The reputation of polling as a tool to determine voting outcomes in political races across the world in recent years has seen two conflicting trends. On the one hand, failures to accurately predict the winners and losers in major elections – such as Donald Trump’s victory in the 2016 US presidential elections and the “yes” majority in the 2016 UK Brexit referendum – have tarnished the credibility of polls as a reliable predictor of political outcomes. Similarly, in Israel, polls suffered a major blow on election night in 1996 after exit polls predicted that Labor leader Shimon Peres had won Israel’s first direct ballot for prime minister. Yet the next morning, Israelis woke up to discover that the real result was that Peres had been narrowly defeated by Likud leader (and current Prime Minister) Benjamin Netanyahu.
No wonder then that Peres, a winner in polls yet almost always a loser in elections, coined a telling aphorism about political surveys: “Polls are like perfume – you can smell [them] but not drink.”
Yet at the same time, polls enjoy unprecedented influence on the political arena in almost all democracies. In Australia over the last decade, consistently low polling numbers were a powerful impetus for leadership spills in both the Liberal and Labour parties, unseating sitting prime ministers Kevin Rudd, Julia Gillard, Tony Abbott and Malcolm Turnbull.
Israel 2019 – elections born and managed by polls
The influence of polls on the April 9 Knesset elections is seminal. Israeli analysts suggest strong polling data gave Netanyahu the confidence to call an early election and, in doing so, attempt to arm himself with a political mandate to combat the serious criminal charges being levelled against him. On the other side of the political map, other polls led two former IDF chiefs of staff Benny Gantz and Moshe Yaalon to merge with Yair Lapid’s Yesh Atid (“There is a future”) to run together as the Blue and White party, in a deal brokered by another former IDF chief of staff, Gabi Ashkenazi.
Poor poll results for veteran politician Tzipi Livni – who narrowly missed becoming prime minister following the 2009 election – ultimately led her to retire from politics entirely. Meanwhile, former Likud MK Oren Hazan, having failed to gain a realistic position on the Likud list in the party’s primary, is trying to find a back door into the Knesset through the long dormant Tsomet political party, in a decision guided again by polls. The list goes on.
Polls dealing with politics are never simple. Universal methodological problems of political polling include, to name a few: incorrect respondent collection methods; inadequate sample representativeness (with regards to age, gender, ethnicity, religious affiliation, place of residence etc.); the tendency of people not to reveal their genuine views in polls or of certain demographics to disproportionately refuse to answer entirely; the impact of previous polls (whether the “bandwagon” effect for people who like backing frontrunners or, conversely, the “underdog” effect for people who deliberately don’t); and the huge influence of the internet, social networks and fake news, sometimes computer-generated (ie. “bots”) – all of which are not always taken into account correctly (or at all) by polls.
Israel’s challenging demographics further complicate the polling challenge. As mentioned, a pollster must first conclude that the sample collected for the poll is representative of the general voters’ population. Certain segments within the Israeli public, such as the ultra-Orthodox and elements of the Arab population, are less accessible to researchers due to technical obstacles (such as, for example, the aversion of ultra-Orthodox Jews to smartphones), or because some members of those groups are hostile to outsiders, or to the idea of voting in Israel altogether.
Another challenge unique to Israel is a result of the laws determining election outcomes. The Knesset is selected via a nationwide, proportional representation system, where 3.25% of valid votes make up the electoral threshold which must be met before a party is eligible to have any seats in the 120-seat parliament. In practice, this means a party needs a minimum of four seats to be in the Knesset. The final allocation of seats is determined after further calculations, which include redistribution of the votes that went to parties that failed to make the threshold to other parties that had a vote redistribution agreement with them.
The practical implications of this maze of complexity is that pollsters may succeed in determining the proportional support of voters in the largest parties, yet often struggle to predict the outcome with respect to the smaller parties, particularly the ones hovering near the electoral threshold.
Concluding whether or not a small party will reach the threshold is often dependent on assessments that exceed the margin of error of any given poll. This measure of uncertainty is a poll-taking reality that can never be completely eliminated.
To complicate matters, 47 parties will be on the ballot on April 9 – a record high number for Israel. This means polling professionals have to accurately factor in the effects of the votes of niche parties, and draw conclusions that can vary widely depending on very small shifts in the number of votes. For a marginal party, simply getting mentioned in a poll report can create a positive ripple effect in future polls, while being ignored can have the opposite effect.
Meanwhile, as an election approaches, Israel’s traditionally large number of undecided or swing voters choose parties. In the past, Israeli pollsters have tracked such votes breaking late towards a specific party or parties, but could not publicise this trend because of a law forbidding publication of polls four days prior to the elections. This is one reason polls failed to accurately predict, for example, the success of the Pensioners party, winner of seven seats in 2006, or Yesh Atid in 2013, which gained 19 seats, while polls predicted only 12 to 13.
Useful, not foolproof
Despite all of their shortcomings, polls in Israel over the past few years have performed quite well and accurately placed Knesset seats in the majority of recent elections.
Israeli elections in April, much like the ones expected in Australia soon after, are yet another test case for both the classic and new issues related to political polling. They pose a plethora of methodological challenges, questions about bias and influence, and new problems specific to 21st century society. Both elections, in that sense, present a good opportunity to learn how to improve political polling, and how to better read and make use of the data polls produce.