NBC News and Marist College have just released 05/24/12 a poll that shows Barack Obama 4 points atop Mitt Romney in the critical swing state of Florida. (These results are 10 points friendlier to President Obama than results released yesterday 05/23/12 from Quinnipiac University, which is 2-hours East of Marist).

But: hidden with the Marist tables is the most important finding for any journalist or scholar who wants to report on public opinion polls in 2012: The difference between the Home-Phone respondents and the Cell-Phone respondents.

Marist is one of only two research companies that routinely cross-tab by how the interview was conducted. Since a majority of public opinion polls released at the state level do not include any cell-phone respondents, and since the few pollsters who do claim to include cell-phone respondents do not cross-tab by how the interview was conducted, Marist’s poll results provide a breathtaking window into the difficulty of trying to reconcile poll results in 2012.

Marist reports a a 26-point difference in how cell-phone and home-phone respondents vote on the Obama-Romney “horserace” question. Romney leads by 3 points among home-phone (aka: “landline”) respondents. But Obama leads by 23 points among cell-phone respondents. When the 2 populations are proportionally blended (28% cell phone, 72% home phone), Marist reports out that Obama is ahead by 4 points.   Had the survey only included land-line respondents, Romney would have been ahead.

These poll results contrast with results released yesterday 05/23/12 by Quinnipiac University, which in Florida show Romney up by 6 points — a 10-point difference. Quinnipiac’s release, though detailed, does not provide insight into how cell respondents vote compared to home-phone respondents. Quinnipiac does provide, in a non-linkable Word document, the following breakdown:

This table would appear to say that 19% of all of the respondents that Quinnipiac interviewed were “cell phone only.” It is unclear whether the 19% CPO that Quinnipiac reached is an apples-to-apples comparison with the 28% cell respondents that Marist reached. Every journalist writing about public opinion polls in 2012 should ask the pollster to disclose how the cell-phone respondents differ from the landline respondents. The answer to that question will go a long way toward reconciling apparently disparate poll results.