October 4, 2012

Dispatch’s Latest Ohio Poll Really Shows Obama-Romney Tie

Overweighted with Democrats and seriously unrepresentative.

________________________________

This post went up yesterday at Watchdog.org with a different title and slight revisions.

________________________________

On Sunday, the Columbus Dispatch published the results of its latest mail poll purporting to show that Democratic Party incumbent Barack Obama has a nine-point, 51 percent to 42 percent lead in Ohio over Republican nominee Mitt Romney in the presidential race.

The paper claims that:

  • “(The) poll echoes four others that show Romney behind by at least five points in Ohio.”
  • “[I]t appears the former Massachusetts governor (Romney) first must turn around his own campaign.”
  • “A surge of Democratic support for Obama has transformed the race since the first Dispatch Poll had the two dead-even at 45 percent just before the Republican National Convention in late August.”

The only thing that has “surged” is the Dispatch’s disingenuousness in describing the results of its own polls. Meanwhile, there is an important factor apparently affecting nearly all surveys which pollsters are choosing to completely ignore at their serious peril.

Reporter Darrell Rowland didn’t name all of the other polls showing a five-point or more Obama lead, citing only the Washington Post and Fox News. One poll Rowland conveniently forgot to note came from Gravis Marketing covering September 21-22. Despite weighting the poll with 41.4 percent Democrats and 31.1 percent Republicans, it showed Obama barely clinging to a 0.9-point lead.

The poll the Dispatch conducted a month ago has far more validity. Rowland himself told us why:

In August, almost exactly the same number of Democrats and Republicans responded to the Dispatch Poll. But after the mail-poll ballots went out this time to registered Ohio voters chosen exactly the same way — at random by a computer — more Democrats returned the poll forms than did Republicans. The breakdown: 43 percent Democrat, 35 percent Republican.

In other words, eight points of the nine-point swing from the Dispatch’s dead heat in August is due to the change in party makeup of its late-September sample. Though it’s difficult to peg exactly, Ohioans’ party preference breakdown is more than likely somewhere between the four-point Republican edge Rasmussen found in early September and the four-point Democratic edge Gallup reported in late September. In other words, the most defensible conclusion one can reach right now is that the Buckeye State’s presidential race is in a statistical dead heat.

A May Pew Research report cited an important reason why the credibility of political polls should be questioned as never before. As briefly summarized by blogger “Zombieat PJ Media on the same day as the Dispatch poll’s release: “Out of every seven people contacted by pollsters, only one will answer the polling question, while the remaining six refuse to answer.” And before getting to that point, one must recognize that pollsters aren’t even able to make contact with 38 percent of those they attempt to reach — up from only 10 percent in 1997.

Pew reported that, when all is said and done, “the response rate of a typical telephone survey was 36 percent in 1997 and is just 9 percent today.” It was 21 percent just six years ago. Since the 2008 presidential race, the poll completion rate has probably dropped by almost half, from an interpolated 17 percent to its current 9 percent. Based on this information, it seems extremely dangerous to believe, as pollsters clearly must, that there is little difference between the views of the tiny remaining minority who complete their surveys and the over nine-tenths of the population who either won’t talk with them, can’t or won’t complete the surveys after trying to get through them, or cannot even be contacted in the first place. A perfect storm of pollster embarrassment may be in the works.

Though the Dispatch’s survey was done by mail, Pew further asserts that “The general decline in response rates is evident across nearly all types of surveys, in the United States and abroad.” It’s hard to believe that the surveys done by Columbus’s only daily newspaper are exempt from this trend. No one should take its latest production as presumptively reflective of where Ohioans really stand.

Share

6 Comments

  1. Correct me if I am wrong, but in the past you have noted a strong Democratic bias in these Dispatch mail-in polls.

    Comment by Tony — October 5, 2012 @ 7:59 am

  2. Yes indeed.

    Comment by Tom — October 5, 2012 @ 8:23 am

  3. Thanks for the thoughts on the poll. I agree with the conclusion. Couple of things to add in 2011 Ohio was listed as a state with a +1% margin in republican voter registration (i.e., 37% R and 36% D). So if the voters selected to receive the poll(based on Ohio voter registration) were truly random one would expect the questionnaires mailed out by the Col. Dispatch to be approximately equal by party affiliation. But the questionnaires returned reflected a +8% democratic bias. Since it is counter-intuitive (but not certain) that enthusiasm on the part of D-voters is greater than R-voters 9i.e., D’s more motivated to return the questionnaire than R’s)– this entire poll seems implausible? I suggest either by error or by intention the Dispatch poll questionnaire was mailed out to more D-registered voters that R. What is odd is that the Dispatch is more conservative than most urban newspapers — why would they do this intentionally?

    Comment by Fargo44 — October 6, 2012 @ 12:38 am

  4. I think the whole concept of polling by mail, which may have been pretty good even 10 years ago, is extremely flawed now.

    As to enthusiasm, I would suggest that people who would tend to answer mail polls will be strong, immovable Dems and Reps (regardless of what they tell the pollsters they are), and that Dems would far outnumber Reps because of media distrust.

    Comment by Tom — October 6, 2012 @ 8:01 am

  5. Thanks for the additional thoughts. I sure would agree with your view that polling by mail is an idea that has outlived its usefulness — it is essentially a glacial pace by current standards. BUT I am still coming back to what the h-e-double-toothpick (that’s “hell” for you young readers)CAUSED such an anomalous return (i.e., +8% D). That is beyond (by ~2X) the statistical power of the survey itself (+/- 3%). Since it should be EASY to send out a randomized sample based on the Ohio voter registration data base — I am starting to think there was intentional bias on the party distribution of the questionnaires sent out. I doubt 8% of Ohio voters switched R to D since July when the respondents were 50% for each party. I know I have to let this go but if the Columbus Dispatch is involved in intentional distortion of polling information (I said “IF”) that carries beyond the election at least for us here in Ohio.

    Comment by Fargo44 — October 6, 2012 @ 9:52 am

  6. Bias in send-outs is possible, but I think it’s more likely that Republicans are simply less inclined to respond.

    Comment by Tom — October 6, 2012 @ 12:39 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.