Yanovitzky and Cappella 1999) have provided evidence that is consistent with reinforcement and/or activation effects. However, none of these studies distinguished activation from reinforcement. Panel findings on conversion effects have been mixed, with some scholars finding supportive evidence (Ladd and Lenz 2009) and others finding no evidence (Yanovitzky and Cappella 1999). No panel study appears to have specified distinct tests for each of the three types of effects.

Moreover, most panel studies (e.g., Barker 2002; Jones 2002; Stroud 2010; Yanovitzky and Cappella 1999) have used lagged dependent variable models, which have a variety of problems (Achen 2001; Allison 1990). Lagged dependent variable models do not assess within-person change in preferences over time but rather change in the rank-order of different people. Nor do these models maximize the potential of panel data to rule out confounders. As discussed below, fixed effects methods provide far more powerful means to analyze panel data by using only within-person variation over time.

Although experimental studies can make the strongest case for establishing causality, they raise generalizability concerns because of their inability to take selective exposure into account (Bennett and Iyengar 2008). Experiments run the risk of identifying the effects of “forced” exposure among those who may not be exposed in the real world, rather than the effects among those who select into an audience.

The Present Study

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

Focusing on the 2008 presidential election, this study uses panel data and fixed effects methods to investigate the potential activation, conversion, and reinforcement effects of exposure to partisan news on television.3 To test for these effects, I employ panel data that measured vote choice at three time points: two pre-election waves (one before the party conventions and one during the fall campaign) and a post-election wave. It is thus possible to examine changes in vote preferences that took place in the months preceding the election as well as changes that manifested in the ultimate vote that partisans cast on Election Day.

Three hypotheses are tested. The first two focus on citizens whose vote choice changed over time—that is, partisans who were activated or converted during the campaign. Hypothesis 1 posits that greater exposure to news slanted toward individuals’ own party increased their odds of changing their vote toward their party's candidate (activation) and decreased their odds of changing their vote choice toward the opposing party (conversion). Hypothesis 2 posits that greater exposure to news slanted toward the opposing party exerted the reverse effects, increasing the odds of conversion and decreasing the odds of activation.

The final hypothesis focuses on the potential effect of partisan news among citizens whose vote choice did not change over time—that is, individuals who maintained a constant vote preference across the three time points. By definition, these voters were neither activated nor converted during the time period examined; rather, partisan news potentially reinforced their constant vote choice, strengthening that candidate preference. Therefore, Hypothesis 3 posits that greater exposure to news slanted toward the party of their preferred candidate increased individuals’ favorability toward that candidate and decreased their favorability toward the opponent.4, 5

Based on theories of motivated reasoning discussed above, one might posit that news slanted toward the opposing party will only spur conversion (and deter activation) among more “persuadable” groups such as less knowledgeable and weaker partisans, while more informed and strong partisans will prove resistant or even boomerang. Though this scenario is possible, its plausibility is undermined by the fact that the most committed partisans, who would be most resistant to counterattitudinal news, are not likely to be among those whose vote choice changed over time. In other words, it seems unlikely that those who changed their vote choice over time held the kind of strong candidate views that would motivate a boomerang response to counterattitudinal news. Still, to provide an empirical answer to this question of potential group differences, I conduct additional analyses to evaluate whether groups that tend to engage in motivated reasoning (highly knowledgeable and strong partisans) registered a distinct pattern of effects of partisan news exposure.

Method

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

Data from the 2008 National Annenberg Election Study's (NAES) Internet panel are used for this study. A nationally representative sample of American adults was initially recruited through random digit dialing (RDD) by Knowledge Networks, of Palo Alto, CA, and then was given Internet access if needed and interviewed via personal computer or WebTV.6 The NAES panel was interviewed five times over the course of 16 months: wave 1 interviews were conducted during the preprimary period (October 2–December 31, 2007); wave 2 was conducted the first three months of primaries (January 1–March 31, 2008); wave 3 spanned the end of the primary campaign through to the party conventions (April 2–August 28, 2008); wave 4 encompassed the general election campaign (August 29–November 4, 2008); and wave 5 was conducted during the post-election period (November 5, 2008–January 31, 2009). Only self-identified partisans (strong, weak, or leaning) who expressed a non-third-party vote choice are included in the analyses (N = 9,008).7, 8

In the models examining activation and conversion, the dependent variable is Vote Choice, measured twice during the general election campaign (waves 3 and 4) and once after Election Day (wave 5). In the models examining reinforcement, the dependent variables are Favorability toward the Preferred Candidate and Favorability toward the Opponent, measured with standard 0–100 feeling thermometer items for John McCain and Barack Obama in waves 3 and 4 (Appendix B provides exact question wording and variable coding).9

To assess exposure to partisan news on television, I used a set of items tapping exposure to the 45 most frequently watched TV news programs, as determined by Nielsen ratings. Respondents were shown lists of 15 programs at a time and asked: “Which of the following programs do you watch regularly on television? Please check any that you watch at least once a month.”10 Respondents could check off as many programs as they wanted or “none of the above.” To capture a variety of potential sources of political news, the lists included network newscasts, cable news programs, political talk shows, and political satire programs. This program-level measure of exposure has been shown to be both reliable and valid (Dilliplane, Goldman, and Mutz 2013).

Each of the 45 TV programs was coded as slanted toward the Democrats, the Republicans, or neither (neutral), based on general public perceptions of that program. To evaluate public perceptions of each program's partisan orientation, I used independently collected survey data on individuals’ perceptions of partisan slant in the news programs they watched. These data came from a separate national telephone survey, which asked a wholly separate sample of respondents to identify the TV program they got most of their campaign information from, and which candidate they thought the program favored, if any.11

For a subset of the programs, there were insufficient survey data on perceived partisan slant because some programs were mentioned by only a few respondents or not at all. In those cases, I combined the survey data with a Lexis-Nexis search of news coverage, which was used to determine if programs or their hosts were generally associated with a partisan orientation or not (assuming that news coverage both reflects and shapes general public perceptions of news sources’ partisan slant). Based on a systematic coding procedure, 15 programs were coded as slanted Democratic, 11 programs as slanted Republican, and 19 programs as neutral. The categorization of programs produced by this coding system is broadly consistent with prior content analyses (e.g., Groeling 2008; Groseclose and Milyo 2005; PEJ 2008) and with how previous studies of partisan media effects have treated various TV channels (e.g., Baum and Groeling 2009; Iyengar and Hahn 2005; Stroud 2010). Moreover, the measures of exposure to each type of programming demonstrate good reliability: .84, .90, and .81 for the number of programs slanted Democratic, slanted Republican, and neutral, respectively. (Appendix A provides full details of the coding system.)

