In the spirit of constructive criticism, here’s some advice on how Tom could’ve done this better. https://twitter.com/tomhunt1988/status/1300022055866380288

First, let’s define the task at hand: Tom wants to know what his constituents think about the TV licence fee.

To try to work this out, Tom sent out a push poll that anyone could answer, asked them to respond and provide a postcode, then stripped out the non-Ipswich postcodes.

Two problems: 1st, we don’t know if these people actually do live in Ipswich even where they submit an appropriate postcode. 2nd, the people that Tom can most easily push his poll to likely share his conservative views (hence the result).

I expect that Tom’s constituency doesn’t contain that many households, maybe somewhere in the low tens of thousands. I expect he also knows which postcodes are in his constituency.

If so, a better way to judge what his constituents think would be to send out a survey to a sub sample of them at random, using the postcode address file or something similar to get a list of addresses.

Randomisation is this case is really useful. In effect, it breaks the link between Tom and the survey’s results, as he would have no influence over who gets to respond to the survey.

What’s more, it would also give you an unbiased estimate of opinion in the constituency at the same time. It’s amazing really. It works like magic.

One last thing, a survey like this is a lot of work, but I see no reason why you couldn’t administer it out of a constituency office. All you need is the raw data and a copy of excel and you’re good to go.

So, to sum up, it’s nice that MPs care about what their constituents think, but it would be even better if they did so in a statistically-sound way.