Polls Underestimate Support for Sanders in South Carolina

Bernie Sanders at a rally in New Orleans. Photo by Nick Solari

In May of last year, a few British friends and I gathered around a laptop to watch the UK election results come in. For weeks, the British press had been alight with doomsday predictions of a fractured parliament. According to the polls, the mainstream parties would be held hostage by upstart fringe groups and Scottish nationalists, who would either force Britain to exit the European Union or else break the United Kingdom apart.

The incumbent Conservative Party ended up winning in a landslide, an outcome so surprising, one leading member of parliament said he would eat his hat if the exit polls were true. No major British polling organization had predicted the outcome because of major flaws in polling methodology. Likely conservative voters were systematically excluded from surveys, and the mix of weights given to certain demographic groups over-represented the Labor Party.

In South Carolina, Hillary Clinton is reckoned to be leading Bernie Sanders by 22 points, according to a recent YouGov poll. Other polls have Sanders trailing by more than 50 points in the state. However, those numbers are probably wrong. Upon closer examination, these polls under-represent the demographic that is most likely to support Bernie Sanders as the Democratic nominee: young white Americans.

Polling methodology is based on the past behavior of groups of people, and young white South Carolinians have not been particularly active in past Democratic primaries. Most whites in the state vote Republican, while most blacks vote Democratic, and far fewer young people vote in elections of all kinds. So, it is reasonable to assume that the response of a middle aged black woman in South Carolina will be a better metric of Democratic voting behavior than that of a young white man in the state.

However, the historical data used to assign weights in polling surveys come from 2008, an anomalous year with unprecedented turnout in the Democratic primary. Young voters and African American voters came out in force to support Barack Obama's candidacy, and he won in the state's primary by 29 points with 295,000 votes. In comparison, John Edwards won the state's Democratic primary in 2004 with only 131,000 votes. Turnout is unlikely to reach 2008 levels this time around.

Polls give more weight to groups who voted in the last Democratic primary. The problem is, many young voters in 2008 have now graduated into an older age group, meaning few of the current batch of young voters participated in the last primary. In the YouGov poll, there is no data whatsoever on the responses of voters aged 18-29 for most questions about Democratic candidates. In other words, the poll takes no account of young people voting in the Democratic primary and attaches practically no weight to their responses.


Another issue is the weight given to African American voters. Indeed, in the YouGov poll, responses from black participants are given about 20-25% greater weight than responses from white participants for questions about Democratic candidates. Although most African Americans are said to be Clinton supporters, there is cause to doubt their enthusiasm.

There is little excitement in the Democratic camp for Hillary Clinton. Her campaign and the press largely portray her as an inevitable candidate, which leaves many Clinton supporters wondering why they should bother to show up on primary day. Meanwhile, the Sanders campaign has been energetically courting young people and African Americans, pushing them to vote in the primary.

Clinton has been less active in South Carolina than Sanders, visiting seven times to his ten. She also has no official Facebook page for the state, though there is one unofficial page with 187 'likes.' The Sanders Facebook page for South Carolina has over twelve thousand 'likes.'

It remains to be seen if Clinton's supposed momentum will defeat Sanders' frenzied campaigning and organization. But, if Clinton wins by as large a margin as projected, I will eat my hat.