Israeli Pollsters Face Potential Collapse of Predictive Models
Hundreds of thousands of Israelis may vote in the upcoming election without being accurately reflected in current polls, according to an analysis. These voters include young people, reservists, evacuees, and citizens whose views have shifted significantly due to the ongoing war and its related crises, such as the hostage situation and a deep public trust deficit. Pollsters may discover that their models represent only those willing to participate in surveys, not the broader Israeli society.
The core issue is not necessarily that the polls are factually incorrect, but that their underlying data may no longer be representative of the electorate. A sample of a few hundred respondents attempts to describe millions of eligible voters, which is only feasible when the sample is well-constructed, response rates are reasonable, and voting patterns are stable. However, post-October 7th Israel is described as a society in emotional, security, social, and political turmoil, making previous assumptions unreliable.
The article highlights that sampling error is only a small part of the problem. It does not account for issues like difficulties reaching young people, discrepancies between stated intentions and actual voting behavior, last-minute decision changes, and flawed assumptions about turnout among different demographics. The critical question is not just what people say, but who agrees to answer, who refuses, who conceals their vote, and who changes their mind on election day.
The most significant flaw in political polling, the author argues, occurs after data collection, during the weighting process. Each poll relies on a model that makes assumptions about turnout for various groups, including youth, the Arab community, undecided voters, and specific demographics like the ultra-Orthodox, new immigrants, and residents of peripheral areas. When these models are inaccurate, they create a misleading illusion of precision. The electoral threshold in Israel further exacerbates these potential errors, as a small shift in votes can cause a party to fall below the threshold, drastically altering the political landscape, as seen in past elections.
This phenomenon is not unique to Israel, with examples cited from the US, Australia, India, France, and the UK, where polls have misjudged voter sentiment, particularly among specific demographics or due to tactical voting. The article contends that Israeli polls have evolved from mere measurement tools into political actors, influencing perceptions of legitimacy, media coverage, fundraising, and even pressuring candidates to withdraw. This influence, presented as objective reporting, poses a democratic challenge. The author calls for greater transparency in polling methodologies, including disclosure of sample size, data collection methods, response rates, weighting procedures, and the client who commissioned the poll. Media outlets are also urged to avoid sensationalizing minor poll fluctuations.
Ultimately, the upcoming election will test whether the Israeli polling industry can accurately gauge the nation's current sentiment, or if it remains trapped by its own outdated assumptions, failing to capture the deep societal shifts occurring beneath the surface.
