Dewey beat Truman again

thanks to wikipedia

Everyone who has ever had anything to do with the social research or polling industry will have heard the story about how all the opinion polls were so positive that Thomas Dewey was going to defeat Harry Truman in the 1948 US presidential election that the Chicago Tribune declared Dewey the winner the morning after the election.   The reason this happened, say opinion pollsters, is that the polls predicting a republican victory were done by telephone, and failed to account for democratic voters being less likely to  have telephones than republicans, telephones still being a luxury in 1948.   This is not necessarily a problem if  pollsters know that say, 90% of the  republican voters have telephones vs 80% of the population that vote democrat.  If this is known, mathematical adjustments are applied, and the poll becomes more accurate.  Because the results were so embarrassing, they figured out what they did wrong, and opinion polling has been fairly accurate up till now.

Alberta just had a provincial election, the incumbent party, the Progressive Conservatives, have dominated the Alberta legislature with almost no opposition representation since 1971.   Without going into the details of Alberta politics, this election was shaping up to be more exciting than the usual PC steamroller.  Wildrose, a right wing populist Alberta version of the tea party was, according to opinion polls, going to drive the comparatively center right PCs out of office.  As in the 1948 US election, the opinion polls got it spectacularly wrong, the PCs are back with a healthy majority.

Based on what this left wing redneck observed, the polls were wrong for the same reason they were supposed to be wrong in 1948, bad sampling.  As the election campaign heated up, I was getting on average one or two robo call a day, 'Press one if you are going to vote for ___, 2 if you are....' etc..  The problem was that I don't answer my land line phone any more, their surveys of my opinion were being left as recorded messages in the in box of my land line.   The only phone I answer is my cell phone.  I did not get one call from an opinion poll on my cell phone, never have, election or no.  The only reason I still have a land line is that it provides me with a phone number I can give out to anyone who I don't want to talk to, like all those stores who want to know your phone number before they will sell you something.  Thanks to privacy laws, my policy of only giving out my cell number to friends and relatives, and that cell phone numbers are fairly well protected from unsolicited calls by cell phone providers, I don't get random calls on my cell phone.  Many people don't even have land lines any more, and the younger they are, the less likely they are to have a tethered phone.  So opinion polls conducted by phoning land lines are more likely to succeed in polling the elderly, technologically illiterate (don't have call display or messaging for screening calls), and rural areas with poor cell phone coverage, while missing the younger, more technologically hip and urban primarily cell phone using population.  If the wireless phone user will vote differently from the wired phone user, polls that exclusively survey land line phones will be wrong.  Which is pretty much how it went in Alberta.

The cities passed on the Alberta Wildrose, who for all their progressive talk, were clearly advocating the values of a past era.   The Wildrose party did elect 17 of the  87 available members to the legislature, from the rural strongholds of traditional social conservative values and apparently the only telephones still connected to polls.

It will be interesting to see if the US opinion polls will get it right in the 2012 elections.


Comments

Popular posts from this blog

Canadian police need more diversity

The deadly gas stove

Artificial Intelligence revisited