Can election polls be trusted?
Ahead of the upcoming elections, we are flooded with mandate polls, and every media outlet presents a different forecast. Behind the numbers hide unrepresentative samples, biased questions, and entire audiences that are barely sampled. Dr. Nimrod Nir from the Hebrew University explains how to read a poll correctly, who creates the gaps between the polls and the actual results, and why it is important to check the fine print first.
Listen to 'The Headline' — the daily news podcast of ynet. This time: on election polls.
'You can get a lot out of polls. Many things in science and technology are based on polls. Based on this, many things happen in the world that work. Therefore, it can be said that a poll is generally a good tool, although there is a significant difference between a mandate poll and a poll used for empirical research. What a poll mainly does is provide a reflection of the current situation, not an assessment of what will happen in a few months. It is a tool designed to reflect what would happen if the elections were held today. It is important to remember this, and that nothing is set in stone. Many things can change,' explains Dr. Nimrod Nir from the Agam Institute and the Hebrew University of Jerusalem.
'There are two dramatic changes that have occurred in the way polls are conducted in the current election campaign compared to previous ones. There is more use of technology to validate and improve accuracy, and there are many more 'commissioned' polls compared to what there was before. Most of the polls conducted in previous elections were done by established and professional research institutes. Today we see all kinds of research institutes run by people who do not come from the field of statistics and data, but come from media and marketing bodies. One can learn from this that polls have become a tool used today more for influence than for measurement and reflection. A poll has enormous power, especially among swing voters.
'The first thing we should ask ourselves when we look at a poll is whether it represents the population. A poll is very influenced by the issue of representation and the characteristics of the sample - how many women and how many men responded, which districts, Arabs, Jews, and so on. For some reason, the vast majority of polls presented in the media do not include the characteristics of the sample. Sometimes the sample looks demographically representative, but when you look at who the respondents voted for in the previous elections, you see that 80% of them voted for the center-left.
'There are several groups that are more challenging to reach. Druze are one of them. Sometimes you have to reach them physically, both because of digital literacy and because of suspicion. Elderly immigrants from the Soviet Union who speak less Hebrew are another group that poses a challenge. 17% of eligible voters in Israel are Arabs, more than 70% are Jews, and there are another 5-6% who belong to 'others'. This is almost not reflected in the polls. That is why, for example, we always see 'Yisrael Beiteinu' surprising in the elections compared to the polls. These are audiences that require other means to sample them.'
If you missed it: you are invited to listen to another episode of ours, on influence operations and how they might affect the elections.
