Ever since opinion institutes and statisticians began conducting surveys to try and determine election behavior, politicians and the media have turned to them for an indication of what voters prefer. The pollsters have become the modern crystal-ball readers of the political landscape.

But a poll is only so effective as long as it is accurate, or at least pretty close to accurate. The moment the first voter returns begin to trickle in, the poll forecasts begin to fade into memory as they become less like the actual results.

During this federal election campaign, pollsters were hard at work surveying German voters every week, sometimes even several times a week, to measure public sentiment for a particular party or candidate.

That's why it's so surprising that the opinion polls were so far off on so many points once the exit polls started coming in on Sunday evening.

Way off the mark

Take the Free Democrats, for example. All the major polling institutes had predicted the FDP receiving about 6.5 to 7 percent of the vote, a considerably worse result than the actual return of 9.8 percent, the party's best in 15 years.

Back in the 2002 federal election, the majority of institutes had erred on the opposite side, predicting the FDP would receive a much higher percentage of the votes than they actually did -- 7.4 percent.

Even more glaringly, polltakers failed to accurately predict the outcome of the Christian Democrats. The CDU with Angela Merkel as the leading contender for chancellor was the clear front-runner in all the surveys throughout the summer campaign. Only during the last couple of weeks of the campaign did polls start showing Gerhard Schröder and the Social Democrats edging closer. Nonetheless, every polling institute gave Merkel a good 6 to 7 points ahead of the SPD to put her party in the 40-percent braket.

It was nothing less than a shock for many to see the CDU coming out at just 35 percent, only one point ahead of the SPD.

2002 embarassment

The last federal election in 2002 already represented a low point for the country's polling institutes. Although they correctly showed the vote as too close to call in the week before the elections, at least one predicted a conservative win a few hours after counting had begun.

That erroneous prediction prompted the 2002 conservative challenger, Edmund Stoiber, to prematurely claim victory before an embarrassing climbdown hours later.

Inaccuracy inevitable?

Pollsters defend such differences between prediction and actual outcome as part of the nature of their work. Because each of the surveys they conduct is based on only a small portion of all eligible voters and so-called margins of error are inevitable, it's impossible to foresee and account for every outcome. The surveys are also a momentary snapshot of voting behavior. The way an individual thinks and reacts to political circumstances can change within days.

Because the 2005 election was such a brief but intense one from the moment the early elections were announced to the debate over their constitutionality, public opinion went back and forth and up and down between the parties. German voters were also much more volatile than ever before, pollsters said in self-defense.

Even as late as Friday, around 25 percent of the electorate were uncertain whom to back or whether to vote at all. Given such indecisiveness, it's no wonder the pollsters can't accurately predict the future.