,
14 tweets,
3 min read
Read on Twitter
Quick thread to explain this rather intemperate response (and apologies for being rather rude in it). Essentially, the problem is that this poll ignores a number of obvious methodological problems which make it very hard to treat it as informative 1/?
2/? First: the poll assumes equal knowledge of the candidates. That's obviously not the case. Boris is far better known than the others. But that wouldn't hold through a leadership campaign, nor would it hold if he were PM. This head start he gets is misleading
3/? Second: The poll assumes equal knowledge of the candidates positions. Brexit party voters likely to be switching more because they know Boris as the "Brexit" and/or "No Deal" candidate while not knowing positions of the others. Another misleading head start
4/? Third: The poll assumes voters are good at predicting their own behaviour in a hypothetical situation. They really aren't. There is a lot of good evidence on that. Just daft to ignore it and pretend people can predict how a new PM will influence how they vote
5/? Fourth (and related to 3): The poll assumes people can predict the impact of a new PM on the political context (and how they would react to that). That's utterly nuts given the year we've just had
6/? Fifth - the production of a projected House of Commons from the poll assumes we know what the pattern of vote changes across seats would be with (a) a new PM (b) a new party competing (c) big shifts in support for all the existing parties. That's not credible either.
7/? Look at it this way: *If* people were well informed, and good forecasters of their own behaviour, and of the political context, and we knew how votes would transform into seats, then this poll would be informative. But none of these things is true.
8/? I understand why media get excited by a poll finding like this. But when they report the results, & none of the shortcomings, they increase the influence of the poll on MPs & members voting in this race. That undermines public debate & incentivises others to make shonky polls
9/? After 2015 & 2017, there was a lot of scrutiny on pollsters' fieldwork methods and why they produced inaccurate polls. But here's the bigger prob - poor understanding of poll methodology among media and politicians, meaning well conducted but badly desinged polls get big play
10/? There's nowt wrong with the fieldwork or representativeness of this poll. Its just badly designed, pretending to give info that its simply not possible for a poll to give. And neither our media, nor our politicians, seem capable on the whole of recognising this shortcoming
11/? I'm far from the only one pointing this out. Lots of other experts in polling and public opinion were saying the same thing last night. Yet, I'm depressingly certain that for the next few days lots of journalists and MPs will wave this poll about.
12/? If we want polling to properly inform public debate then behaviour like this needs to change. We need to be able to separate good polling from bad polling. Regulating pollsters' fieldwork won't help us when the problems lie in design & interpretation. /ENDS/
I have deleted the original tweet (screengrab attached) because it was intemperate, personalised and uninformative, not good enough really, apologies to all. 
For those confused about what this thread is discussing, its the "Boris would get a landslide" poll/reporting from yesterday evening: uk.news.yahoo.com/poll-suggests-…
