Surveying the polls

Last week's Knesset election was marked by what some observers dubbed "polls mania".

elections06.article.298 (photo credit: )
(photo credit: )
Election Day doesn't only deliver the ultimate verdict for candidates; it also grades the performance of those who attempt to predict the outcome of the race - the pollsters. Last week's Knesset election was marked by what some observers dubbed "polls mania." Never had there been such a surfeit of surveys. In fact, political reality was measured by polls as never before: Since Kadima was a brand new protagonist, testing its strength and electability for the first time, the only way to gauge its viability was to rely on polls. To some extent it can be said that Kadima derived its validation from polls. So how did the pollsters do? In the broadest terms, they did forecast the outcome. Kadima did emerge as the largest Knesset faction, Labor did nearly retain its pre-election number of mandates and the Likud was indeed trounced. Successfully envisaging the big picture isn't to be taken for granted or scoffed at. In previous campaigns, after all, pollsters failed to do even that. Those whom they prematurely crowned as winners sometimes ended up in the opposition. This year, however, there was an unprecedented proportion of undecided voters, constituting a riddle which the pollsters never quite managed to solve. That's where the bad news begins. The pollsters utterly failed to appraise the scopes of victory and defeat and especially to realize what was happening to the smaller lists, most notably this campaign's wild card - the Pensioners' Party. In other words, with some exceptions, they got the numbers wrong. No poll foretold the full extent of Kadima's decline in recent weeks, the rise in Shas's fortunes, the dramatic quadrupling of Avigdor Lieberman's parliamentary contingent or even the retention of the Arab lists' strength. The fact that the pensioners managed to elect seven MKs was something not a single poll came close to anticipating. The problem may be in polling samples, which, it's suggested, fail to fully estimate swings among the youngest voters. The reason may be mundane: Polling is conducted via landline phones while many first-time voters are glued to cell phones. But it may be worse yet. Young voters especially are given to sudden fads which polls can't discern, like those triggered by showbiz celebrities who urged their fans to vote for the pensioners rather than abstain. This idea spread quickly during the last few days of the campaign. The impetus, however, was ironically generated by the polls themselves, when some showed the pensioners just crossing the Knesset entry threshold. That made their list a viable vehicle for protest. The polls, thus, aren't merely reflectors of a given situation but players which help determine it. Especially in the Israeli threshold system, their predictions can become catalysts that kick-start dynamic developments. Similarly, Kadima's fortunes soared when the polls gave it 40-plus mandates. That inspired over-confidence, as evinced in Ehud Olmert's declaration that the elections "were already won." His hubris may have driven some voters away and lulled others into complacency, if not apathy. Polls create expectations or turn into self-fulfilling prophecies - for better or worse. The fact that polls influence voters has led many democracies to prohibit their publication for a given period before Election Day. All this assumes, of course, that polls are conducted on the up-and-up. A Channel 10 expose in February challenged this assumption. It showed underpaid temporary polling firms' employees filling out questionnaires themselves to complete their quotas instead of conducting proper telephone interviews. There's almost no supervision of the polling organizations and they're loath to agree to a code of ethics. This is no trifling matter and should certainly not be left to pollsters themselves to sort out. Legislative guidelines may have to be buttressed and enforcement of existing regulations guaranteed. Most of all, those who read polls ought to be cautioned about how much credence to accord the results. Perhaps warnings like those on cigarette packages are warranted. They should inform us that the margin of error is a statistic that only takes into account sampling error, and only applies if undecided voters vote in the same proportions as those who have indicated how they will vote. We should be apprised that factors such as a large number of undecided voters can result in significant inaccuracies. As happened this time.