Thread by @robfordmancs

This is doing the rounds but I'm really not sure we can conclude much/anything from it. People aren't good at estimating large numbers, or proportions of large numbers. It is a task that requires some thought, and survey respondents don't have an incentive to think hard 1/?

https://twitter.com/LukeTryl/status/1467778968589901826

2/? If you give people a hard question and no incentive to do the work, they will substitute an easy question. In this case, "what am I aware of that the government spends money on?" People are aware of the NHS, net zero and MPs' pay, so they score high

Pensions and benefits are low salience in debate, so that scores low. Also, note how many things average around 10% - a low effort heuristic number probably used to convey "I think they spend quite a lot on this, but no idea how much precisely".

4/? So an alternative summary of the findings is "Public think govt spends a lot on NHS, and will whack 10% on any category you ask them about, without really thinking about it". Not really a very important finding.

5/? In general I think detailed qs like this which require a lot of the respondent are only worth asking if you incentivise people to think hard about them (e.g. pay them based on how close they get to the correct answer). Otherwise all you get is guesses & expressive politics

Also, as Graeme points out, the percentages seem to sum up to well over 100% so it seems respondents didn't have to make their figures sum to 100 (whereas the real budgets do). Even easier to just click through going "Education? 10% Police? 10% MPs salaries? 10%..."

CORRECTION - Luke has advised the answers did have to sum to 100, but the averages don't have to. That would suggest an awful lot of spread in the underlying answers - my guess is partisan biases and expressive politics will be driving a lot of that.

Or some people with no clue what to do just allocate equal shares to everything asked about, per this suggestion:

https://twitter.com/stejormur/status/1467793659332546560