For the analyses of activation and conversion, the key independent variables are the Number of programs slanted toward one's own party and the Number of programs slanted toward the opposing party, based on respondents’ own partisanship reported in wave 1. For the analyses of reinforcement among those with a constant vote choice, the key independent variables are the Number of programs slanted toward the preferred candidate's party and the Number of programs slanted toward the opponent's party.

Method of Analysis

This study employs fixed effects regression, which offers important advantages over cross-sectional regression models as well as lagged dependent variable models. First, fixed effects models capitalize on repeated measures to produce estimates of predictors’ effects on within-person change over time. This advantage is crucial because the hypothesized effects of partisan news imply that exposure produces change within individuals over time.

Second, fixed effects regression maximizes the robustness of the results against spuriousness. Unlike lagged dependent variable models, fixed effects models only use within-person variation, comparing each person to him- or herself at an earlier point in time. Thus, fixed effects regression automatically controls for the constant effects of all preexisting stable variables, whether measured or unmeasured, as if they had been included in the model (Allison 2009). For example, one might suspect that younger or less sophisticated voters are both more susceptible to conversion and more likely to be exposed to counterattitudinal media. The fixed effects estimation method used in this study accounts for this possibility—indeed, that is one of its key strengths. Because the models only use within-person variation, any characteristic that happens to lead some partisans both to watch counterattitudinal news and vote for the opposing party (or, conversely, to watch like-minded news and vote for their own party) is ruled out as an alternative explanation.

Third, fixed effects models enable estimation of the effects of stable levels of exposure—i.e., citizens’ habitual programs of choice—as well as the effects of changes in levels of exposure over time. Prior research indicates that TV consumption tends to be a habitual behavior (Adams 2000; Rosenstein and Grant 1997) and that citizens exhibit considerable stability in viewership, reporting not only the same number of programs but the same exact programs over time (Dilliplane et al. 2013). In the present study, a similar picture of stability emerged for partisan news exposure across the three waves in which exposure was measured (waves 2, 4, and 5).12

Relatively stable levels of exposure do not preclude effects. Even if the number of programs one consumes does not change, the amount of coverage that candidates receive does change during an election. Due to the massive increase in media attention to presidential candidates that occurs during the general election campaign, even citizens who watch an unchanging number of partisan programs would be exposed to an infusion of campaign-related information, which could, in turn, produce changes in vote preferences. In other words, the content of individuals’ regular programs of choice changes, creating the potential for effects.

To capture stable levels of exposure to partisan programs, I took the average number of programs that each individual reported watching across the three waves in which the measures appeared. To estimate a stable predictor's impact on change in a dependent variable in a fixed effects model, the predictor must be interacted with a variable for time (captured by the Wave variable). This is because the main (i.e., constant) effects of all stable predictors automatically drop out of the model, as there is no way that the unchanging influence of an unchanging characteristic could cause within-person change in the dependent variable. The interactions with time reveal whether one's stable level of partisan news exposure (e.g., habitually watching more like-minded programs versus habitually watching fewer) affects the rate of change in candidate favorability or changes the odds of a given vote choice.

Despite relatively high levels of stability, some individuals did exhibit change in the number of partisan programs they watched. To examine the possibility that a change in exposure produced a change in the dependent variables, I ran a second set of fixed effects models estimating the impact of within-person change in the number of partisan programs watched on within-person change in the outcomes.13

Results

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

I begin with an overview of patterns in vote choice and partisan news exposure during the 2008 presidential campaign, which provides a useful backdrop against which to interpret the results of the hypothesis tests. Partisans proved to be highly stable in their vote choice over the course of the general election campaign. A large majority (72.7%) reported the same major party vote choice across waves 3, 4, and 5—that is, from the spring/summer months prior to the conventions through to the post-election report of vote choice. Among these unchanging, or “constant,” voters, almost all supported their own party's candidate (90.8% of constant voters, or 66.1% of the sample), though there was a small group of staunch defectors. Because these partisans stuck with the initial decision they reported in the spring or summer of 2008 (wave 3), they were not part of the pool of voters whose exposure to partisan news potentially produced a change in vote choice (i.e., activation or conversion). In other words, for most partisans, it seems that the potential influence of partisan news is limited to reinforcement of initial vote preference.

The partisans who did exhibit a change in vote choice over time comprised 25.7% of the sample.14 Some shifted their vote choice between the preconvention period and the fall campaign (between waves 3 and 4), others shifted between the fall campaign and Election Day (between waves 4 and 5), and still others shifted both times. The changes included moving in and out of indecision, shifting from one major party candidate to the other, and changing from support for a candidate to abstention. When it came to reporting their Election Day decision (wave 5), a little more than a third of these shifting partisans (38.8%) ultimately decided to vote for their own party's candidate, while about half as many (18.9%) decided to defect, and the remainder (42.3%) ultimately abstained.

As might be expected, patterns in partisan news exposure differed by voting behavior (see Figure 1). Among partisans with a constant vote choice, exposure to like-minded news (i.e., slanted toward the preferred candidate's party) on average dwarfed exposure to news with the opposite partisan slant (toward the opponent's party). Partisans with a changing vote choice also favored like-minded news (i.e., slanted toward their party) over news with the opposite partisan slant (toward the opposing party), but the difference was not as glaring as among constant voters. Interestingly, although both groups exhibited some tendency toward news with a congenial slant, there was little evidence that partisans walled themselves off into like-minded media bubbles. Among those watching like-minded programming, 58.6% also watched news with the opposite slant, and 89.1% also watched neutral programming. Only 5.7% of constant and changing voters watched only like-minded programming.15

image

Figure 1. Exposure to Partisan TV Programs among Partisans Whose Vote Choice Changed or Remained Constant over Time

Note: The figure shows mean levels of exposure reported by respondents across three waves of data. For partisans whose vote choice changed over time, programs were coded as slanted “toward” or “away from” respondents’ own views based on their party identification. For partisans whose vote choice was constant, program slant was coded based on the party of their preferred candidate. For both groups, neutral news exposure was, on average, significantly greater than exposure to news slanted toward respondents’ views, which was in turn significantly greater than exposure to news with the opposite slant (p < .001 in all comparisons, based on paired-samples t-tests).

Download figure to PowerPoint

Activation and Conversion

Turning to the hypothesis tests, the first set of results focuses on the capacity of partisan news exposure to motivate changes in vote choice over time. Hypothesis 1 posits that greater exposure to news slanted toward one's own party increased the odds of activation and decreased the odds of conversion. Hypothesis 2 posits that greater exposure to news slanted toward the opposing party exerted the reverse effects.

