Start-up nation takes on election numbers in 'data-thon'

Data analysts find new parties disrupt accuracy of polls; immigrants in last 25 years are three times more likely to be right-wing than rest of population.

By LAHAV HARKOV
An Israeli flag is seen in the background as a man casts his ballot for the parliamentary election (photo credit: REUTERS)
An Israeli flag is seen in the background as a man casts his ballot for the parliamentary election
(photo credit: REUTERS)
Polls become more accurate the closer they are to Election Day, according to results of the first-ever Israeli election “datathon” obtained by The Jerusalem Post Sunday.
One of the factors making election polls as notoriously inaccurate as they are is new parties, one team found; but polls become more accurate as Election Day nears.
In a related finding, another team analyzed the success of new parties and compared them to current polls, to predict Koolanu’s seats in the next Knesset. They estimated that the party will have between eight and 10 MKs.
In last week’s polls, the party averaged 8.1.
Sisense, a company whose technology analyzes and visualizes massive amounts of data, hosted a data-thon Friday, in which a group of analysts, engineers, data finders and other tech professionals formed teams to review and cross-section demographic data, voting patterns, social media and other publicly available information to come up with insights on the upcoming election.
The data-thon was inspired by “hackathons,” events in which developers spend hours trying to solve a problem, answer a question or come up with new ideas.
One team, from the Hamidgam polling app, found most voters who support Koolanu, previously voted the Likud and Yesh Atid, as did most Likud voters.
Another team found that immigrants who came to Israel after 1990 are three times more likely than the rest of the population to vote for the Right than the Left; and people who live over 50 km. away from Tel Aviv are twice as likely to vote for the Right than the Left.
However, another team found, towns near the Gaza border tend to vote for the Left, and the North is more right wing than the South.
The winning team in the data-thon were Technion students Omri Dor and Danny Rasin, who used an algorithm to segment voters into demographic groups and associated them with the party for which they were most likely to vote, by comparing results of the 2013 election with Central Bureau of Statistics data, which gave them details about the demographic makeup of the voters in each polling place.
Their data showed that if there had been a 100 percent voter turnout in 2013, the results of the election would have been much different, with the Likud gaining eight seats, the three Arab parties gaining six and United Torah Judaism losing five and Yesh Atid going down by three.