Israel Election Forecast Model Explains Discrepancy With Polls
Translated & summarized from JDN by baba
The 'Barometer' election forecast model predicts 58 seats for the coalition, four more than current polls suggest. The model incorporates demographic trends, past local results, and voting patterns to adjust for potential polling biases. It assigns different weights to polling firms based on past accuracy and analyzes specific sectors like the Haredi and Arab communities. The methodology aims to provide a more robust prediction by not solely relying on survey data.
The story in 6 lines · by baba
- The 'Barometer' election model forecasts 58 seats for the coalition, exceeding the 54 seats indicated by average polls.
- The model integrates demographic data, past election results, and voting patterns to refine poll predictions.
- Polling firms are weighted based on their accuracy in the last three election cycles.
- Demographic growth projections suggest a gradual shift in the political balance over the coming years.
- The model accounts for potential underestimation of certain parties, like Shas, in opinion polls.
- Methodological limits are applied to the forecast to prevent extreme deviations from other data sources.
A forecasting model called 'The Barometer,' developed by JDN News, predicts 58 seats for the coalition parties in the upcoming elections, a figure four seats higher than the average of current polls, which show 54 seats. The model's methodology differs from simple polling averages by incorporating demographic data, past election results in specific localities, and voting patterns. It aims to identify potential biases in polls, particularly when multiple surveys show similar deviations, as occurred in the 2021 and 2022 elections where Netanyahu's bloc was underestimated by an average of 3.5 seats.
The Barometer assigns different weights to polling firms based on their accuracy in the last three election cycles, prioritizing accuracy in bloc predictions, party data, and consistency. Direct Polls and Next Data each receive about 19% of the weight, Kantar and Maagar Mochot about 15%, and other firms around 8%. The model also translates poll discrepancies into voter numbers, estimating a difference of up to 422,000 votes between the most optimistic and pessimistic poll interpretations.
Demographic trends are factored in, with projected annual growth rates for different population groups (4.2% for Haredi, 2% for religious, 1.9% for Arab citizens, 0.7% for traditional, and 0.6% for secular). These growth rates are expected to gradually shift the political balance, potentially reducing the center-left bloc's share from 39.7% in 2022 to 36.4% by 2034, assuming consistent voting patterns. The model also analyzes local trends in approximately 1,200 localities, projecting future voting behavior based on past performance, with greater reliance on regional trends for smaller towns.
Special attention is given to the Haredi and Arab sectors, where polls may be less reliable. For instance, polls underestimated Shas by 2-3 seats in 2022. In these sectors, polls are given half the weight, with the other half split between demographic and geographic components. The model predicts Shas will receive around 9.9 seats, higher than the 7.9 estimated by polls alone. For Arab parties, the model assumes a 58% voter turnout, though it acknowledges the volatility seen in past elections, such as the 44.6% turnout in 2021 with a similar list structure.
Finally, The Barometer applies methodological limits to its forecast, preventing the coalition's seat count from falling below a certain threshold derived from non-polling models, thus explaining the divergence from the raw poll averages. A specific rule also determines if a list is projected to pass the electoral threshold, based on the combined weight of the polling firms predicting its success.
