Here’s what he said on Twitter:
Me: My super-rough rule of thumb is to look at the average sample size in each seat (about 130 in this case); see how many multiples you need to get up to 1,000 (8 in this case); and then concluded that if a factor is present in that many seats then MRP should be able to cope.
Ben: That is generally the calculation I do as well!
The understanding which MRP-based data can give about individual constituencies can be supplemented, of course, by constituency polling.
Get polling news and analysis by email
Sign up here if you would like to receive the Polling UnPacked occasional email newsletter, highlighting the best in analysis and news about British political opinion polling from a carefully curated range of high-quality sites (no more than one email a day and usually much less frequent):