Estimating Constituency Opinion in Britain
Lead Research Organisation:
Durham University
Department Name: Government and International Affairs
Abstract
Today in Britain we have a very clear picture of political opinion in the general population. By surveying a nationally representative sample of citizens, researchers regularly make accurate inferences regarding, for example, what proportion of British voters approve of a specific policy, or which political issue is most important to the largest number of Britons.
Yet when we zoom in to look at political opinion at the level of the Parliamentary constituency, the picture is not nearly as clear. For even though they are the basic unit of representation in British politics, each choosing a single Member of Parliament (MP) to send to Westminster, we have very little systematic information about political opinion in individual constituencies. This is because gathering such information using traditional polling techniques is prohibitively expensive: researchers would need to survey a sufficient number of voters in each separate constituency, meaning an extremely large overall sample size.
This all matters because we need to measure political opinion at the constituency-level if we are to answer fundamental questions about the role of constituencies in British representative democracy. For example, is there a meaningful representative link between individual MPs in the House of Commons and their constituents? To what extent do the personal political preferences or political behaviour of individual MPs reflect not just the wishes of the national political party to which they belong, but also political opinion in the constituency they represent? And also, how do changes to constituency boundaries impact upon the electoral fortunes of political parties? So far, research concerning these questions has been hampered by a lack of direct measures of political opinion in constituencies.
Our project will remedy this situation. We will exploit major recent methodological advances which have enabled political scientists in the US and Germany to estimate political opinion for multiple small local areas without having to survey an unfeasibly large overall sample of respondents. Instead, these new methods combine information from several existing, publicly available data sources. First, we use data from the British Election Study, which records the political opinions, socio-demographic type and constituency location of a large number of survey respondents, to estimate a statistical model predicting an individual's political opinions as a function of their socio-demographic type and constituency location. We also include Ordinance Survey data on which constituencies neighbour each other, thus allowing for the tendency of people who live in constituencies near to one another to have more similar political opinions than those who live in constituencies further apart. In the second step we generate an estimate of public opinion in each constituency by combining the predictions of our statistical model with Census-based information on the number of people of each socio-demographic type living in that constituency.
Through this procedure, we will generate the first systematic constituency-specific measures of: public opinion on specific political issues such as the Iraq War or European Union membership; and opinion on over-arching political questions such as the appropriate balance between government taxation and spending or which is the most important issue facing the UK. We will also use our method to generate new constituency-specific measures of electoral support for each political party.
Not only will these new measures allow us to shed new light on important questions concerning the role of the constituency in British representative democracy. They can also help to improve the quality of the link between British citizens and politicians, offering MPs systematic information about the political preferences of their constituents and thus enabling them to be more responsive to the people they represent.
Yet when we zoom in to look at political opinion at the level of the Parliamentary constituency, the picture is not nearly as clear. For even though they are the basic unit of representation in British politics, each choosing a single Member of Parliament (MP) to send to Westminster, we have very little systematic information about political opinion in individual constituencies. This is because gathering such information using traditional polling techniques is prohibitively expensive: researchers would need to survey a sufficient number of voters in each separate constituency, meaning an extremely large overall sample size.
This all matters because we need to measure political opinion at the constituency-level if we are to answer fundamental questions about the role of constituencies in British representative democracy. For example, is there a meaningful representative link between individual MPs in the House of Commons and their constituents? To what extent do the personal political preferences or political behaviour of individual MPs reflect not just the wishes of the national political party to which they belong, but also political opinion in the constituency they represent? And also, how do changes to constituency boundaries impact upon the electoral fortunes of political parties? So far, research concerning these questions has been hampered by a lack of direct measures of political opinion in constituencies.
Our project will remedy this situation. We will exploit major recent methodological advances which have enabled political scientists in the US and Germany to estimate political opinion for multiple small local areas without having to survey an unfeasibly large overall sample of respondents. Instead, these new methods combine information from several existing, publicly available data sources. First, we use data from the British Election Study, which records the political opinions, socio-demographic type and constituency location of a large number of survey respondents, to estimate a statistical model predicting an individual's political opinions as a function of their socio-demographic type and constituency location. We also include Ordinance Survey data on which constituencies neighbour each other, thus allowing for the tendency of people who live in constituencies near to one another to have more similar political opinions than those who live in constituencies further apart. In the second step we generate an estimate of public opinion in each constituency by combining the predictions of our statistical model with Census-based information on the number of people of each socio-demographic type living in that constituency.
