Thread by @TomKibasi

THREAD on the @BestForBritain tactical voting tool.

The tool is based on an MRP with a sample size of 46,000 conducted in October 2019.

What are the issues?

(1/6)

The first issue is the notion that it is more important to reflect recent volatility in the polls than real past voting behaviour from 2017.

But by definition, if volatility is high then polls could shift very significantly in the campaign. So why release the tool now?

(2/6)

The second issue is that it doesn’t pass “the sniff test”. Is it really plausible that 25k people in a single seat that never voted LD will do so now?

There would have been merit in a more complex set of rules eg capping the swing size at the largest swing this century.

(3/6)

If that had been done, it would have stopped the model producing implausibly large swings which have seen people take to Twitter with 2017 results that show the LDs 20k votes behind but recommended for Lab/Tory marginals.

It just looks daft or biased as it is.

(4/6)

The third issue is that it makes political judgements where the bar is set at “stop Brexit” not “will vote for a referendum”.

This generates a strong bias towards LD candidates and away from Labour candidates who would back a referendum if its in their party manifesto.

(5/6)

So what next? Well there’s not really any good reason to have a tactical voting tool available until the last couple of weeks of a campaign.

B4B would be better off pulling it for now, making key methodological changes, and relaunching it at the end of November.

(6/6)