The recent British election is the most recent example of the pre-election polls getting it wrong. The pollsters had David Cameron’s Tories in danger of losing to Labour, when in fact the Labour vote collapsed and Tories won a majority—something no one had predicted. In the BC election in May, Angus Reid had the Christy Clark Liberals losing by 9 points, but instead they won a majority government. Pollsters predicting the outcome of the 2012 US election failed to forecast the Republican takeover of both houses. A week before Ontario’s election last year Ekos, Ipsos Reid and Forum Research all had the Liberals and Conservatives in a dead heat, but the Liberals ended up winning a majority with an 8 point margin over the Tories.
There are other examples of polls gone awry, leading voters to ask if polling can be trusted anymore. No less an expert than Cliff Zukin the past president of the American Association for Public Opinion Research, and a Rutgers Political Science professor writing in the New York Times last month; says polls and pollsters are going to continue to be less reliable than in the past. Why? Technology has a lot to do with it, most importantly the explosive growth of computer and cell phone usage. In 2005 only about 6 percent of respondents indicated they had given up the landline phone in favour of the cell phone. In 2014 the number of cell-only respondents had increased to 43 percent with another 17 percent saying they “mostly” used cell phones. The problem in the US at least is that the use of automatic dialers to reach cell phones is illegal. Automatic diallers are necessary for pollsters to draw a true random sample at a cost that media outlets can afford. He points out that to reach 1,000 respondents it might be necessary to dial 20 times that number of potential respondents. Manually dialling 20,000 numbers puts the cost of accurate polling out of the reach of most organizations, so they begin to rely on less reliable methodologies like robo calls that largely only reach landline users. 97 percent of young people use the Internet but they only make up 13 percent of voters. Conversely 40 percent of those 65 or older don’t use computers but make up 22 percent of voters.
The other problem is that even if you reach a live prospect, increasingly they are unwilling to participate. You can thank telemarketing for peoples’ reluctance to have anything to do with strangers on the phone. In the 1970’s when people were flattered to be asked for their opinion, possibly even viewing participation as a civic duty; pollsters could get about an 80 percent response rate. By 1997 the response rate had declined to 36 percent and today response is down to 8 percent. Pollster Nate Silver told the Times, “The foundation of opinion research has historically been the ability to draw a random sample of the population. That’s become much harder to do.”
The other problem with political polling is that about a third of respondents are not truthful about their likelihood of voting. Not being able to accurately predict the actual voter turnout can seriously skew the results. Zukin strikes a pessimistic stance on the future prospects of political polling. He says there isn’t an immediate solution and that the polling industry is going to have to go through a period of experimentation in search of a solution. In the meantime he advises extreme caution in relying on any political polling in the near to medium term.