About

Populus
Northburgh House
10 Northburgh Street
London EC1V 0AT

Tel: +44 [0]20 7253 9900
Fax: +44 [0]20 7253 9911

How did the polls get it so wrong – again?

The major national polls in the US weren’t that far off the final result. Most predicted a Clinton win in the popular vote, and she looks set to narrowly clinch this.

What went really wrong were the slew of state polls produced throughout the campaign. In Wisconsin, for example, more than 80 polls were conducted and not a single one in the last few weeks put the eventual winner – Trump – ahead.

Trump’s unexpected wins in Wisconsin, and to a lesser degree Pennsylvania and Michigan, were a function of these state polling errors, which Populus had identified last week.

In state after state and poll after poll analysis, the 2012 results and 2016 estimates were so tightly aligned that it suggested the Republican and Democrat coalitions had barely changed from four years ago. The qualitative evidence suggested that this was almost certainly not the case.

Surely Donald Trump would be doing better with certain demographic groups, such as non-college educated whites, than Mitt Romney, and less well with other voting blocs such as those who lived in cities, and areas with high numbers of ethnic minorities? The difference between the results and the pre-election polls bore this thesis out.

Take non-college educated whites; America’s traditional non-graduate working class. This group has been a key part of the Republican coalition since 2000. In election after election, a clear linear relationship could be observed: the more non-college educated whites in a state, the higher the Republican vote.

In 2016 this relationship was still present, but now steeper – super charged if you will – as the number of white non-college educated voters increased, so Donald Trump’s vote share increased even faster than other Republican candidate’ vote share had before him. The inverse is also true for states and areas with fewer white non-college educated voters.

These polling errors, and the incorrect statistical relationships between voting habits and demographics they had imbedded within them, had real and devastating consequences for the strategic decisions of the Clinton campaign.

It led to a misreading of swing states by both Mrs Clinton’s aides and observers. States like Ohio and Iowa saw a swing to Trump that was much greater than expected. Midwestern states like Minnesota, Michigan, and Wisconsin, thought to be relatively safe for Clinton, turned out to be highly competitive, with Trump crucially taking Wisconsin and Michigan, and laying out a path to the presidency that state polls had suggested would be near impossible.


Back to previous page
Webflow to WordPress theme development by whois: Andy White