Through this procedure, we will generate the first systematic constituency-specific measures of: public opinion on specific political issues such as the Iraq War or European Union membership; and opinion on over-arching political questions such as the appropriate balance between government taxation and spending or which is the most important issue facing the UK. We will also use our method to generate new constituency-specific measures of electoral support for each political party.
Not only will these new measures allow us to shed new light on important questions concerning the role of the constituency in British representative democracy. They can also help to improve the quality of the link between British citizens and politicians, offering MPs systematic information about the political preferences of their constituents and thus enabling them to be more responsive to the people they represent.
Planned Impact
This research will have impact on three identifiable groups: Members of Parliament (MPs); public bodies interested in estimating local opinion; and the public.
First, MPs benefit because they will better understand the opinions of their constituents on a range of specific political issues and also the general political orientation of their constituents. MPs claim to value correspondence with their constituents, and public opinion as revealed `on the doorstep', as means of gauging opinion in their constituency. We therefore believe that they will also value statistical information provided at the level of the individual constituency. We will maximise the impact of our research on MPs through a number of dissemination events in Westminster and through the publication of region-specific booklets giving information on local public opinion. Our dissemination events will enable us to demonstrate, to individual MPs, the state of public opinion in their constituency. Although much public opinion research could be said to benefit MPs by providing information on national public opinion, our impact is qualitatively different because we provide systematic information about political opinion in the individual constituencies MPs represent.
Second, public bodies interested in estimating local opinion benefit because our techniques can make it less costly for them to research local opinion. Public opinion research in general is costly. Generating representative samples at local level is costlier still. Attempts by individual councils are unlikely to be cost-effective. Attempts at national level, such as the Department for Communities and Local Government (DCLG)'s Place Survey, are extremely expensive. Our impact therefore comes through meeting with these organisations and offering them techniques to deliver estimates of local opinion with greatly reduced sample sizes (and thus at a far lower cost). This impact is financial in nature, and is conditional on public bodies commissioning some kind of public opinion research across multiple local authority areas. However, the impact could be considerable if, through MRP+SCR, we could reduce the sample size in each local authority area by a factor of three or four. We note in our Pathways to Impact statement how we will meet with DCLG and suggest ways in which our research might be utilized.
Finally, our research benefits the general public by enriching public discourse on politics in Britain. In order to secure this impact, we have identified a number of regional newspapers to serve as suitable outlets for our findings. Citizens will also be able to use our outputs to find out about how political opinion varies geographically across Britain and about political opinion in their local constituency. The availability of systematic and easily digestible information on these matters will help citizens evaluate how well the political system responds to the will of the public and can help inform public debates concerning major political reforms, such as the recent proposal to introduce the AV electoral system in the UK.
We also note that the general public may benefit indirectly from our aforementioned impact on MPs. Survey evidence indicates that the general public want MPs to pay more attention to the opinions of their constituents when making political decisions (Kellner 2012). Thus, if MPs change their behaviour in response to information about constituents' views, then our research will benefit the public through its impact on MPs.
First, MPs benefit because they will better understand the opinions of their constituents on a range of specific political issues and also the general political orientation of their constituents. MPs claim to value correspondence with their constituents, and public opinion as revealed `on the doorstep', as means of gauging opinion in their constituency. We therefore believe that they will also value statistical information provided at the level of the individual constituency. We will maximise the impact of our research on MPs through a number of dissemination events in Westminster and through the publication of region-specific booklets giving information on local public opinion. Our dissemination events will enable us to demonstrate, to individual MPs, the state of public opinion in their constituency. Although much public opinion research could be said to benefit MPs by providing information on national public opinion, our impact is qualitatively different because we provide systematic information about political opinion in the individual constituencies MPs represent.