Table 1 reports the results of the models estimating the effects of stable levels of partisan news exposure on activation and conversion. Four binary logit fixed effects models are shown: two correspond to activation (predicting change toward one's own party's candidate between waves 3 and 4 and between waves 4 and 5) and two correspond to conversion (predicting wave 3–4 and wave 4–5 change toward the opposing party's candidate).16 These models include only respondents whose vote choice changed over time; by design, it is not possible to estimate a fixed effects model predicting vote choice for respondents whose vote choice did not change over time because there is no within-person variation in the outcome variable.17 The surprisingly small sample sizes reflect the fact that only a minority of partisans (N = 2,317) changed their vote choice over time, and among these respondents, only some of them changed between any two adjacent waves. Moreover, the models examining wave 4–5 change necessarily drop abstainers, thus reducing the sample size further.18

Table 1. The Impact of Stable Levels of Partisan News Exposure on the Odds of Activation and Conversion
 ActivationConversion
 Wave 3–4Wave 4–5Wave 3–4Wave 4–5
 Coef. (SE)Coef. (SE)Coef. (SE)Coef. (SE)
Note
  1. Table reports unstandardized coefficients from binary logit fixed effects models. Number of Programs Slanted toward Own Party refers to the average number of programs that an individual respondent watched across three waves of data, which is used as the indicator of stable exposure; the same applies to Number of Programs Slanted toward Opposing Party and Number of Neutral Programs. Activation refers to change in vote choice toward one's own party's candidate; conversion refers to change in vote choice toward the opposing party's candidate. The wave 4–5 models exclude respondents who reported abstaining (in wave 5).

  2. †p < .10, *p < .05, **p < .01, ***p < .001.

Number of Programs Slanted toward Own Party*Wave.094*.112−.111−.210*
 (.042)(.079)(.059)(.089)
Number of Programs Slanted toward Opposing Party*Wave−.178***−.323**.237***.249*
 (.050)(.100)(.061)(.110)
Number of Neutral Programs*Wave−.005.139*−.033.018
 (.027)(1.293)(.038)(.063)
Wave.505***1.293***−.355**1.244***
 (.093)(.163)(.118)(.197)
N1035502656332

Beginning with the top row of coefficients (the Number of Programs Slanted toward Own Party*Wave interaction), the results provide moderately strong support for Hypothesis 1. In the “Activation” columns, the positive coefficients indicate that greater exposure to news slanted toward one's own party increased the odds of shifting one's vote toward one's own party's candidate between each pair of waves, though the effect fell short of statistical significance between waves 4 and 5 (possibly due to the smaller sample size). Moreover, in the “Conversion” columns, the negative coefficients indicate that greater exposure to news slanted toward one's own party had the expected effect of decreasing the odds of conversion across both pairs of waves.

Turning to the second row of coefficients (the Number of Programs Slanted toward Opposing Party* Wave interaction), the results clearly support Hypothesis 2. In the two columns labeled “Activation,” the negative coefficients indicate that greater exposure to news slanted toward the opposing party significantly decreased the odds of activation between both pairs of waves. In addition, the positive coefficients in the two columns labeled “Conversion” reveal that greater exposure to programs slanted toward the opposing party had the expected effect of increasing the odds of conversion between both pairs of waves. In fact, across all four models, these coefficients proved to be larger and more consistently significant than the coefficients for news slanted toward citizens’ own party, suggesting that news with a counterattitudinal partisan slant was a stronger influence on the odds of activation and conversion.

Note that, in each model, the coefficient for the Wave variable captures the average total effects of all other time-varying influences producing change in the odds of a given vote choice over time. Thus, each partisan news interaction coefficient indicates the change in odds that occurred above and beyond the change over time driven by other time-varying factors. In addition, because the models simultaneously include interactions between Wave and each of the three types of TV programming, each interaction coefficient represents the effect of one type controlling for the effects of the other two types. As explained above, the models also automatically control for the constant (main) effects of all stable characteristics.

To translate these findings into more substantive terms, I calculated the predicted probability of activation and conversion for partisans watching the mean number of each type of programming, compared to partisans whose exposure to like-minded or counterattitudinal programming was two standard deviations above the mean (other variables held at their means). Among partisans watching the mean number of each type of programming, the predicted probability of wave 3–4 activation was .61, and the predicted probability of wave 3–4 conversion was .41. By comparison, among partisans whose exposure to news slanted toward their own party was two standard deviations above the mean, the predicted probability of activation increased to .69, and the predicted probability of conversion decreased to .31. Among partisans whose exposure to news slanted toward the opposing party was two standard deviations above the mean, the predicted probability of activation dropped to .48, and the predicted probability of conversion jumped to .58. In other words, the changes in the odds of activation and conversion driven by higher levels of like-minded news exposure were relatively smaller than the changes driven by higher levels of exposure to news with the opposite partisan slant. A similar pattern emerged for wave 4–5 changes in vote choice.

Table 2 reports the effects of changes in partisan news exposure over time. Only paltry evidence emerged that a change in exposure produced a change in the odds of activation or conversion. In the models for wave 4–5 change in vote choice, there was support for Hypothesis 1: an increase in the number of programs slanted toward one's own party predicted both an increase in the odds of activation and a decrease in the odds of conversion. This pattern suggests that an increase over time in like-minded news exposure helped bring partisans home to their party and discourage defection late in the campaign. By contrast, no significant effects emerged for changes in exposure to programs slanted toward the opposing party. Overall, these lackluster findings suggest that changes in vote choice were driven less by changes in viewing habits and more by changes in the content of the programs that partisans habitually watched. The massive infusion of campaign information that was injected into partisans’ regular programs during the presidential race appears to have largely swamped any increases in exposure to campaign information produced by changes in individual viewing habits.19

Table 2. The Impact of Change in Partisan News Exposure on the Odds of Activation and Conversion
 ActivationConversion
 Wave 3–4Wave 4–5Wave 3–4Wave 4–5
 Coef. (SE)Coef. (SE)Coef. (SE)Coef. (SE)
Note
  1. Table reports unstandardized coefficients from binary logit fixed effects models. The models predicting wave 3–4 change in vote choice use wave 2–4 change in exposure to the three types of TV programming; the models predicting wave 4–5 change in vote choice use wave 4–5 changes in exposure. Abstainers were excluded from the wave 4–5 models.

  2. †p < .10, *p < .05, **p < .01, ***p < .001.

Change in Number of Programs Slanted toward Own Party−.021.187−.046−.379**
 (.049)(.103)(.069)(.132)
Change in Number of Programs Slanted toward Opposing Party.062−.100−.021.066
 (.059)(.122)(.073)(.142)
Change in Number of Neutral Programs−.021−.005.044.055
 (.033)(.061)(.040)(.070)
Wave.434***1.481***−.313***1.346***
 (.065)(.117)(.080)(.140)
N1035502656332

Interestingly, additional analyses revealed little evidence of differential effects across levels of knowledge or partisan strength.20 In particular, it appears that the effects of stable levels of exposure to news slanted toward the opposing party (observed in Table 1) were not confined to less informed and weaker partisans, groups typically characterized as susceptible to influence. For the most part, highly knowledgeable and strong partisans appeared to be swayed just as much.

