AAPOR dissects New Hampshire polling failures
US-- An investigation into last year's New Hampshire presidential primary has blamed polling failures on the compressed primary calendar, an influx of first-time voters and underrepresentation of harder-to-reach groups who were likely to support Hillary Clinton.
Most pollsters put Barack Obama ahead in last January's New Hampshire Democratic primary, failing to predict Hillary Clinton's surprise win.
The study was carried out by a committee formed by the American Association for Public Opinion Research (AAPOR) and headed by Michael Traugott.
Among the reasons identified were:
• the unusually short timescale of last year's primaries, which meant pollsters may have been unable to capture late shifts in voter preference;
• the huge number of first-time voters, which made ‘likely voter' models less reliable;
• the fact that many groups who were likely to vote for Clinton, including union members and those with less education, were among the hardest respondents to reach, and were underrepresented.
AAPOR president Richard Kulka said the findings “raise significant questions for research on pre-election polling methods”.
The study also identified factors that did not appear to have influenced the polls, including the so-called ‘Bradley effect', whereby voters supposedly say they support a black candidate when polled, but vote for a white candidate in the privacy of the polling booth. “We found no evidence that white respondents overrepresented their support for Obama,” said Traugott.
The investigation also said the exclusion of cell-only individuals from surveys did not appear to have had influenced the results.
Author: Robert Bain