Attitudes, Non-Attitudes, & Constraint

POLSCI 240 / PSY 225: Political Psychology

January 27, 2025

Questions for this section

What is the nature of public opinion?

  • How should people organize their opinions about politics? How do they?
  • Do they even have opinions?
  • How do people answer questions on surveys? What does it mean for public opinion?
  • How do core values and traits lend structure to opinions?

Belief systems

“…a belief system [is a] configuration of ideas and attitudes in which the elements are bound together by some form of constraint or functional interdependence.”

“In the static case, ‘constraint’ [means] the success we would have in predicting, given initial knowledge that an individual holds a specified attitude, that he hold certain further ideas and attitudes.”

“In the dynamic case, ‘constraint’ [means] the probability that a change in the perceived status (truth, desirability, and so forth) of one idea-element would psychologically require, from the point of view of the actor, some compensating change(s) in the status of idea-elements elsewhere in the configuration.”

Converse (1964), “The Nature of Belief Systems in Mass Publics”

Left vs right in U.S.

Source

Networked beliefs

Source

The value of structured beliefs (?)

What are some reasons why we might think it desirable that citizens have structured (constrained) beliefs

  • Derived from 1st principles

    • Reason from basic values (e.g., equality) to specific opinions (e.g., taxes)

  • Similarity to, and understanding of, “elite” ideology (e.g., liberal vs conservative)

    • Easier for everyone to talk about politics

  • Beliefs fit together, are logically consistent

    • Changes in one belief lead to changes in related ones

Converse


Converse’s findings

  1. Most people don’t understand well the terms “liberal” and “conservative”
  2. When people talk about likes and dislikes of parties & candidates, rarely talk about ideology or even policy

Levels of conceptualization

Converse’s findings

  1. Most people don’t understand well the terms “liberal” and “conservative”
  2. When people talk about likes and dislikes of parties & candidates, rarely talk about ideology or even policy
  3. Expected correlations among issue attitudes are small (weakly related, e.g., abortion and taxes)

But…

Why should people hold opinions that look like partisan elites?

Converse’s findings

  1. Most people don’t understand well the terms “liberal” and “conservative”
  2. When people talk about likes and dislikes of parties & candidates, rarely talk about ideology or even policy
  3. Expected correlations among issue attitudes are small (weakly related, e.g., abortion and taxes)
  4. Issue attitudes are unstable over time (except for party identification)

Instability

Looking closer at opinion change

  • We can predict opinions at time 3 just as well from time 1 as time 2
  • Why is this kind of weird?

A series with “memory”

If change is systematic, opinions more correlated when closer in time

The “black and white” model of public opinion

What kind of model could explain the weird pattern in Converse’s data?

  • A small percent of public with perfectly stable attitudes

  • A much larger percent with random responses to survey questions (“non-attitudes”)

    • Random responses mean that there is no correlation across time (so equally correlated t1 -> t2 and t1 -> t3)
    • Perfect stability is correlation of 1 no matter what (so equally correlated t1 -> t2 and t1 -> t3)
    • When you combine the two subpopulations, you get exactly the correlation observed, where the size of the correlation between time points a function of the size of the perfectly stable subpopulation

Converse’s conclusions

  1. Few people are understand and use “elite” ideology to navigate politics
  2. Most people use social groups to understand political conflict
  3. Political opinions are not well “organized”, constrained
  4. People change their opinions, a lot, over time
  5. Nature of instability suggests most people don’t even have political opinions

Much of political psychology in the subsequent years is a response, in one way or another, to these conclusions

An alternative view: measurement error

Classic measurement theory

Classic measurement theory says that a measure of something is the sum of the “true score” plus “random error”

\[ \begin{align} \text{measured score} &= \text{true score} + \text{error} \\ x &= \tau + e \end{align} \]

  • For example, get your height measured every day for 100 days
  • What do you think you would see?

Measuring height with error

Let’s say you are 6ft tall, and measurements of your height are typically accurate to about 1/8 of an inch:

Implications

Height is not changing (perfectly stable), but measured height changes randomly across time

  • When we look at many people across time, we end up with an imperfect correlation between time points
  • People tall at time 1 are also tall at time 2 and 3, but error creates some changes in ordering
  • Error is random and independent of time, so adjacent time points are no more correlated than those far away
  • The size of the correlation is smaller when measurement error is larger

Implications for politics

If same is true of political opinions, measurement error can account for the pattern found by Converse!

