Attitudes and polling

Why polling in the 2020 election was inaccurate?

Polling in 2020 systematically underestimated support for Republicans because non Response Bias, Underweight less

educated, “Shy” Trump voters, Likely worsened by pandemic

How does sample size impact accuracy

Determines margin of error, it’s usually around 3%

How accurate are polls generally

• Average state polling error is 4%

• Usually very accurate, especially later on

• No clear biases

• National POTUS polls are about 3% and state is 5%

• Polls predict the winner about 75-80% of the time

Evidence of how accurate polls generally are

• 2012: predicted 50/50 state presidential outcomes

• 2016: predicted popular vote within 1pt, state polls in upper midwest off

• 2018: predicted House vote within tenth of a percent, winner 97% of the time, and 87% of all contested race

Social desirability bias

People will often tell you what they believe is socially acceptable rather than what is true

Problems with social desirability bias

• Difficult to get accurate opinion on major identity issues

• Bradley effect- Controversial candidates may have undercounted support

Meaningful attitudes

Beliefs about parties, candidates, groups and/or policies that feature:

• Stability: attitudes generally do not change over time

• Robustness: attitudes not fragile to superficial elements like question order and wording

Stability

Increases with interest in politics, stronger identification with a party, and greater knowledge of where parties stand on an issue

When public fails to consider tradeoffs

• Supporting cut offs but also the increase/decrease of things they want or need. Often under informed

• People have unstable and contradictory attitudes when forced to consider these objects as tradeoffs