Haaretz dodgy data in Livni poll story

A story in Haaretz today, based around a new Dialog poll of Israeli voters, announces “The Left Bloc contracts further“. The Haaretz piece reveals that their new poll shows Livni’s HaTnua party winning 7 seats and claims:

“What Livni has managed to do is to severely damage Yair Lapid’s “Yesh Atid” with the loss of five mandates compared to the last survey, and to a lesser extent Yachimovich’s Labour party which also loses five mandates”

This should make us instantly suspicious. Voters are complicated and voting models more so, but it’s not at all clear why the establishment of Livni’s new party should move votes from Labour to Meretz.

In fact, the story is misleading.

Haaretz chose to compare today’s poll with a Channel 10/Dialog poll published a month ago, on the 28th of October. This poll was quickly commissioned and published as a reaction to the merger of the Likud and Yisrael Beiteinu lists a few hours earlier. It had a relatively low number of seats for Likud Beiteinu (35) and a relatively high one for Lapid’s Yesh Atid (14).

A lot’s happened since then – Operation Pillar of Defence and its conclusion, Barak’s withdrawal, the re-election of Barak Obama, Naftali Bennett’s victory at Jewish Home and the start of the slow announcements of of party lists.

Comparing numbers between different polling companies is very problematic for a number of technical reasons. So while there have been 19 other polls since then, Haaretz is right not to use any of them as its comparator. Any of them except one.

That one is another poll by Dialog. Because it’s from the same company we can use it to compare numbers and look for movement. That poll was published much more recently, on the 19th of November – nine days ago – in, yup, Haaretz.

Using this poll as the basis for comparison you get these figures:

Comparison of the two polls. NB Arab parties were not included in the most recent Dialog poll so have been removed from this table

This shows Livni’s HaTnua party taking just 3 seats from Labour and only one from Yesh Atid. Overall, it shows the “right bloc” contracting by one seat, though all of the movements in other parties apart from Livni’s are within the margin of error, making it difficult to guess at the true picture.

Why did Haaretz choose to use the older data? Was it simply after the sensationalist story or did it have another agenda?

The Times of Israel quickly changed its story once it realised it was based on misleading data, though the headline is still far from ideal. But Haaretz now has form with dodgy reporting of polls. All news organisations should examine their numbers closely in the future.