Even more striking, the effects of a change in exposure to such programming were more powerful among strong partisans compared to weaker partisans: an increase in the number of programs slanted toward the opposing party produced a significantly stronger negative effect on the odds of wave 3–4 activation and a significantly stronger positive effect on the odds of wave 3–4 conversion among strong partisans (with a similar pattern in the wave 4–5 models)—quite the opposite of a boomerang effect. The only (nonsignificant) hint of a boomerang response among high-knowledge respondents was found in the effects of change in exposure to news slanted toward the opposing party on the odds of wave 3–4 activation and conversion. One explanation for the general lack of resistance among knowledgeable and strong partisans is that individuals who changed their vote choice over time (i.e., were activated or converted during the campaign) did not hold the type of strong candidate preference that would likely motivate sturdy resistance to counterattitudinal news.21

Reinforcement

What about the majority of partisans who maintained a constant vote choice across the general election campaign? Hypothesis 3 posits that exposure to news slanted toward the party of the preferred candidate increased favorability toward that candidate and decreased favorability toward the opponent—i.e., reinforcing the initial vote choice.

Table 3 shows the results for stable levels of partisan news exposure. Very weak support for the hypothesis emerged. The positive coefficient for the Number of Programs Slanted toward Preferred Candidate's Party*Wave interaction suggests that greater exposure produced a marginally significant increase in favorability toward the preferred candidate. However, this type of programming also exerted a significant positive effect on favorability toward the opponent—precisely the opposite of expectations. These contradictory effects imply that such exposure produced no net benefit for the preferred candidate.

Table 3. The Impact of Stable Levels of Partisan News Exposure on Reinforcement of Constant Vote Choice
 Change in FavorabilityChange in Favorability
 toward Preferred Candidatetoward Opponent
 Coef. (SE)Coef. (SE)
Note
  1. Table reports unstandardized coefficients from linear fixed effects models. Number of Programs Slanted toward Preferred Candidate's Party refers to the average number of programs that an individual respondent watched across three waves of data, which is used as the indicator of stable exposure; the same applies for Number of Programs Slanted toward Opponent's Party and Number of Neutral Programs. The models predict wave 3–4 change in candidate favorability.

  2. †p < .10, *p < .05, **p < .01, ***p < .001.

Number of Programs Slanted toward Preferred.140.234**
Candidate's Party*Wave(.078)(.086)
Number of Programs Slanted toward.057.120
Opponent's Party*Wave(.170)(.187)
Number of Neutral Programs*Wave−.164*−.099
 (.076)(.084)
Wave6.572***−2.636***
 (.303)(.334)
Constant71.051***24.470***
 (.138)(.152)
N64736381

What might explain the underwhelming evidence of reinforcement? One possibility is that reinforcement occurred as a result of increases over time in the number of partisan programs watched, rather than stable levels of exposure. Perhaps the programs that these partisans habitually watched helped them form their candidate preference in the first place, but additional increases in like-minded program exposure were needed to produce any subsequent effect on the intensity of that preference.

Table 4, which reports the effects of changes in exposure, provides some support for this possibility. An over-time increase in the number of programs slanted toward the preferred candidate's party predicted a significant increase in favorability toward that candidate, in line with Hypothesis 3. However, this reinforcement effect appears quite modest in size. Increasing one's exposure to news slanted toward the preferred candidate's party by one program predicted a mere 0.29-point increase in favorability (on a 100-point scale). Moreover, an increase in exposure to this type of programming failed to predict a corresponding decrease in favorability toward the opponent. News slanted toward the preferred candidate's party appears to be more effective at burnishing the image of that candidate as opposed to tarnishing the opponent's image.22 In sum, partisan news with a like-minded slant does serve a reinforcement function, but it is a fairly limited one, perhaps because constant voters already have relatively developed attitudes toward the candidates.

Table 4. The Impact of Change in Partisan News Exposure on Reinforcement of Constant Vote Choice
 Change in FavorabilityChange in Favorability
 toward Preferred Candidatetoward Opponent
 Coef. (SE)Coef. (SE)
Note
  1. Table reports unstandardized coefficients from linear fixed effects models. The models use wave 2–4 change in exposure to the three types of TV news programming to predict wave 3–4 change in candidate favorability.

  2. *p < .05, **p < .01, ***p < .001.

Change in Number of Programs Slanted toward.288*.113
Preferred Candidate's Party(.124)(.136)
Change in Number of Programs Slanted toward Opponent's Party.032.262
 (.198)(.219)
Change in Number of Neutral Programs−.093.026
 (.103)(.114)
Wave6.380**−2.291***
 (.199)(.220)
Constant70.672***23.938***
 (.363)(.403)
N64736381

The weak evidence of reinforcement is surprising, given that this effect has been argued to be the most likely outcome of partisan news exposure (Bennett and Iyengar 2008). To see whether a stronger reinforcement effect of like-minded news was lurking among certain groups, such as those most likely to engage in motivated reasoning, I investigated potential differences across levels of knowledge and partisan strength.23 The results yielded somewhat larger reinforcement effects, though not consistently for a particular group. Stable levels of exposure to news slanted toward the preferred candidate's party had a significantly stronger reinforcement effect among highly knowledgeable citizens, while changes in exposure appeared to exert stronger reinforcement effects among the less informed and weaker partisans. Yet even among those registering a stronger impact, the size of the reinforcement effect was still relatively small (around a half of a point in favorability per additional program watched).24 In short, the evidence for a reinforcement effect remained modest.

Discussion

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

This study has marshaled the strengths of panel data, an unusually detailed set of program-level exposure measures, and fixed effects methods in order to evaluate the potential effects of partisan TV news exposure on changes in vote preferences over time. The results provide only modest evidence that partisan news serves a reinforcement function, while more robust evidence emerged for its role in motivating activation and conversion. Particularly noteworthy is the significant influence of programs slanted toward the opposing party; instead of repelling viewers, habitual levels of exposure to such programs appear to sway partisans away from their own party and into the arms of the opposing party.

In evaluating these findings, careful consideration must be paid to questions of causality. Given the use of fixed effects methods, the case against spuriousness is far stronger than in cross-sectional models and even lagged dependent variable models. Because the constant effects of all preexisting individual differences (i.e., the huge amount of unobserved heterogeneity plaguing between-person analyses) were automatically controlled, one can confidently lay to rest typical concerns about the effects of unobserved third variables explaining away the observed relationships. Further, the Wave variable captured the average total effects of all other time-varying influences producing within-person change in the dependent variables.

