Regular readers will know that I'm a fan of the growing use of "word cloud" graphics to depict the results of open-ended questions. As such, it's a pleasure to pass along some interesting new examples posted today by Patrick Murray, the director of the Monmouth University Polling Institute. The examples below shows the word or phrase that "mainly unaffiliated voters" mention when asked about the candidates for Governor in New Jersey:
Do click through for full-size versions as well as the word clouds for independent Chris Daggett and for the image or message that voters say stand out from the campaign ads they have seen.
Another intriguing aspect of this project is that it comes from "an online panel study with mainly unaffiliated voters" that the Murray says Monmouth is conducting "as part of our polling coverage." I emailed him to ask for more details. Here is his response:
We expected that the New Jersey Governor's race would be volatile and so it would be interesting to supplement our standard trend telephone surveys with a panel to track individual level changes over the final six weeks of the campaign. We also wanted to experiment with some ideas that can't be easily done in a telephone poll (i.e. reactions to visual images). That is why we are not releasing marginals (i.e. full percentages). We'll continue to be conducting our standard telephone polls for representative sampling purposes.
Approximately 1/3rd of the panel respondents were recruited from prior Monmouth University telephone poll respondents who indicated an interest in participating in online surveys. The remainder came from a purchased random voter list (of which 60% were "unaffiliated", meaning they are registered with no party - although that does not necessarily mean they are politically independent). Our target was to arrive at a fairly even partisan distribution and we assumed that unaffiliated/independents would be less likely to participate. Self-identified party affiliation among the panel's Wave 1 participants was 29% Democrat, 31% Republican, 41% independent (i.e. a little less D and a little more R than a representative New Jersey voter poll). The panel is also somewhat more male and Caucasian than our representative voter polls.
We found that the demographic skews in the panel have little effect on the relative frequency of the words mentioned for each candidate, which is what the clouds illustrate.
The subject of online panels and their use is source of great controversy, but this experiment is indicative of the way pollsters trained in traditional methods are starting to experiment with online techniques: They are cautious and careful, but also eager to exploit the opportunities created by the online mode.
Follow Mark Blumenthal on Twitter: www.twitter.com/MysteryPollster