Anyone living in the United States who doesn’t know that tomorrow is “Super Tuesday” — the day when voters in more than 20 states will go to the polls to cast their votes for presidential candidates in their parties — has either been in a coma or is truly able to avoid all contact with the media. I for one will be glad when this first phase of the presidential selection process will be over, if only because it will mean a break from the drumbeat of political advertising and punditry.
All the chatter about politics has gotten me to thinking about one of the most ubiquitous features of modern political discourse–discussion of poll results. Is Obama gaining in California? Is McCain fading in Missouri? Poll results often come to dominate that discourse and, seemingly, to drive the decisions of candidates on everything from where to focus their energies (and cash) to what to talk about where. The history of modern political polling is fairly well known. George Gallup created the first of what we would call “scientific” samples for his national political polls in 1935 and achieved national recognition when his American Institute of Public Opinion correctly predicted the results of the 1936 presidential elections.
One sample from news reports about his polling is from the New York Times on August 30, 1936. In a story headlined Democratic Doubt Vanishes in the South, the paper reported, “Additional evidence of Mr. Roosevelt’s strength in the area is seen in the fact that the Southern States regularly return larger pro-Roosevelt majorities than those of any other section in the American Institute of Public Opinion polls. From Kentucky’s 61 per cent to Mississippi’s 91 per cent, the entire region seems to be solidly aligned in the Democratic Party once more.” Change the names (and the political affiliation of the American South) and these sentences could have been written today.
Of late, polling organizations such as Zogby, Harris, and others have experimented with a number of new methodologies, including interactive polls, while others have promoted the use of market mechanisms to predict political results.
In a world where search engine results become destiny, what if we were to use search queries to determine political popularity? GoogleTrends allows us to do just that. Do you want to know how the Democratic primary race in California is going? On the basis of search queries (I used hillary clinton, barack obama as my query), the race is neck and neck.
Is there a predictive value here? If we take a look at the Republican race in California, what we see is that volume of search queries would predict that Ron Paul is going to win the biggest prize on Super Tuesday–a result that even his most ardent supporters would have to admit is highly unlikely. However, this result does not mean that there is no predictive value. If the two marginal candidates (in California) are eliminated from the graph and we look at only the two front runners in the Republican race, what we see is a “search primary” that has narrowed to a dead heat, just like the Democratic case.
We’ll check these data again on Wednesday to see if there might be some predictive value for a “search primary”. If so, historians may well look back at this campaign as one where new kinds of data helped shape the campaigns. Or not. For now, all we can say is “stay tuned.”