Case Study: Media – Political Campaign Tracking
Case 2: Tracking Political Campaigns
A second way that we have helped media is through using custom-designed research and strategic polling tracking trends in public opinion about public policy issues and political campaigns. Our most recent polling about a political campaign for our client, The Commercial Appeal, like all of our published polls, very clearly demonstrates the accuracy of our political polling.
- To compare the accuracy of the last polls taken by various firms before the election, two things have to be kept in mind.
- First, they are not comparable because things can change even up to the day before and the day of the election.
- Second, if you’re going to go ahead and make that comparison anyway, you must adjust for undecided voters, otherwise the poll numbers tend to be much lower than the election results.
- Regarding the issue of adjusting for the undecided voters, two points need to be made from other authoritative sources:
- “Mythically, undecided voters are the most serious students of the process, reserving their judgment until all the facts are in; they make up their minds as they enter the voting booth. In reality, undecided voters usually don’t know or care who’s running and are most likely not to vote at all. . . . So, don’t worry about the undecideds.” (The Irrelevance of the Undecided Voters, inThe New Prince, Dick Morris, Renaissance Books, Los Angeles, 1999, pp. 220-221).
- It is common practice in the polling industry when comparing the latest poll results taken before an election to the election results to take the “undecided’ and “refused” responses out of the base and then re-percent the figures to the base of people who answered the question.
- This is known as the “Proportional Allocation Method” (see, “In Defense of Polling,” by Kenneth Warren, Westview Press, Boulder, Colorado, 2003, pp. 287-288; 319-326).
- This method assumes, like the Dick Morris quote above, that most “undecided” voters are not as knowledgeable and end up not voting at all and if they do vote, in all but the rarest of cases, they tend to split among the candidates in the same proportions as those who were decided when the poll was taken.
- Using the proportional allocation method, the results of the last polls before the 2007 Memphis mayoral election are compared to the actual election results in the following table. Using the standard adjustment procedure that is described above, in the table below, the “undecided” and “refused” voters were taken out of the base and the numbers were “re-percented” to the base of people who answered the question. For example, in the Ethridge poll, 35% were either “Undecided” (22%) or “refused” to answer the election question (14%).
- As this table shows, on both an unadjusted (5 points) and an adjusted basis (8 points), Ethridge & Associates, L.L.C.’s poll more accurately predicted the spread between the top two candidates (8 points) than did either the Rasmussen poll (1 point unadjusted; 1 point adjusted) or the Yacoubian poll (2 points adjusted; 3 points unadjusted). All of the other polling firms, even though their polls were taken closer to election day than ours, showed the race “too close to call” (within the margin of error). Ours was the only poll that showed Willie Herenton would win by a significant margin.
- The astute reader of the table above will wonder why Ethridge’s numbers for the top two candidates were deflated compared to the other polls. The reason is that The Commercial Appeal had a standard practice of telling respondents in the introduction to the survey that the poll was being done for The Commercial Appeal. In this particular race, incumbent Mayor Willie Herenton had made a convincing case to his voting base that the newspaper was biased against him. This caused many of his voters, and some of Chumney’s, who were polled to refuse to answer the election question: a phenomena that we have often seen in campaigns involving Herenton. This meant that to be interpreted properly, the responses to this question had to be adjusted to a new base, as shown in the “re-percented” column above. A reporter for The Commercial Appeal got this right, when he reported on his blog, Eye on Politics that:
- Herenton’s supporters often declare that polls have been very wrong when it comes to measuring his actual level of electoral support, the idea being that white voters who like him are reluctant to say so and black voters are suspicious of answering polls. In a final poll conducted for the 2007 race . . . Ethridge surmised that if many of the people who were refusing to participate in the poll were indeed Herenton supporters, he could likely get more than 40 percent of the vote — very close to his final tally of 42 percent. Zack McMillin, Eye on Politics, MemphisNewsBlog.com, on August 13, 2009 2:42 PM
Note that we conducted several polls for The Commercial Appeal in this race. Earlier polls showed that Herenton would have a difficult time winning (note from the table above that he won with less than 50% of the vote). Later polls showed trends shifting away from Chumney and toward Morris. However, after our last poll was conducted for The Commercial Appeal, our private polling for City Council races showed that the trend shifted away from Morris toward Chumney for second place. Those private polls were not published. This trend in Chumney’s favor in the race for second place began after our latest poll for The Commercial Appeal and thus was not reflected in any of the reporting on our polls.