That's the message Israelis have heard from their media for months now. So my message today is: Don't trust the polls. Anything could still happen.
1. Strategic voting
This year more than most, I get the sense that many Israelis are voting strategically. That is, they plan to cast their ballots not for the party which best represents their views, but for one which will either influence the structure of the Knesset in some direction, or will "send a message" of protest.
Strategic voting effectively relies on the accuracy of the opinion polls. That is, a voter who supports Kadima may assume that Kadima has the election locked up (as Olmert foolishly stated a few weeks ago), and thus feel free to vote for a different party to raise other issues of importance to him. Ironically, Kadima's success in the polls is probably responsible for the recent boost in support for Labor. No one wants to see Amir Peretz as prime minister, but if Olmert is prime minister, many voters (though not me!) would like to see Peretz have significant influence in the coalition.
The problem, of course, is that the strategic voter implicitly assumes that no one else is voting strategically, that all other Kadima supporters (for the sake of this example) will continue to vote Kadima, so he can safely vote Labor without affecting Kadima's victory. This is obviously absurd. If everyone tries to influence Olmert by voting for their second preference party, Kadima won't win in the first place.
It's what scientists call a feedback loop. People decide how to vote based on what yesterday's polls say other people decided. Solving those equations would tax the greatest practitioners of chaos theory.
Strategic voting, I suggest, explains much of Kadima's recent fall in the polls. Once voters take Kadima's victory for granted, more of them feel comfortable switching their votes to other parties. If Kadima slips too far, though, voters will shift back to it. How many will do each? We'll know that only on Wednesday morning.
2. Small parties
This year more than most, several small parties have been hovering near the threshold of votes needed to enter the Knesset. Strategic voting and feedback loops are at play here too. Many voters would like to support some small party or other, but they are reluctant to waste their votes on someone who will fail to pass the electoral threshold. If all the supporters of Green Leaf, or the Pensioners, or Baruch Marzel, were to vote for those parties, they would clearly succeed. But the chance of failure deters enough voters that "can't pass the threshold" is a self-fulfilling prophecy.
Each small party which does or doesn't get in shifts at least 2-3 Knesset seats, possibly affecting the coalition balance.
The big unmentionable in this realm affects the Arab parties. Polls see three Arab parties getting 8-9 seats in total, meaning 2-3 seats per party. This is dangerously close to the threshold (which is higher than last time), and some detailed surveys have indicated that one or even two of them may fail to get in. This could crush the Arab factions to just 3-6 seats, enlarging all the other parties proportionally.
3. Voter turnout
All the indications are that turnout this year will be lower than ever. More and more voters are disillusioned with the parties (I know I am), and many of them apparently do not plan to vote, or will vote for parties guaranteed to fail. This is a new phenomenon in Israel, where until 2001 turnout had consistently been over 80%, effectively including nearly all able-bodied resident adults.
As a result, Israeli pollsters don't have much experience developing a turnout model, that is, a way to forecast who will actually cast a ballot. Turnout is likely to depend on one's political and religious orientation, one's ethnic background, and other unknown factors. The pollsters don't know enough about it to give meaningful results.
Another unusual factor this year is the high rate of undecided voters this late in the campaign. Though I've already stated my preference, I still toy with the alternatives and have reconsidered my choice several times over the last few days. (Though so far I've ended up in the same place.)
How will people actually vote? All the pollsters can say is that "the undecideds generally break down like the rest of the population". But will they? No one knows.
5. Kadima, etc.
It's the elephant in the room. Kadima is an unprecedented phenomenon, perhaps in any democracy. A popular prime minister breaks away from his own party a few months before the election and establishes a new one, in cooperation with leading members of the opposition. Then he falls ill and is succeeded by his unloved deputy.
And have we mentioned the rise of Hamas? And the aftershocks of disengagement, which undermined many people's faith in Israeli democracy in general, and the mainstream parties in particular?
Too much has changed in Israeli politics to rely on the usual determinants of voter behavior. Opinion polls ask people what they would do if elections were held today. No one can predict will people actually will do behind the curtain.
6. The polls don't agree
How many seats will Yisrael Beitenu get: 7 or 15? Depends on which poll you believe. Maagar Mochot said 15; Dialogue said 7. (Dahaf and Teleseker said 12.)
Will NU/NRP get 8 or 12? That's a big difference in influence, but just about 3% of the votes.
It's hard to get an accurate forecast of the Knesset, since small changes cause large effects. A typical survey of 500 participants has a sampling error of 4%, or 5 seats in the Knesset. And that's without considering all the other sources of error in election polling. How can small parties be meaningfully forecasted with such tools?
Ultimately, it comes down to the methodologies of the different polling companies. They can be more important than what people actually tell the pollsters.
7. The polls' record is poor
In Israel's last election, in January 2003, you probably remember the polls as being pretty accurate in forecasting a landslide for Sharon and Likud. But how accurate were they really?
In the following chart, the first column of numbers is the result of the 2003 elections (in Knesset seats); the second column is the range of results of opinion polls taken the week before the elections.
Notice that in most cases the polls underestimated support for the right - substantially in the case of Likud itself - and overestimated support for the left and the Arab parties. None of the polls gave Likud anywhere near 38 seats.
Likud had been slipping in the polls in the runup to the election; that may have induced supporters to vote Likud on Election Day. Kadima's recent slide may also be good for it at in the polling booth. Or it may reflect a genuine drop in support. The opinion polls don't distinguish between these effects.
And let's not mention that Shimon Peres is today the head of the Labor Party, having defeated Amir Peretz as the polls indicated. This came ten years after he was elected prime minister over Binyamin Netanyahu, as the polls indicated.
8. Message: Make your vote count
My point is this: Nothing is over until the votes are cast. Make yours count. Vote for the party you most wish to see represented in the Knesset.
If you give up and throw away your vote, you're letting the pollsters determine the election.