Borrowed from physics

• In physics, polarization is the orientation of a wave
• The original meaning is alignment, orientation, separation
• Nothing about conflict or hostility. Social science supplied that.

A family of concepts

• Broadest usage: alignment, separation, sorting, divergence
• At least four mathematically distinct things share the name
• They can move in opposite directions in the same data

Three further axes

• Bipolar vs multipolar: nothing requires two camps
• State vs process: a snapshot vs a trend
• Object vs target: who is polarized against whom?

Figure 3: Bipolar and multipolar configurations. On the left, two camps on one dimension. On the right, several camps spread over several dimensions.

Elite versus mass

• Elites (legislators, activists): divergence since the 1970s is not disputed (Fiorina et al. 2008)
• The mass public: does the electorate follow its elites?
• Two further distinctions sit inside that question
  • Ideological divergence: people move apart on issues, towards the extremes
  • Issue alignment: positions bundle together, and with party, without anyone moving

Is the public polarized?

• After 2000 and 2004: red states vs blue states, a “culture war” narrative
• Culture War? The Myth of a Polarized America (Fiorina et al. 2006): the public is closely but not deeply divided
• Journal of Politics 70 (2), 2008, critique and reply back to back
  • Abramowitz and Saunders (2008): “Is polarization a myth?”
  • Fiorina et al. (2008): “Misconceptions and misreadings”
• Same data, same decades, opposite verdicts

Introduction

• Puzzle: Fiorina says today’s voters look like the barely ideological public Converse (2006) described for the 1950s
• But much has changed since then
  • Share with some college education: 19% in 1956, 61% in 2004
  • Party elites have moved apart for decades
• Question: is mass polarization a myth?
• They test five of Fiorina’s claims: moderation, partisan and geographic polarization, social cleavages, and voter engagement

Theory

• Education and elite cues: a more educated public, watching polarized elites, should think more ideologically
• The engaged: polarization should be deepest among the interested, informed and active, “whose opinions matter most to candidates and officeholders”
• Mobilization: polarization raises the stakes of an election, so it should energize voters, not turn them off

Research design

• Data: American National Election Studies 1972 to 2004, 2004 National Exit Poll
• Ideological consistency: seven issues, each collapsed to liberal, moderate or conservative
  • Score = |number of liberal positions minus number of conservative positions|
  • A 16-item scale for 2004, recoded to 11 points
• Engagement: interest, knowledge and participation, split into thirds
• Partisan polarization: correlation of party identification with ideology over time
• Geography and religion: red vs blue state voters, logit of the Bush vote on religiosity

Results: consistency

• The average correlation between the seven items rises from .20 to .26 to .32

Results: the engaged

• Same mean, nearly double the spread. The engaged third is 37% of respondents and close to half of all voters.

Results: sorting

• Among the engaged the same correlation runs from .47 to .77, and the gap between the parties’ mean ideology doubles

Results: geography

Results: other evidence

• Religion: among white voters, religiosity predicts the Bush vote better than any other social characteristic
  • P(Bush vote) .34 for the least religious, .81 for the most religious
• Engagement in 2004: turnout 54% to 61%, persuasion attempts 32% to 48%
  • Voters feeling strongly about Bush, either way, were the most likely to vote and campaign

Conclusion

• Polarization is not just an elite phenomenon
• The centre is held by the least interested, informed and active
• Elite divisions “reflect real divisions within the American electorate”
• Polarization energizes the electorate rather than turning it off

“We have met the enemy and he is us” (Abramowitz and Saunders 2008, quoting Pogo)

Introduction

• A point-by-point reply to the five criticisms
• Before the reply, a rule about evidence:
  • Votes, election returns and approval ratings “can not be used as evidence of polarization, for or against”
• Everything else follows from that rule and from one definition

Theory

• Polarization = movement from the centre towards the extremes, so the test is rising bimodality
• Sorting is something else: party groups grow more distinct while the overall distribution stays put
• Centrist voters register polarized choices when the candidates polarize, without moving themselves
• So a changed relationship in the data may sit on the candidates’ side, not the voters’

Research design

• Raw distributions instead of recoded and aggregated indices
• Ideological self-placement in three surveys: NES, GSS and Gallup, 1970s to 2000s
• The same NES issue scales as A&S, compared position by position, 1984 vs 2004
• Pew Research Center: party differences on 40 items, 1987 to 2007
• Abortion positions of strong partisans in 2004

Results: raw data

• Only jobs and standard of living loses its centre to both extremes

Results: limited sorting

• Party sorting “certainly has occurred”, but far less than among elites

Results: the rest

• Abortion, 2004: a third of strong Democrats and two fifths of strong Republicans break with their party’s position
• Red vs blue states: majorities disagree on one NES issue, homosexual adoption
• 2004 turnout: the parties’ ground game rather than polarization, with party contacts up 8 points
• Religion vs class: A&S attribute a claim to them that they say they never made

Conclusion

• Without “multiple recodings and aggregations”, the A&S case “disappears”
• Sorting yes, polarization no
• Red-blue divides are a “misleading exaggeration”
• 2004 engagement rests on one election and ignores mobilization
• “No reason to revise the conclusions of Culture War?”

Two definitions

“adopting the common sense notion of polarization as a movement from the center toward the extremes, one searches in vain for evidence of increasing bimodality” (Fiorina et al. 2008)

Pair task

For each claim, note what evidence each side uses and which concept from Figure 2 or the three axes it rests on.

1. “The public is more polarized than in the 1980s.”
2. “Democrats and Republicans are further apart.”
3. “The public is polarized where it counts.”
4. “Voters are divided: look at the red and blue map.”
5. “The data show it.”

Then decide: is the disagreement about a definition, a measurement or a fact?

The mapping

Three kinds of dispute

• Mostly definitional: both accept that party sorting happened. They disagree about whether to call it polarization.
• Partly measurement: coding “don’t knows” as moderates and aggregating items into an index change the answer
• A little substantive: was 2004 turnout driven by polarization or by party mobilization? Does religion outweigh class?

Questions for the group

• If both sides accept that sorting happened, what exactly is left to disagree about?
• Is a public polarized if its engaged third is? Whose opinions should count?
• If votes cannot show polarization, what evidence could?
• Does any of this travel to a multiparty system like Germany’s?

Recap: takeaways

• Polarization is a family of concepts: dispersion, bimodality, sorting, distance, plus three axes
• Elites diverged without dispute. The contested question is the mass public.
• A&S find rising consistency and sorting, deepest among the engaged
• FAP find a stable middle and call the rest sorting
• Most of the dispute is about which concept and which population, some about coding, little about facts

Recap: learning goals

Goal 1. Elite vs mass, divergence vs alignment

• Elite polarization concerns office holders and activists, mass polarization the electorate
• Divergence means moving towards the extremes. Alignment means positions bundling together, and with party.

Goal 2. Disagreeing without disagreeing about a fact

• A&S and FAP largely agree on the facts, but attach the word to different concepts and populations

Goal 3. Which concept a claim depends on

• Ask what distribution changed, in whom, and whether it is positions or choices

Before you go

• The term paper is empirical: it confronts a claim with data you analyse yourself
• It is two things: the written paper, and the code that reproduces its analysis
• Any technology, R, Python, Stata or anything else. Nothing is prescribed.
• What is required is that a result comes out of a script that runs again, not out of clicks nobody can retrace
• The detail is in the syllabus