  • People have stable underlying opinions, measured with (lots of) random error
  • Some correlation across time (but weak if error is high)
  • Correlations of adjacent time points no more correlated than non-adjacent

Thus, an alternative view of Converse’s data: people have roughly stable opinions, but these are measured with lots of error

Summary

We have empirical facts about democratic publics (true for the most part, even in contemporary politics)

  • Most people’s opinions are weakly correlated across issues
  • Most people’s opinions are unstable across time

We have two different models (theories) that can account for these facts

  • Converse’s “black and white” model (most have “non-attitudes”)

    • But do most people really have completely random opinions? Seems hard to believe…

  • Measurement error: hard to translate (stable) opinions into survey responses, so lots of random error

    • But how hard is it really to answer a survey question?
    • And why are constraint and stability higher for people who pay attention to politics?

BREAK

How do people answer questions?

Take a few seconds and answer the following questions - keep answers to yourself, but really try, as if you were taking a survey

  1. Do you approve, disapprove, or neither approve nor disapprove of the job Joe Biden did as President?

  2. Do you think tariffs on imported goods from China to the United States should be higher than they are now, lower, or should they stay the same?

  • Now reflect on how you answered those questions? What was the process like? What did you do?

Two general models for thinking about question-answering

Simple recall

  • Opinions are stored in long-term memory
  • When asked a question, you recall the opinion into working memory
  • You do your best to map the recalled opinion to the answer choices

Constructionist models

  • When asked a question, you recall relevant “considerations” into working memory
  • You combine the considerations somehow (e.g., add them up) into an overall opinion
  • You do your best to map the recalled opinion to the answer choices

Example of constructionist

Imagine you are asked whether you approve or disapprove of President Trump, one time each day, for 100 days

  • You have 10 (relevant) considerations in long-term memory
  • Each is \(+1\), \(0\), or \(-1\) for Trump
  • Let’s say you have the following: \((0,1,1,-1,0,-1,-1,-1,0,1)\)
  • Each time you are asked, you randomly “sample” 3 considerations, add them up, and report “approve”, “disapprove”, or “neither approve nor disapprove”

   approve disapprove    neither 
        22         50         28

Ambivalence

We might expect that most people are ambivalent when it comes to most political issues and objects

  • They have, stored in long-term memory, both “positive” and “negative” considerations relevant to evaluating an object

  • When they construct their opinion, they get a random sample

  • So their opinions are likely to be:

    • Moderate on average
    • Not highly correlated across opinions
    • Highly variable from one time to another
    • If considerations are stable, adjacent time points no more correlated

What comes to mind?

When you think about whether you approve or disapprove of Trump, what comes to mind and why?

  • There are things that are always very likely to come to mind (e.g., “Republican”, “President”)
  • There are things that are always very unlikely to come to mind (e.g., Atlantic City)
  • There are things that are likely to come to mind at this moment because you have recently seen or thought about them (e.g., Colombia)

What’s going on here more generally?

Accessibility

Accessibility is how likely it is that something in long-term memory gets brought into working memory

  • Baseline accessibility: general, stable tendency to come to mind due to prevalence in environment
  • Temporary (or “spreading”) accessibility: short-term, due to connections with recently thought-about ideas

But where do considerations come from?

And do people just take in whatever considerations they come across?

Zaller’s RAS model

Receive, Accept, Sample

  • Partisan political elites generate messages and disseminate them to public through media

  • Politically attentive people are more likely to receive political messages

  • Politically attentive people are less likely to accept messages that conflict with their political values and identities (e.g., partisanship)

  • Political opinions are the average of sampled considerations from long-term memory

    • The probability a consideration is sampled is determined by accessibility (baseline + temporary)

Implications

Public opinion is highly responsive to “context”

  1. Agenda-setting: the issues that get coverage have disproportionate influence on public opinion

  2. Priming: potentially irrelevant aspects of political coverage can alter public opinion through spreading activation

  3. Framing: highlighting certain aspects of issues over others can change public opinion

  4. Partisan polarization: public opinion divides by partisanship when elite partisans are divided, and polarization is larger for more attentive citizens

Examples: agenda setting

Iyengar and Simon (1993)

Examples: priming

Source

Examples: framing

Nelson et al. (1997)

Examples: partisan polarization

Berinsky (2007)