Second, public bodies interested in estimating local opinion benefit because our techniques can make it less costly for them to research local opinion. Public opinion research in general is costly. Generating representative samples at local level is costlier still. Attempts by individual councils are unlikely to be cost-effective. Attempts at national level, such as the Department for Communities and Local Government (DCLG)'s Place Survey, are extremely expensive. Our impact therefore comes through meeting with these organisations and offering them techniques to deliver estimates of local opinion with greatly reduced sample sizes (and thus at a far lower cost). This impact is financial in nature, and is conditional on public bodies commissioning some kind of public opinion research across multiple local authority areas. However, the impact could be considerable if, through MRP+SCR, we could reduce the sample size in each local authority area by a factor of three or four. We note in our Pathways to Impact statement how we will meet with DCLG and suggest ways in which our research might be utilized.
Finally, our research benefits the general public by enriching public discourse on politics in Britain. In order to secure this impact, we have identified a number of regional newspapers to serve as suitable outlets for our findings. Citizens will also be able to use our outputs to find out about how political opinion varies geographically across Britain and about political opinion in their local constituency. The availability of systematic and easily digestible information on these matters will help citizens evaluate how well the political system responds to the will of the public and can help inform public debates concerning major political reforms, such as the recent proposal to introduce the AV electoral system in the UK.
We also note that the general public may benefit indirectly from our aforementioned impact on MPs. Survey evidence indicates that the general public want MPs to pay more attention to the opinions of their constituents when making political decisions (Kellner 2012). Thus, if MPs change their behaviour in response to information about constituents' views, then our research will benefit the public through its impact on MPs.
Organisations
People |
ORCID iD |
| Nick Vivyan (Principal Investigator) | |
| Chris Hanretty (Co-Investigator) |
Publications
Hanretty C
(2016)
Comparing Strategies for Estimating Constituency Opinion from National Survey Samples
in Political Science Research and Methods
Hanretty C
(2016)
Dyadic Representation in a Westminster System
in Legislative Studies Quarterly
Hanretty C
(2019)
An Introduction to Multilevel Regression and Post-Stratification for Estimating Constituency Opinion
in Political Studies Review
Hanretty, C.J.
(2014)
Constituency Opinion in Britain: Technical Report
Hanretty, C.J.
(2014)
Sex, Lies and the Ballot Box: 50 Things You Need to Know About British Elections
| Description | Our project had two over-arching objectives. Our methodological objective was to validate new estimates of political opinion in Westminster parliamentary constituencies. Our substantive objective was to utilise these new estimates to gain new insights into the role of the constituency in British politics. On each front, we have made major contributions. First, we made a methodological contribution by showing that accurate estimates of political opinion in Westminster constituencies can be obtained using multilevel modelling and post-stratification (MRP) to combine national survey samples of 4,000 or more with information from Census data and geodata. Our evidence comes from two validation studies: an external validation, where we compared MRP estimates of parties' constituency vote shares with actual results at the 2010 general election; and a cross-validation, where we split a very large survey sample and compared MRP estimates of constituency opinion based on one portion of the sample with `true' opinion in the remainder. Second, we made a methodological contribution to the MRP literature more broadly. MRP can potentially involve several different elements, each drawing on different types of demographic or geographic information. For researchers with limited resources and access to data, it is important to understand how much each MRP element is likely to improve constituency opinion estimates. Compared to previous validation studies, ours assess a more comprehensive set of MRP elements. We find that although using more elements is generally better, in a setting like the British one, where the number of constituencies is large, the largest and most consistent improvement in MRP performance comes from including constituency-level predictors. This is encouraging: researchers can achieve large marginal gains in the quality of constituency opinion estimates through a relatively low-cost search for relevant constituency-level data (e.g average age or earnings). Third, we made a substantive contribution by providing a new mapping of the geography of political opinion in Britain. Combining survey data from the British Election Study, population data from the UK Census, and geodata from Ordinance Survey, we have generated new estimates of 2014 constituency opinion on three key topics: income redistribution; Britain's membership of the