We’ve been compiling tracking statistics on our sales of items in support of the various possible 2008 Democratic presidential contenders since November of 2004. I have a pretty good personal sense of what these stats mean: they are the expressions of committed support for presidential contenders, the kind of support so committed that it is ready to be expressed publicly. That’s a strong connection to a candidate: one that costs a bit of money, one that is ready to withstand the challenge that can come from public expression, one that is meant to be durable over time. Polls responses, on the other hand, are an expression of a weak connection to a candidate: one that costs no money to express, one that is kept anonymous and only reported in the aggregate, and one that can change as soon as a respondent hangs up the phone. We should expect the two to be related to each other in some way, but not to be identical in character.
Of course, that’s just my own thought about the meaning of our statistics. It’s possible they are weak in some regards as well. Unlike data captured in a national representative survey, we know that the data we grab from sales is unrepresentative. Part of that unrepresentativeness isn’t a problem: if we’re talking about committed support, then surely we wouldn’t want to collect data from uninformed and uncommitted Americans. The minority that is engaged and has formed commitments is an important subset in its own right, even though and especially because it is not drawn randomly from the U.S. population. These committed individuals are the sorts who are likely to form constituencies of core support for presidential candidates. But what if Irregular Times sales are drawn more strongly from a particular subset of those interested in Democratic presidential contenders, those interested perhaps more in one candidate than another? That might be a problem. So I’ve decided today to look at that potential problem and winkle a bit of insight out of it.
Comparisons by Google
I unfortunately can’t answer the question of veracity directly, since I just don’t have complete data on all those strongly committed to all presidential candidates. Nobody does, although many people try to get some version of that data. But we can try to figure out if some candidates are privileged over others in the likelihood that they’ll find their way to our collections of stickers, buttons, shirts and so on. If people search for a candidate’s name and are looking for a bumper sticker, for instance, what will they find? To be more specific, if we type in a candidate’s name and “bumper sticker” for a google or yahoo search, how prominently are links to our shops located, and how much does that vary across presidential contenders? Let’s give it a try and see what turns out.
Joe Biden: Google Search Result #1, Yahoo Search Result #1
Hillary Clinton: Google Search Result #1, Yahoo Search Result #4
Christopher Dodd: Google Search Result #1, Yahoo Search Result #1
John Edwards: Google Search Result #2, Yahoo Search Result #1
Al Gore: Google Search Result #3, Yahoo Search Result #1
Mike Gravel: Google Search Result #1, Yahoo Search Result #1
Dennis Kucinich: Google Search Result #1, Yahoo Search Result #1
Barack Obama: Google Search Result #1, Yahoo Search Result #2
Bill Richardson: Google Search Result #1, Yahoo Search Result #1
All these 2008 Democratic candidates appeared in the first handful of search results for both Google and Yahoo (and more commonly so than for any other vendor). Candidates who consistently appear as number ones are, if anything at all, slightly less common in our sales results, which means that our measures of sales would lead to results that are conservative in the statistical sense, minimizing differences between the candidates. For any of these major contenders (and yes, for more minor and undeclared contenders too, as you can see for yourself), a big portion of those seeking an election item is going to come through Irregular Times, making the resulting patterns more meaningful.
Comparisons to Google Trends
Another way to assess the extent to which our results for committed support among 2008 contenders is an accurate reflection is to compare our measure to another. Google has recently come out with a feature, called Google Trends, that provides information on the popularity of various Google searches carried out by people over time. The following are the Google trends in searches for the five most popular presidential candidates, from January of 2006 to the most recent month for which data is available, February 2007. The second, smaller graph indicates the appearance of news articles about the candidate in the Google News search index.
Apologies for vertical cuts in that graph: it was too wide to fit on one line in its original form. For comparison, here is our month-to-month data over the same period for sales of items supporting the same 2008 democratic contenders:
The two are not identical, as they shouldn’t be, since they measure different things: a Google search is a weak curiosity, and a bumper sticker purchase is a strong commitment. But they are quite similar in the areas that matter: the Gore surge mid-2006, the small Richardson surge in January 2007, the overall ordered ranking of the candidates across time, and so on. These results lend credence to the supposition that our bumper stickers sales data are showing something real and important, even if the source of the information is somewhat unique.