January 21, 2014

Understanding the Science of Elections

Print More


“Polling makes a great narrative for the media,” said Prof. Peter Enns, government. But who can really tell how accurate a poll is? Before elections, public opinion is meticulously measured in polls which are picked through in excruciating detail by everyone from journalists and newscasters to campaign managers.

But Enns and Brian Richman ’11 suggest that current polling methods fail to engage potential respondents enough to accurately predict certainty about elections and issues that influence their voting choices.

Individuals do not treat polls the same way they treat voting. According to Enns, psychological studies have shown that people often use a strategy when answering surveys called satisficing, where they only expend the minimal amount of effort to satisfy the questioner. This causes a mismatch in political poll responses between when elections are months away and when they are imminent.

According to Enns, when an election is several months away, voters are more likely to satisfice, so polling data will not be as accurate. Voters also tend to give different weighting to issues such as the economy when the election is in the distant future than when the election is days or weeks away. The most accurate weighting of each issue will be seen in polls just before election day.

Repeated polling can also affect the accuracy of polling results. Enns and Richman used data from the National Annenberg Election Survey (NAES) from the 2000 election to look at how voters’ poll answers change as they were polled repeatedly over the course of the campaign season.

Voters who seemed committed to answering polling questions accurately early in the campaign season appeared to become more apathetic over repeated polls as the season progressed, according to Enns.

When voters care less about the poll, they would give a less accurate impression of what issues were important to them and how they would eventually vote. The change in interest in polling correlates with the difference between poll answers and how voters actually voted when election day came. This supports Enns and Richman’s hypothesis that interest in the poll itself is related to accuracy of the polling data.

Overall, polling is not obsolete, according to Enns. Further research must be performed in order to encourage more accurate polling, especially early in campaigns. As campaign strategists can better understand how much weight voters place on the issues, they can more successfully target their efforts to influence public opinion. Enns plans to do further research in this area during the 2016 presidential election.