European Union; and the cultural impact of immigration. These new estimates of constituency opinion improve our understanding of demographic and geographic patterns in popular views on highly salient political topics. Fourth, we made a broader substantive contribution by providing new evidence of a representational bond between MPs and their constituents. Across four cases (including whipped and un-whipped parliamentary behaviour), we show that the policy positions MPs take respond to opinion in their constituencies. This association between constituency opinion and MP position holds even when we look at MPs from within the same party. This suggests that constituency influence does not operate solely through choice of party, but that MPs are at least to some extent individually responsive to their constituents. Our results suggest the MP-constituent link is an under-appreciated channel for policy representation in Britain, and have implications for debates about electoral reform. |
| Exploitation Route | First, our validation evidence regarding the relative contribution of different MRP elements informs future methodological work aimed at improving MRP methods. It also provides guidance for applied researchers estimating constituency opinion in settings similar to the British one. Second, our new estimates of constituency opinion will be useful as explanatory variables for academics researching constituency-level phenomena including electoral outcomes, party campaign strategies, candidate positioning, and various aspects of MP behaviour (e.g. Commons votes and speeches). The estimates also offer MPs and prospective parliamentary candidates new information about the policy preferences of their constituents. This supplements the potentially less representative information that such actors receive through mailbags and canvassing, and can help politicians better understand those they represent. Third, our finding that MPs are individually responsive to constituency opinion can be taken forward by academics and non-academics. Future academic research can look into whether such responsiveness is driven more by selection mechanisms or by re-election concerns. Our approach can also be used to test for responsiveness in other political systems. From a non-academic perspective, our evidence that there is some MP-constituency policy linkage can inform future debates about electoral reform in Britain, particularly with regard to the maintenance of single-member districts. |
| Sectors | Government, Democracy and Justice |
| URL | http://constituencyopinion.org.uk |
| Description | This research project has generated three types of impact: first, informing MPs about political opinion in their constituencies; second, shaping public understanding of the geography of political opinion in Britain; third, informing methods used by commercial pollsters. We describe each of these in turn. Our first type of impact has been achieved through providing new information to Members of Parliament concerning public opinion in their respective constituencies. Our research generated new estimates of public opinion in each Westminster constituency on a number of policy topics. These estimates have been valuable to MPs wishing to gather information about the views of those they represent, as there were previously no publicly available, systematic and comprehensive measures of public opinion at the constituency level. To ensure that MPs were aware of and had access to our constituency opinion estimates, we collaborated with the House of Commons Library, who hosted a dissemination event entitled "Constituency Opinion and Representation in Britain" at Westminster in October 2014. At this seminar, we presented our methods and findings to an audience made up of MPs' aides, journalists (from national newspapers and national broadcasters), think tank employees and Commons Library staff. As a follow-up to this, we produced and printed a briefing booklet summarising our findings for a general audience. This booklet was mailed to all MPs, and was also posted on our website. Additionally, the investigators took care to mention the estimates in personal meetings with MPs. We have several reasons to be confident that our constituency opinion estimates have been useful for MPs. This is indicated, first, by the fact that the Commons Library - a body whose main purpose is to provide "impartial information and research services for Members of Parliament and their staff in support of their parliamentary duties" - recognised the relevance of our research and was willing to support dissemination is in itself a form of impact. A second indication of the utility of our estimates is the fact that staff from MPs' offices took the time to attend our dissemination event. A third, yet stronger indication of utility is that we have received a number of direct requests from MPs' offices for further information on our estimates. A fourth indicator is that the Commons Library has also reported several requests for further information from MPs after our dissemination event. These all indicate that our constituency opinion estimates have been of use to MPs at Westminster. Our second type of impact has been to shape public understanding of the geography of political opinion in Britain. Given the absence of pre-existing