Contact
← All work/ RESEARCH / PROJECT 61

ELECTIONS · 2003–2015

Project 61.Who gets the polls right?

Every Israeli election poll published in the final month of five campaigns, scored against the real result. Pollsters are rated on their record, and the weighted average of them all beat every single one.

WSJ · FiveThirtyEight · Times of Israel

5 / 5

elections where our weighted average had the lowest error

99

Project 61 rating, vs a pollster average of 80

14

polling institutes rated

0.18

error score in 2015, vs 0.40 for the median pollster

Pollster rating

0–100, weighted toward recent campaigns. Higher is more accurate.

Project 6199
Mina Tzemach88
Midgam84
Market Watch84
Panels Politics82
Teleseker81
Midgam - Mina & Mano81
Rafi Smith80
Hagal Hachadash79
Kamil Fuchs79
Shluv Millward Brown77
Geocartography76
Maagar Mohot74
Shvakim Panorama69

Error by election

Average error of each pollster’s final-month polls. Lower is better.

Maagar Mohot: 0.45Millward Brown: 0.44Kamil Fuchs: 0.42Panels Politics: 0.40Midgam (G&T): 0.40Rafi Smith: 0.39Teleseker: 0.32Hagal Hachadash: 0.32Geocartography: 0.25Project 61: 0.18
Maagar Mohot0.45Millward Brown0.44Kamil Fuchs0.42Panels Politics0.40Midgam (G&T)0.40Rafi Smith0.39Teleseker0.32Hagal Hachadash0.32Geocartography0.25Project 610.18

How the score works

Each poll is compared with the actual result of that election. Errors are combined with sampling error and the time between publication and election day. Pollster scores are compared with the average error of that campaign, regressed to the mean, and weighted toward recent campaigns.