Citizens' Assembly for Frome

Here’s an initial description of the problem from Rob C:

I’m excited about the possibility that you Frome techies may be able to help us out - effectively with a home made sortition process. Although this is specifically to help with getting a Citizens’ assembly running on the Climate Emergency … if we can crack the process, it would help with further CAs or similar kinds of events. The basic idea is to find a way of arriving at somewhere between 20-50 people in a room discussing an issue (in an informed and facilitated way). The process of selecting those people needs to be demonstratively fair and achieve a convincingly representative democratic profile of the Frome population.

I think the task is this:

Optical character read the pdf version of the Electoral register into either a data base, or spreadsheet (approx. 205,000 entries)

Randomly select approx. 1750-2000 names and addresses. [letters then get sent out inviting participation in a CA]

We’d apparently normally expect approx. 3% positive response rate to those 2000 letters … the letters would ask for further information, i.e. confirming gender, nationality, educational background, etc, .

Selective randomisation selection on those 60 responses would then aim to provide a convincingly representative democratic profile of the Frome population.

We’d also need to find some way of coming up with a accurate demographic profile of the people in Frome. Not sure how we set about that.

and here’s a snapshot of the whiteboard that came out of our initial conversation:

1 Like

This is actually quite an interesting statistics problem because of the fact it requires the response of the 2000 people that are notified.
They will have to consider that people will in essence self select and it will introduce a skew in the final set of people that replied.
I am assuming the goal is to have a jury that is as demographically balanced as possible?

What would make more sense is to build a profile of the response rates as a function of the demographic (it might be harder to get one group involved than another) So when sending the requests for participation more can be sent to the groups that tend to respond less and and thus the final outcome should be more balanced.

Interesting way to slice! One thing which came up while we were discussing this is that while it’s tempting to come up with ideas about how to make this work, there must be studies by people with expertise who will know more about various failure modes. And indeed, Rob shared and we found some resources with some case studies - one approach which appealed to me was to select a sufficient set (perhaps with spares) and to perform home visits, educating and persuading people of the importance of taking part. If you can get a much higher hit rate, then you don’t have to worry quite so much about self-selection.

See below for some resources which turned up - the state of the possible collaboration with Frome Town Council is that the ball is in their court, while they seek funding and advice and assistance.

There’s a 36 page PDF here, with case studies: Citizens Participation Using Sortition, A Practical Guide.

A site with info about stratified sampling, and more besides:

Brett Hennig’s site has lots of resources:

The Sortition Foundation looked like it might have useful resources:

This was quite interesting to read about. I do really like the idea of sortition given its a reasonable solution to the issue of corruption and lobbying.

I was wondering, how about a double page spread (around £440.00 for the council) in the “Frome Times” (14,800 distributed) of an infographic, example below.

The infographic would show and tell what Frome Town Council needs from (variety) Fromeians* in order to set up a fair and balanced “Citizens’ Assembly” . Also, add a web address with more indepth info, if required?

Not sure how much an infographic would cost, but the savings on stamps, envelopes, paper/printing and distribution would be thousands (2000 stamps = £781.00 at a business discount).


Link back to Trello card:

1 Like