measures of public opinion in each parliamentary constituency, our project outputs have been able to inform media and public discussions by, for example, providing new evidence about which parts of the country are more or less Eurosceptic, or more or less opposed to immigration. We have achieved significant public engagement in two ways. First, we engaged with national and regional media to maximise the chances that our findings reached a broad audience. For example, in October 2014 our findings were the focus of a page 2 article in The Times newspaper (print and online readership, 1,006,000), which also included on its website an interactive map showcasing our data. In addition to this, our estimates were reported by the Financial Times FT Data blog (the FT has over 500,000 online subscribers), the Press and Journal, and on regional television shows such as ITV's Around the House and Anglia Late Edition. We also published a summary chapter on our constituency opinion estimates in the popular social science book "Sex, Lies and the Ballot Box" (Eds. Cowley, P. and Ford, .R, Biteback 2014). Second, we developed a website, constituencyopinion.org.uk. This website, which was designed for a general audience, housed a blog, data on our opinion estimates, and accompanying documentation. Importantly, it also hosts (heavily annotated) code, which can be downloaded and adapted by anyone who wants to generate new constituency opinion estimates from publicly available data. To date the website has had more than 30,000 visits and continues to garner upwards of 1500 visits per month. There is also a final, indirect, way in which our research has had an indirect impact on public discourse concerning the geography of political opinion in Britain: the methods tested, codebase generated and data assembled during the project have all contributed to the subsequent development of the statistical model underlying the electionforecast.co.uk website. This website, developed by the PI and Co-I together with Benjamin Lauderdale, provided constituency-level forecasts of the results of the 2015 General Election. Not only did the site receive significant public attention (over 1 million visits), but the forecasting model was used to provide regular in-depth analysis for BBC's flagship current affairs programme, Newsnight, and for the popular US website fivethirtyeight.com. The resulting features helped inform the public as to the geography of political opinion in Britain, and about the consequences of the British electoral system. The predictions were also featured in the Telegraph and Financial Times. Our third type of impact is on the methods used by commercial survey research companies. This grant continues to deliver commercial benefits for one large public opinion research firm. Following on from the early grant-funded proofs-of-concept and applications of MRP to the United Kingdom, co-I Hanretty began a commercial relationship with Survation. This relationship has led to a three year partnership agreement, as a result of which Survation will fund a PhD studentship in small area estimation. Hanretty continues to produce estimates of constituency and local authority opinion for Survation, which have benefitted a range of end users and Survation clients. The most public demonstration of the potential for MRP was shown in the Channel 4 programme, "Brexit: What the Nation Really Thinks", where Hanretty and Survation produced estimates of how each local authority would vote in a second referendum, and how each local area viewed key issues in relation to Brexit (approval for the Prime Minister's proposed deal; views on freedom of movement, and so on). |
| First Year Of Impact | 2014 |
| Sector | Government, Democracy and Justice,Other |
| Impact Types | Societal |
| Title | Code to generate constituency opinion estimates |
| Description | We have made a available code (with detailed annotation) that allows users to generate constituency opinion estimated within the open source software R and WinBUGS. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2014 |
| Provided To Others? | Yes |
| Impact | This code enables anyone to generate new estimates of constituency opinion in Britain using new sources of survey data. |
| URL | http://constituencyopinion.org.uk/wp-content/uploads/2013/05/code_documentation.zip |
| Title | Data on constituency opinion estimates |
| Description | This data set contains the new estimates of constituency opinion we have generated in our project. For each opinion, we include the point estimate for average opinion in the each constituency, as well as the lower and upper bounds of the 95% confidence interval. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2014 |
| Provided To Others? | Yes |
| Impact | Our data has been utilised by academics and journalists, including authors at the Financial Times FT Data blog and The Times. |
| URL | http://constituencyopinion.org.uk/data/ |
| Description | Anglia Late Edition |
| Form Of Engagement Activity | A press release, press conference or response to a media enquiry/interview |
| Part Of Official Scheme? | No |
| Geographic Reach | Regional |
| Primary Audience | Public/other audiences |
| Results and Impact | Dr Hanretty gave an interview on constituency-level Euroscepticism in East Anglia. The findings informed a panel debate amongst local MPs and PPCs. |
