Implicated as potential cause of:

• Ideological polarization (Levendusky, 2013)
• Affective polarization (Garrett et al., 2014)
• Participation (Dilliplane, 2011; Wojcieszak, Bimber, Feldman, & Stroud, 2016)
• Perceptions of media bias (Arceneaux, Johnson, & Murphy, 2012)
• Uncivil political discussion (Gervais, 2014)

• Selective approach (I choose my side's programming)
  • Seems more prevalent than active avoidance of the other side (Garrett, 2009)
• But what causes it?
  • Partisanship (Iyengar & Hahn, 2009; Johnson, Zhang, & Bichard, 2011; Stroud, 2008; Wicks, Wicks, & Morimoto, 2014)
  • Other leads: Emotional motivations (e.g., Valentino, Banks, Hutchings, & Davis, 2009)
• Let's take a closer look at partisanship

• Party affiliation as a group membership rather than symbolic statement of issue positions or rational choice (Huddy, Mason, & Aarøe, 2015; Iyengar, Sood, & Lelkes, 2012)
• We can apply social identity theory to partisanship (Green, Palmquist, & Shickler, 2004)

• Media are used to maintain social identities
  • Only use as much media as needed for homeostasis
  • Different identities may require different media sources
• Threatened identity = more identity-relevant media use
  • Necessary maintenance
  • Threat removed = less identity-relevant media use

• Media use is a response to the environment rather than perpetually reinforcing
  • Motivation is only for maintenance
  • Competing identities and diverse systems prevent positive feedback loops

• Numerical distinctiveness
  • Being outnumbered raises identity salience and affiliation (Bettencourt, Miller, & Hume, 1999; McGuire, McGuire, Child, & Fujioka, 1978)
• Threats to value (Branscombe, Ellemers, Spears, & Doosje, 1999)
  • Questioning the merit of the group's practices or morality
• Generally, stronger identification = more entrenchment in the face of threat

• Red states and blue states
  • Urban-rural divide
  • E.g., are you outnumbered politically in your local area?
• Questioning the beliefs and moral status of opposing parties is normatively acceptable
  • E.g., discussions with those who disagree

H1a: Being a member of the losing party locally will be associated with greater partisan media selectivity.

H1b: Discussing politics with members of opposing parties will be associated with greater partisan media selectivity.

H2a: The effect of being a member of the losing party locally will be amplified by stronger partisan identity.

H2b: The effect of discussing politics with members of opposing parties will be amplified by stronger partisan identity.

• N = 1000 sample from YouGov
  • 800 general population sample, 200 black oversample
• Party ID measured on 7-point scale from two questions
• Do you identify as…Republican, Democrat, or Independent
• If Independent, asked if lean toward one or other party

• If one of the parties, asked if "strong" or "not strong"

• Those who provided no party or party lean were excluded

• Used program list technique (Dilliplane, Goldman, & Mutz, 2013) to measure media use
  • 67 sources from TV, radio, and websites
  • Yes or no response to watch/listen to/visit source "regularly"
• Sources categorized as left-leaning, right-leaning, or non-partisan
  • 5 political comm. researchers rated each source with option to say "don't know"
  • If majority agreed with identification, then it is used
  • Otherwise, used Lexis-Nexis technique described in Dilliplane (2011)

• Respondent ZIP code -> county
  • County-level identity threat = average presidential margin of loss/victory in 2008 and 2012 elections
  • Threat = R margin of victory for Ds, D margin of victory for Rs
• What portion of the people you talk politics with are Republicans, Democrats, members of other parties, and non-party identifiers?
  • Forced sum to 100%
  • Network threat = % out-parties

Linear regression predicting % of in-party partisan sources out of total number

Model 1: H1a, H1b

Model 2: H2a (county by identity interaction)

Model 3: H2b (discussion by identity interaction)

Focal predictors:

• County-level identity threat (out-party POTUS vote share)
• Network-level identity threat (% out-party discussants)
• Strength of political identity constructed measure
  • 0 = nonpartisan, 1 = lean/not strong partisan, 2 = strong partisan
  • 0 = moderate, 1 = liberal/conservative, 2 = very liberal/conservative
  • Very liberal strong Democrat/very conservative strong Republican = 4
  • Moderate not strong partisan = 1

Controls:

• Party (1 = Republican)
• Indicator for whether R discusses politics with others
• Number of people with whom R discusses politics with
• Whether any media sources used
• Number of media sources used
• Age, race (black, Hispanic), gender, education, Christian religion

Model 1

H1a (county-level threat): Supported

H1b (network-level threat): Not supported

H2a (county threat by identity strength interaction): Supported

H2b (network threat by identity strength interaction): Supported

Overall:

• Heterogeneous networks associated with more selectivity
  • Especially for stronger partisans
• RSM is a theory invested in over-time effects
  • Those aren't measured here