One still might suspect that the changing effect of some stable characteristic is driving the relationships. For example, one might suspect that the impact of partisan strength on the odds of activation increased over time (i.e., partisan strength exerted a stronger impact later in the campaign). While it is difficult to conceptualize how this type of changing effect of a stable variable would confound the relationship between stable and changing levels of partisan news exposure and changes in the dependent variables, supplementary analyses were conducted to test for this more complex level of spuriousness. Each of the models was replicated, this time with interactions between the Wave variable and age, gender, race, education, income, strength of partisanship and ideology, political interest, political knowledge, and interpersonal political discussion. The replications also included variables capturing exposure to campaign information in newspapers, on the radio, on the Internet, and in news magazines and exposure to political advertising on television.25 The results (reported in Appendix C) suggest largely the same substantive conclusions, though there were a few signs of a weakening in the effects of programs slanted toward citizens’ own views.26 Particularly robust were the findings for stable exposure to programs slanted toward the opposing party; the negative effect on activation and the positive effect on conversion remained strongly significant and even increased in size in some cases.

Although panel data make it clear that change over time is occurring, it is possible that the causal arrow runs in the reverse direction. For example, it is possible that some other source produced changes in vote choice or favorability, and this, in turn, produced changes in exposure to partisan news. Without experimental data, this possibility cannot be definitively ruled out. However, it is difficult to conceive of a force that could accomplish this beyond those already ruled out above. Moreover, reverse causation is a weaker threat to the extent that exposure has effects even when it is stable over time. As programs that individuals habitually watch begin covering the presidential candidates during the campaign, the resulting change in the content of exposure is more truly exogenous because it occurs without individuals taking action (i.e., changing their viewing selections). The fact that stable exposure, far more than changing exposure, predicted changes in vote choice thus bolsters the case against reverse causation. In sum, though further research is needed, this investigation takes an important step forward in making a strong causal case for partisan news effects on electoral preferences.

This study's results did not support the argument that, in the current high-choice environment, the only media effects are most likely to be reinforcement, rather than conversion. Because of selective exposure, the argument goes, there will be few opportunities for partisans to be exposed to counterattitudinal information, and any counterattitudinal exposure that does occur will fail to produce conversion effects due to defensive processing.

In contrast, this study reveals a more robust role for news slanted away from partisans’ own views. Consistent with other recent findings (Garrett 2009; Gentzkow and Shapiro 2010), I find that most partisans do prefer news with a like-minded slant over news with the opposite slant, but many are willing to watch a certain amount of the latter (and few consume only like-minded sources). More importantly, exposure to news slanted toward the opposing side proved capable of motivating changes in vote choice, exerting a consistent influence across the campaign. Even groups expected to be highly resistant appeared to sway to the tune of such counterattitudinal news, with highly knowledgeable and strong partisans exhibiting largely the same effects as less informed and weaker partisans.

The relatively weak evidence for reinforcement effects found in this study may in part reflect the fact that news with a like-minded slant offers information that is largely redundant with beliefs that partisans already hold. This information may be more influential among those who are undecided or have strayed into the opposing party's camp—as illustrated by the observed activation effects—but an equally strong punch appears to be packed by news slanted away from citizens’ own views. This is not to say that partisan news spurred mass defection; the point is that conversion appears to be as important a consequence of partisan news exposure as reinforcement is.

In terms of normative implications, it is possible to draw positive or negative conclusions, depending on one's perspective. For example, the conversion effect of partisan news may be viewed as cause for concern, providing evidence that partisans are the unwitting victims of media sources that dupe them into a vote choice inconsistent with their interests. This interpretation, which essentially portrays viewers as being pushed around willy-nilly, seems somewhat dubious in light of the fact that even highly knowledgeable partisans exhibited effects of counterattitudinal news exposure. An arguably more persuasive (and optimistic) interpretation would be to understand these effects as a welcome signal that partisans are not so deeply entrenched in their beliefs that they refuse to update their political judgments in the face of challenging information. Partisans’ exposure to news sources slanted against their own views may be understood as approximating standards of deliberative democratic theory, which posits that citizens should expose themselves to views from both sides of the debate in order to make an informed decision.27 Of course, conversion does not necessarily signal thoughtful decision making. Still, the fact that some partisans exhibited effects of exposure to news slanted away from their own views suggests a willingness to engage in the kind of open-minded thinking that is often mourned as missing in contemporary American politics.

The effects of like-minded news may likewise be cast in either a positive or negative light. For example, to the extent that activation and reinforcement effects occur, they may be understood as encouraging blind partisan allegiance and more biased views of the candidates. An alternative is to see these effects as encouraging the kind of stronger political preferences that typically enhance citizens’ propensity to participate in politics, thus supporting the aims of participatory democracy. This view is consistent with evidence showing a mobilization effect of like-minded news exposure (Dilliplane 2011; Stroud 2011). Activation effects may also be understood as helping to bring partisans’ attitudes and behaviors into line with their political orientation, which would be consistent with the concept of constraint (Converse 1964).

Regardless of whether one finds the pessimistic or the optimistic perspective more compelling, the results of this study indicate that exposure to partisan news sources did matter in the 2008 presidential campaign. Though it is important to avoid exaggeration of their influence over citizens, partisan news organizations do wield real power in the battle for ballots. In response to the question of whether partisan media influence Americans’ vote choice, the answer appears to be: yes, they can.

  1. 1

    Similar to research on vote choice, studies of partisan media effects on policy preferences and evaluations of public figures have yielded conflicting conclusions (see, e.g., Feldman 2011; Jamieson and Cappella 2008; Jones 2002; Owen 1997).

  2. 2

    See Finkel (1993) and Hillygus and Jackman (2003) for evidence of activation effects of political campaigns and campaign events.

  3. 3

    Television, though certainly not the only source of partisan news, remains the public's top news source (Pew Research Center 2010), making this medium a natural focus of scholarly inquiry.

  4. 4

    As indicated by these hypotheses, activation and conversion involve a different dependent variable (change in vote choice) from reinforcement (change in the strength of candidate preference). The concept of reinforcement implies that like-minded media exposure will bolster citizens’ vote choice, preventing them from straying into indecision or the other party's camp; that is, exposure prevents change in vote choice. But one cannot estimate a model assessing the capacity of media exposure to produce no change; in order to estimate a model assessing the capacity of partisan news exposure to reinforce vote choice, it is necessary to have within-person variation in the dependent variable. Given that individuals who did not change their vote choice over time, by definition, did not have within-person variation in vote choice, I examined within-person changes in candidate favorability in order to detect the potential reinforcement effect of partisan news exposure. The reasoning is that, if like-minded news reinforces vote choice, it should be possible to detect at least some strengthening of candidate preference over time, as captured by the favorability items.

  5. 5

    Because it is possible that some partisans maintained a constant vote preference for the opposing party's candidate (e.g., a Democrat who maintained a constant preference for John McCain), the test for reinforcement effects focuses on news slanted toward the party of an individual's preferred candidate rather than news slanted toward the individual's own party.