| Year(s) Of Engagement Activity | 2014 |
| Description | Commons Library Seminar |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | Yes |
| Geographic Reach | National |
| Primary Audience | Policymakers/politicians |
| Results and Impact | The talk triggered several questions and discussion afterwards, concerning our method, the constituency opinion estimates we have produced, and our findings regarding the MP-constituency policy link. The talk was very recent so some impact are yet to be felt. However, the Commons Library have reported that they have received a number of requests for further information about our research in the weeks since the event. |
| Year(s) Of Engagement Activity | 2014 |
| Description | Financial Times Data Blog -- Attitudes to migration and where migrants live |
| Form Of Engagement Activity | A press release, press conference or response to a media enquiry/interview |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Public/other audiences |
| Results and Impact | The FT Data blog utilised our constituency opinion estimates to look at the relationship between attitudes to immigration and levels of immigration across constituencies. This blog post relied heavily on the new data we had produced during our project and had made available publicly. The author utilised our new estimates with other information on levels of immigration in constituencies to draw new insights concerning the immigration debate in the UK. In this sense, our project output directly influenced the output of a major news outlet which reaches a global audience: the FT Data blog is part of the Financial Times website, a site which had over 500,000 subscribers in 2014. |
| Year(s) Of Engagement Activity | 2014 |
| URL | http://blogs.ft.com/ftdata/2014/11/19/public-attitudes-to-migration/ |
| Description | ITV Around The House |
| Form Of Engagement Activity | A press release, press conference or response to a media enquiry/interview |
| Part Of Official Scheme? | No |
| Geographic Reach | Regional |
| Primary Audience | Public/other audiences |
| Results and Impact | Dr Vivyan gave an interview explaining our estimates of constituency opinon in the North East of England. Our estimates were discussed by a local MP. |
| Year(s) Of Engagement Activity | 2014 |
| Description | LSE British Politics and Policy Blog |
| Form Of Engagement Activity | A magazine, newsletter or online publication |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Public/other audiences |
| Results and Impact | Dr Hanretty authored a post on the LSE British Politics and Policy blog about our constituency opinion estimates. The blog is a very popular resource for those with an interest in British politics. |
| Year(s) Of Engagement Activity | 2014 |
| URL | http://blogs.lse.ac.uk/politicsandpolicy/how-eurosceptic-is-rochester-and-strood/ |
| Description | Press and Journal Article |
| Form Of Engagement Activity | A magazine, newsletter or online publication |
| Part Of Official Scheme? | No |
| Geographic Reach | Regional |
| Primary Audience | Public/other audiences |
| Results and Impact | An article reporting on our constituency Euroscepticism estimates was published in the Press and Journal on 24 October 2014. This paper has a circulation of over 60,000 in 2013. |
| Year(s) Of Engagement Activity | 2014 |
| URL | https://www.pressandjournal.co.uk/fp/news/politics/westminster/380210/most-scots-want-to-stay-in-the... |
| Description | Times article 1 |
| Form Of Engagement Activity | A magazine, newsletter or online publication |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Public/other audiences |
| Results and Impact | An article on our research project was published on page 2 of The Times newspaper on 21 October 2014, accompanied by an online interactive map for people to look at constituency opinion estimates in more detail. After the article, we have received several requests for further information from both the public and from other journalists. |
| Year(s) Of Engagement Activity | 2014 |
| URL | http://www.thetimes.co.uk/tto/news/politics/article4242712.ece |
| Description | Times article 2 |
| Form Of Engagement Activity | A magazine, newsletter or online publication |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Public/other audiences |
| Results and Impact | An article reporting on the differences in Euroscepticism between English, Welsh and Scottish constituencies, which was based on our constituency opinion estimates, was published on the front page of The Times Scottish edition on 23 October 2014. We had numerous requests for further information. |
| Year(s) Of Engagement Activity | 2014 |
| URL | http://www.thetimes.co.uk/tto/news/uk/scotland/article4245089.ece |
| Description | constituencyopinion.org.uk |
| Form Of Engagement Activity | A magazine, newsletter or online publication |
| Part Of Official Scheme? | No |
| Geographic Reach | National |
| Primary Audience | Public/other audiences |
| Results and Impact | Our blog disseminated information about our method of constituency opinion estimation and the results of our study. Our blog has triggered interest from newspapers and from the Commons Library. |
| Year(s) Of Engagement Activity | 2013,2014 |
| URL | http://constituencyopinion.org.uk |