Inferring too much, we aren’t alone

Running through a European airport during a business trip a couple weeks ago, I picked up a British newspaper and was both delighted and saddened to learn that we are not alone in over-interpreting polls. Indeed, the London Independent went further than I can imagine almost any U.S. news outlet going in distorting the meaning of its findings.

Having performed miserably in elections to the European Parliament earlier this month, Prime Minister Gordon Brown was facing a simmering rebellion within his Labour Party — and in the British system, such insurgencies are not just a headache for leaders, they can bring down the prime minister.

Into the breach rides the Independent with a headline screaming, “ ‘Independent’ Poll Reveals That New Leader Could Transform Labour’s Prospects.” It seems that “Alan Johnson [another Labour politician] would deny [Conservative] David Cameron an overall majority at the next general election if Labour ditched Gordon Brown and installed him [Johnson] as Prime Minister according to a new poll for the Independent.”

The newspaper — no doubt delighted to become part of its own story — quoted unnamed members of Parliament describing the findings as “stunning” and asserting the poll “could influence Labour’s agonized debate over whether it should back or sack the beleaguered Prime Minister.”

No one can deny Brown is in serious trouble. In this poll and every other, with him at the helm, Labour runs far behind the Tories.

However, the question raised by the Independent is whether anyone else could do better — and that is where the evidence wears as thin as a 30-year-old Burberry.

With a sample size of 1,000, the poll has a margin of error of about +/- 3 percentage points on each number. According to the survey, the flailing Brown would allow the Tories to garner 38 percent of the national vote, while Johnson, the alleged savior, would hold them to a mere 36 percent. Needless to say, 36 percent and 38 percent are well within the margin of error of each other, and quite close by any accounting.

The Independent proceeds to pile error on top of error in estimating that while Brown as prime minister would give the Tories a 74-seat majority in Westminster, Johnson would hold the Conservatives at bay, six seats short of a majority.

In a system based on proportional representation, that calculation is fairly straightforward. However, in a constituency-based system, where MPs run from geographically defined areas, how do they translate national votes into seats? The easy way — by assuming a consistent shift across each constituency from the vote in the last election to the number they garner in this poll.

Of course, the easy way is pretty much guaranteed to be wrong. The shift in votes from one election to another has never been uniform across constituencies. If the swing had been uniform in the last general election, Tony Blair would have emerged with a 100-seat margin, but in fact, the election gave him 34 fewer seats than that.

Historically, the variation across constituency has been significant, and in no two general elections since 1945 has the ratio of votes received to seats won ever been the same, making prediction a fuzzy affair, at best.

Adding the poll’s margin of error to the unknown error in translating votes to seats leaves the Independent with little other than yet another confirmation of Labour’s troubles and a sensationalized, nearly fabricated, headline proclaiming that one Labour leader would give the Conservatives a majority that another leader would deny them.

Compared to our British cousins, the American press seems a model of propriety, sobriety and responsibility in reporting on polls.

Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the majority leaders of both the House and Senate.