  6. 6

    Chang and Krosnick (2009) show that the sample representativeness and response quality of Internet surveys using probability samples (i.e., the Knowledge Networks panel) are as good as traditional RDD telephone interviewing.

  7. 7

    The hypotheses for activation and conversion do not apply to “pure independents” (those who did not lean toward either of the major parties) because these individuals do not have a partisan predisposition that can be activated or from which they can be converted. Although the reinforcement hypothesis could extend to pure independents who maintained a constant vote choice over time (because the hypothesis focuses on the impact of news slanted toward the party of the preferred candidate, rather than citizens’ own party), I limited the sample to partisans in order to be consistent across analyses. This decision only dropped a tiny proportion of respondents (N = 139) from the analysis of reinforcement. Moreover, the findings for reinforcement reported below did not change at all when the analyses were rerun with these independents included.

  8. 8

    Of the total sample interviewed in wave 1 (N = 19,190), 54.6% completed all five waves (N = 10,472). The demographic composition of this five-wave sample is comparable to the July 2008 Current Population Survey sample. Within the five-wave sample, 444 independents and 1,020 respondents who expressed a third-party vote choice (or refused) were dropped, yielding the final sample size of 9,008.

  9. 9

    Wave 5 favorability is not used because it may have been influenced by the election results.

  10. 10

    Defining “regular” exposure as at least once a month is a relatively low bar. A person who watched a single program once during the whole month would not be exposed to very much political information, and thus he or she would be unlikely to register a strong impact. Moreover, an individual who watched a program five times per week, every week, would be counted as having the same exposure level as an individual who watched the same program only twice a month. These limitations of the measure likely make it somewhat more difficult to detect significant effects of exposure, and the results should be interpreted with this in mind. On the positive side, by using a broad definition of regular exposure, the program-level measure efficiently captures an individual's exposure to multiple, diverse types of programs, including those that only air on a weekly basis or even less regularly (e.g., PBS's Frontline), while also avoiding the heavy cognitive demands of asking respondents to mentally tally their average number of hours or days of exposure per week (Prior 2009).

  11. 11

    The data came from the 2008 NAES telephone component, which employed a large, nationally representative sample and a rolling cross-sectional design (N = 45,105). More information is available at http://www.annenbergpublicpolicycenter.org/NewsDetails.aspx?myId=263.

  12. 12

    Using paired-samples t-tests, mean changes in exposure to partisan news programs were as follows. For partisans whose vote choice changed: mean change in number of programs slanted toward their own party was .031 (p = ns) between waves 2 and 4 and .044 (p < .10) between waves 4 and 5; mean change in programs slanted toward the opposing party was .003 (p = ns) between waves 2 and 4 and .036 (p < .10) between waves 4 and 5. For partisans with a constant vote choice: mean change in number of programs slanted toward the preferred candidate's party was .267 (p < .001) between waves 2 and 4 and .084 (p < .001) between waves 4 and 5; mean change in programs slanted toward the opponent's party was −.02 (p < .10) between waves 2 and 4 and −.01 (p = ns) between waves 4 and 5.

  13. 13

    For the models predicting wave 3–4 change in vote choice and wave 3–4 change in candidate favorability, the media exposure measures from waves 2 and 4 were used. For the models predicting wave 4–5 change in vote choice, the media exposure measures from waves 4 and 5 were used.

  14. 14

    The remaining 1.6% was undecided in both waves 3 and 4 and then abstained in wave 5.

  15. 15

    This percentage reflects partisans who reported watching only like-minded programs across the three waves in which exposure was measured, including those who watched solely like-minded programming in one or two of the waves and no programming at all in the remaining wave(s).

  16. 16

    For the models predicting wave 3–4 change in vote choice, an alternative to the binary logit models would be to run a multinomial logit model. The drawbacks of the multinomial logit model are that the fixed effects estimates are population averaged instead of subject-specific (Allison 2009), which might bias the estimates, and the results are more complicated for readers to interpret. Given that the binary logit and multinomial logit models yielded the same substantive conclusions, I opted to report the binary logit results for ease of presentation.

  17. 17

    Indeed, if a fixed effects model predicting change in vote choice were run with individuals whose vote choice did not change over time, they would simply drop out of the model because fixed effects regression only uses within-person variation on the dependent variable. Thus, it is not possible (nor is it theoretically logical, as explained in note 4) to run a single fixed effects model that incorporates all three types of effects (activation, conversion, and reinforcement) simultaneously because, by definition, activation and conversion involve change in vote choice over time, whereas reinforcement involves no change in vote choice (but rather a change in the strength of candidate preference).

  18. 18

    In addition, the binary logit models drop respondents who stayed 0 values across two time points (e.g., a partisan who was undecided and then shifted her vote toward her own party's candidate would remain a 0 value in the model predicting change toward the opposing party).

  19. 19

    In addition, to the extent that there is little within-person change in program exposure over time, it would be difficult to detect much of an effect (Allison 2009).

  20. 20

    To investigate differential effects across groups, I replicated each of the models in Tables 1 and 2 twice: the first replication added interactions between each predictor and a dummy for political knowledge (high vs. low), and the second added interactions between each predictor and a dummy for partisan strength (strong partisans vs. weak/leaning partisans). Appendix D provides full details and results of these analyses. Note that, given the small sample sizes, caution should be exercised in drawing firm conclusions about subgroups. See Appendix B for details of question wording and variable coding for partisan strength and political knowledge.

  21. 21

    There was suggestive evidence that highly knowledgeable and strong partisans were influenced somewhat more strongly by like-minded news exposure than less informed and weaker partisans, but the differences fell short of statistical significance.

  22. 22

    One might wonder whether these lackluster findings for reinforcement are attributable to the minority of staunch defectors—the small group of partisans who consistently expressed a vote choice for the opposing party's candidate. Perhaps the reinforcement effect really only applies to those who resolutely support their own party's candidate. To examine this possibility, I re-ran the models, this time limiting the sample to constant voters who supported their own party's candidate. The conclusions are substantively the same. The only difference is that the previously marginal positive effect of stable like-minded news exposure on favorability toward one's preferred candidate becomes slightly stronger, passing the .05 threshold for significance.

  23. 23

    To examine this possibility, I replicated each of the models in Tables 3 and 4 twice—adding interactions between each predictor and dummies for political knowledge and partisan strength, respectively, as in the previous analyses of activation and conversion.

  24. 24

    Moreover, stable levels of exposure to news slanted toward the preferred candidate's party still exerted the odd positive effect on favorability toward the opponent across all the groups. Surprisingly, there was also evidence that, among the highly knowledgeable, stable levels of exposure to news slanted toward the opponent's party exhibited a negative effect on favorability toward the preferred candidate, which suggests an erosion effect on initial candidate preference. In this regard, the highly knowledgeable differed significantly from the less knowledgeable, who did not appear to exhibit this effect. See Appendix D for full results.

