> Executive Summary
Citizens’ Assemblies, in which a representative selection of members of the public
are invited to consider policy on contentious areas, are increasingly in the news.
Supporters claim they will enhance public confidence in democracy, and could
also break the deadlock on issues from assisted dying to climate change. They are
claimed to give politicians and policy makers insight into what an informed
common ground might look like.
This paper examines the case for these claims. We have reviewed over 700
initiatives covered in an OECD database, focusing in particular on 17 examples
from Ireland, several US states and two Canadian provinces where the
deliberation of a citizens’ assembly was followed by a public vote on the same
subject. We have then analysed the results in the light of academic literature on
political behaviour and opinion forming.
Our conclusion is that citizens’ assemblies are a poor predictor of what the public
is likely to decide if asked. With the recommendation of citizens’ assemblies
rejected on 10 out of 17 occasions, they are worse at forecasting the public mood
than tossing a coin. Even in cases where assemblies were praised for anticipating
the popular vote, like on abortion or gay marriage in Ireland, the winning margin
at the assembly was around 40 percentage points higher than at the referendum.
The error is consistently in the same direction - assemblies were more supportive
than the general public of progressive policies on 15 of 17 occasions1 and the
proportion of people who voted for the progressive option was, on average, 25
percentage points higher in the citizens’ assembly than in the subsequent
referendum. Even when every effort is made to conduct them robustly, the
structure of these assemblies seems highly vulnerable to a series of biases, in
particular selection bias, issue framing and ‘polarization effects’ – a type of group
think. There is good evidence that the contentious issues for which they are most
often proposed, like assisted dying, might be the very ones for which citizens’
assemblies are least suited.
> Executive Summary Citizens’ Assemblies, in which a representative selection of members of the public are invited to consider policy on contentious areas, are increasingly in the news. Supporters claim they will enhance public confidence in democracy, and could also break the deadlock on issues from assisted dying to climate change. They are claimed to give politicians and policy makers insight into what an informed common ground might look like. This paper examines the case for these claims. We have reviewed over 700 initiatives covered in an OECD database, focusing in particular on 17 examples from Ireland, several US states and two Canadian provinces where the deliberation of a citizens’ assembly was followed by a public vote on the same subject. We have then analysed the results in the light of academic literature on political behaviour and opinion forming. Our conclusion is that citizens’ assemblies are a poor predictor of what the public is likely to decide if asked. With the recommendation of citizens’ assemblies rejected on 10 out of 17 occasions, they are worse at forecasting the public mood than tossing a coin. Even in cases where assemblies were praised for anticipating the popular vote, like on abortion or gay marriage in Ireland, the winning margin at the assembly was around 40 percentage points higher than at the referendum. The error is consistently in the same direction - assemblies were more supportive than the general public of progressive policies on 15 of 17 occasions1 and the proportion of people who voted for the progressive option was, on average, 25 percentage points higher in the citizens’ assembly than in the subsequent referendum. Even when every effort is made to conduct them robustly, the structure of these assemblies seems highly vulnerable to a series of biases, in particular selection bias, issue framing and ‘polarization effects’ – a type of group think. There is good evidence that the contentious issues for which they are most often proposed, like assisted dying, might be the very ones for which citizens’ assemblies are least suited.