  25. 25

    In the models examining stable levels of exposure, interactions between Wave and exposure to newspapers, radio, the Internet, news magazines, and political television ads were used; in the models examining changing levels of exposure, variables capturing changing exposure to each type of media were used. See Appendix B for all question wording and variable coding.

  26. 26

    In particular, the positive effect of changing exposure to programs slanted toward citizens’ own party on wave 4–5 activation, as well as the positive effect of stable exposure to programs slanted toward the preferred candidate's party on change in favorability toward that candidate, faded to nonsignificance.

  27. 27

    Of course, neutral news could also provide exposure to diverse perspectives.

References

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information
  • Achen, Christopher H. 2001. “Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Independent Variables.” Prepared for the Annual Meeting of the Political Methodology Section of the American Political Science Association, UCLA, Los Angeles.
  • Adams, William J. 2000. “How People Watch Television as Investigated Using Focus Group Techniques.”Journal of Broadcasting & Electronic Media 44(1): 7893.
  • Allison, Paul D. 1990. “Change Scores as Dependent Variables in Regression Analysis.” Sociological Methodology 20: 93114.
  • Allison, Paul D. 2009. Fixed Effects Regression Models. Los Angeles, CA: Sage.
  • Bai, Matt. 2008. “Working for the Working-Class Vote.” New York Times Magazine, October 15. http://www.nytimes.com/2008/10/19/magazine/19obama-t.html?_r=1&ref=magazine&oref=slogin&pagewanted=print (accessed January 30, 2012).
  • Barker, David C. 2002. Rushed to Judgment: Talk Radio, Persuasion, and American Political Behavior. New York: Columbia University Press.
  • Barker, David C., and Susan B. Hansen. 2005. “All Things Considered: Systematic Cognitive Processing and Electoral Decision-Making.” Journal of Politics 67(2): 31944.
  • Bartels, Larry M. 1993. “Messages Received: The Political Impact of Media Exposure.” American Political Science Review 87(2): 26785.
  • Baum, Matthew A., and Tim Groeling. 2009. “Shot by the Messenger: Partisan Cues and Public Opinion Regarding National Security and War.” Political Behavior 31(2): 15786.
  • Baum, Matthew A., and Phil Gussin. 2007. “In the Eye of the Beholder: How Information Shortcuts Shape Individual Perceptions of Bias in the Media.” Quarterly Journal of Political Science 3(1): 131.
  • Beck, Paul A., Russell J. Dalton, Steven Greene, and Robert Huckfeldt. 2002. “The Social Calculus of Voting: Interpersonal, Media, and Organizational Influences on Presidential Choices.” American Political Science Review 96(1): 5773.
  • Bennett, W. Lance, and Shanto Iyengar. 2008. “A New Era of Minimal Effects? The Changing Foundations of Political Communication.” Journal of Communication 58(4): 70731.
  • Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press.
  • Brader, Ted. 2005. “Striking a Responsive Chord: How Political Ads Motivate and Persuade Voters by Appealing to Emotions.” American Journal of Political Science 49(2): 388405.
  • Chang, LinChiat, and Jon A. Krosnick. 2009. “National Surveys via RDD Telephone Interviewing versus the Internet: Comparing Sample Representativeness and Response Quality.” Public Opinion Quarterly 73(4): 64178.
  • Chong, Dennis, and James N. Druckman. 2007. “A Theory of Framing and Opinion Formation in Competitive Elite Environments.” Journal of Communication 57(1): 99118.
  • Cialdini, Robert B., Melanie R. Trost, and Jason T. Newsom. 1995. “Preference for Consistency: The Development of a Valid Measure and the Discovery of Surprising Behavioral Implications.” Journal of Personality and Social Psychology 69(2): 31828.
  • CNN. 2009. “Interview with White House Communications Director; Obama Wins Nobel Peace Prize.” Reliable Sources, October 11. http://transcripts.cnn.com/TRANSCRIPTS/0910/11/rs.01.html (accessed April 7, 2012).
  • Converse, Phillip E. 1964. “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, ed. David E. Apter. New York: Free Press, 20661.
  • D'Alessio, David, and Mike Allen. 2000. “Media Bias in Presidential Elections: A Meta-Analysis.” Journal of Communication 50(4): 13356.
  • Dalton, Russell J., Paul A. Beck, and Robert Huckfeldt. 1998. “Partisan Cues and the Media: Information Flows in the 1992 Presidential Election.” American Political Science Review 92(1): 11126.
  • DellaVigna, Stephano, and Ethan Kaplan. 2007. “The Fox News Effect: Media Bias and Voting.” Quarterly Journal of Economics 122(3): 11871234.
  • Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know about Politics and Why It Matters. New Haven, CT: Yale University Press.
  • Dilliplane, Susanna. 2011. “All the News You Want to Hear: The Impact of Partisan News Exposure on Political Participation.” Public Opinion Quarterly 75(2): 287316.
  • Dilliplane, Susanna, Seth K. Goldman, and Diana C. Mutz. 2013. “Televised Exposure to Politics: New Measures for a Fragmented Media Environment.” American Journal of Political Science 57(1): 23648.
  • Druckman, James N., and Michael Parkin. 2005. “The Impact of Media Bias: How Editorial Slant Affects Voters.” Journal of Politics 67(4): 103049.
  • Eagly, A. H., P. Kulesa, L. A. Brannon, K. Shaw, and S. Hutson-Comeaux. 2000. “Why Counterattitudinal Messages Are as Memorable as Proattitudinal Messages: The Importance of Active Defense Against Attack.” Personality and Social Psychology Bulletin 26(11): 13921408.
  • Feldman, Lauren. 2011. “The Opinion Factor: The Effects of Opinionated News on Information Processing and Attitude Change.” Political Communication 28(2): 16381.
  • Finkel, Steven E. 1993. “Reexamining the ‘Minimal Effects’ Model in Recent Presidential Campaigns.” Journal of Politics 55(1): 121.
  • Garrett, R. Kelly. 2009. “Politically Motivated Reinforcement Seeking: Reframing the Selective Exposure Debate.” Journal of Communication 59(4): 67699.
  • Gasper, John T. 2007. “Ideological Shift: Explaining the Liberal Media Myth.” Working paper. http://www.andrew.cmu.edu/user/gasper/research.html.
  • Gentzkow, Matthew, and Jesse M. Shapiro. 2010. “Ideological Segregation Online and Offline.” Working Paper No. 55. Initiative on Global Markets. Chicago: University of Chicago.
  • Gerber, Alan S., Dean Karlan, and Daniel Bergan. 2009. “Does the Media Matter? A Field Experiment Measuring the Effect of Newspapers on Voting Behavior and Political Opinions.” American Economic Journal: Applied Economics 1(2): 3552.
  • Groeling, Tim. 2008. “Who's the Fairest of Them All? An Empirical Test for Partisan Bias on ABC, CBS, NBC, and Fox News.” Presidential Studies Quarterly 38(4): 63157.
  • Groseclose, Tim, and Jeff Milyo. 2005. “A Measure of Media Bias.” Quarterly Journal of Economics 120(4): 11911237.
  • Hart, William, Dolores Albarracin, Alice H. Eagly, Inge Brechan, Matthew J. Lindberg, and Lisa Merrill. 2009. “Feeling Validated versus Being Correct: A Meta-Analysis of Selective Exposure to Information.” Psychological Bulletin 124(3): 55588.
  • Heise, David R. 1969. “Separating Reliability and Stability in Test-Retest Correlation.” American Sociological Review 34(1): 93101.
  • Hillygus, D. Sunshine, and Simon Jackman. 2003. “Voter Decision Making in Election 2000: Campaign Effects, Partisan Activation, and the Clinton Legacy.” American Journal of Political Science 47(4): 58396.
  • Hillygus, D. Sunshine, and Todd G. Shields. 2008. The Persuadable Voter: Wedge Issues in Presidential Campaigns. Princeton, NJ: Princeton University Press.
  • Iyengar, Shanto, and Kyu S. Hahn. 2009. “Red Media, Blue Media: Evidence of Ideological Selectivity in Media Use.” Journal of Communication 59(1): 1939.
  • Jamieson, Kathleen H., and Joseph N. Cappella. 2008. Echo Chamber: Rush Limbaugh and the Conservative Media Establishment. Oxford: Oxford University Press.
  • Jones, David A. 2002. “The Polarizing Effect of New Media Messages.” International Journal of Public Opinion Research 14(2): 15874.
  • Kahn, Kim F., and Patrick J. Kenney. 2002. “The Slant of the News: How Editorial Endorsements Influence Campaign Coverage and Citizens’ Views of Candidates.” American Political Science Review 96(2): 38194.
  • Kurtz, Howard. 2008. “MSNBC Drops Olbermann, Matthews as News Anchors.” Washington Post, September 8. http://www.washingtonpost.com/wp-dyn/content/article/2008/09/08/AR2008090800008.html (accessed July 18, 2011).
  • Ladd, Jonathan M., and Gabriel S. Lenz. 2009. “Exploiting a Rare Communication Shift to Document the Persuasive Power of the News Media.” American Journal of Political Science 53(2): 394410.
  • Lazarsfeld, Paul F., Bernard Berelson, and Hazel Gaudet. 1948. The People's Choice: How the Voter Makes Up His Mind in a Presidential Campaign. New York: Columbia University Press.
  • Lee, Gangheong, and Joseph N. Cappella. 2001. “The Effects of Political Talk Radio on Political Attitude Formation: Exposure versus Knowledge.” Political Communication 18(4): 36994.
  • Marcus, George E., W. Russell Neuman, and Michael MacKuen. 2000. Affective Intelligence and Political Judgment. Chicago: University of Chicago Press.
  • McGerr, Michael. 1986. The Decline of Popular Politics. New York: Oxford University Press.
  • McGuire, William J. 1986. “The Myth of Massive Media Impact: Savagings and Salvagings.” In Public Communication Behavior (Vol. 1), ed. G. Comstock. New York: Academic Press, 173257.
  • Owen, Diana. 1997. “Talk Radio and Evaluations of President Clinton.” Political Communication 14(3): 33353.
  • Patterson, Thomas E., and Robert D. McClure. 1976. The Unseeing Eye: The Myth of Television Power in National Politics. New York: Putnam.
  • Petty, R. E., and J. T. Cacioppo. 1996. Attitudes and Persuasion: Classic and Contemporary Approaches. Boulder, CO: Westview Press.
  • Pew Research Center for the People & the Press. 2010. “Americans Spending More Time Following the News.” September 10. http://people-press.org/2010/09/12/americans-spending-more-time-following-the-news/.
  • Prior, Markus. 2007. Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. Cambridge: Cambridge University Press.
  • Prior, Markus. 2009. “Improving Media Effects Research through Better Measurement of News Exposure.” Journal of Politics 71(3): 893908.
  • Project for Excellence in Journalism (PEJ). 2008. The Color of News. Washington, DC: Pew Research Center for the People & the Press.
  • Rosenstein, Aviva W., and August E. Grant. 1997. “Reconceptualizing the Role of Habit: A New Model of Television Audience Activity.” Journal of Broadcasting and Electronic Media 41(3): 32444.
  • Schudson, Michael. 1995. The Power of News. Cambridge, MA: Harvard University Press.
  • Stroud, Natalie J. 2010. “Polarization and Partisan Selective Exposure.” Journal of Communication 60(3): 55676.
  • Stroud, Natalie J. 2011. Niche News: The Politics of News Choice. Oxford: Oxford University Press.
  • Sunstein, Cass R. 2007. Republic.com 2.0. Princeton, NJ: Princeton University Press.
  • Taber, Charles S., and Milton Lodge. 2006. “Motivated Skepticism in the Evaluation of Political Beliefs.” American Journal of Political Science 50(3): 75569.
  • Valentino, N. A., V. L. Hutchings, and I. K. White. 2002. “Cues That Matter: How Political Ads Prime Racial Attitudes during Campaigns.” American Political Science Review 96(1): 7590.
  • Yanovitzky, Itzhak, and Joseph N. Cappella. 1999. “Effect of Call-In Political Talk Radio Shows on Their Audiences: Evidence from a Multi-Wave Panel Analysis.” International Journal of Public Opinion Research 13(4): 37797.
  • Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.
  • Zaller, John. 1996. “The Myth of Massive Media Impact Revived: New Support for a Discredited Idea.” In Political Persuasion and Attitude Change, ed. Diana C. Mutz, Paul M. Sniderman, and Richard A. Brody. Ann Arbor: University of Michigan Press, 1779.

Biography

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information
  • Susanna Dilliplane is a George Gerbner Postdoctoral Fellow at the Annenberg School for Communication at the University of Pennsylvania, 3620 Walnut Street, Philadelphia, PA 19104

Supporting Information

  1. Top of page
  2. Abstract
  3. The Theoretical Basis for Partisan Media Effects
  4. Inconclusive Empirical Findings
  5. The Present Study
  6. Method
  7. Results
  8. Discussion
  9. References
  10. Biography
  11. Supporting Information

Disclaimer: Supplementary materials have been peer-reviewed but not copyedited.

FilenameFormatSizeDescription
ajps12046-sup-0001-Suppmat.pdf844K

Appendix A: Procedure for Coding the Partisan Slant of TV News Programs in NAES

Appendix B: Question Wording and Variable Coding

Appendix C: Supplementary Analyses

Appendix D: Fixed Effects Regression Models for Analysis of Moderators

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.