British Election Study - 2005 Harold Clarke, David Sanders, Marianne Stewart, Paul Whiteley.

Slides:



Advertisements
Similar presentations
Internet Surveys and Political Attitudes: Evidence from the 2005 British Election Study David Sanders, Harold Clarke, Paul Whiteley and Marianne Stewart.
Advertisements

Voters and Voter Behavior Chapter 6
Measuring Turnout – Who Voted in 2010? British Election Study Harold Clarke, David Sanders, Marianne Stewart, Paul Whiteley, University of Essex.
Random effects as latent variables: SEM for repeated measures data Dr Patrick Sturgis University of Surrey.
Introduction to Statistics: Political Science (Class 7) Part I: Interactions Wrap-up Part II: Why Experiment in Political Science?
The images of the main party leaders are important in influencing voters. Trust in the party leader and their perceived ability to become prime minister.
THE 2009/10 BRITISH ELECTION STUDY RESEARCH OPPORTUNITIES TSEPOP INITIATIVE Harold Clarke David Sanders Marianne.
United States Politics 1 Public Opinion and Polling.
Bush's lead gets smaller in poll By Susan Page, USA TODAY WASHINGTON — President Bush leads Sen. John Kerry by 8 percentage points among likely voters,
Non-Experimental designs: Surveys & Quasi-Experiments
Improving Advertising Conversion Studies Chapter 47 Research Methodologies.
Modelling Campaign Effects in the 2005 British Election Study Harold Clarke, David Sanders, Marianne Stewart and Paul Whiteley The 2005 British Election.
Conventional Participation Voting Contact officials Attend meetings Attend political rallies Contributing to a campaign Volunteering in a campaign Running.
Survey Research & Understanding Statistics
Media Effects and Public Opinion GV917. Mediated and Unmediated Opinions Mediated opinions are those which come from the mass media rather than personal.
An Introduction to Logistic Regression
5. The Structuring of Beliefs. Are Many Opinions Random? (re Converse) GV917.
Internet Experiments in the 2005 BES David Sanders Harold Clarke Marianne Stewart Paul Whiteley.
BES 2005 The 2005 British Election Study Principal Investigators Harold Clarke David Sanders Marianne Stewart Paul Whiteley RO: Kristi Winters RA: Paul.
Sampling Theory and Surveys GV917. Introduction to Sampling In statistics the population refers to the total universe of objects being studied. Examples.
Survey Design. Models, Questions, Hypotheses Model of Vote and Voters Vote Choice = Party Identification + Issues + National Conditions + Ideology + Candidate.
What is an Opinion Survey or Poll?
Electoral Choice in Britain, 2010: Emerging Evidence From the BES Harold Clarke David Sanders Marianne Stewart Paul Whiteley June 25, 2010 ©Copyright 2010:
Voting Behavior of Naturalized Citizens: Sarah R. Crissey Thom File U.S. Census Bureau Housing and Household Economic Statistics Division Presented.
PUBLIC OPINION AND POLITICAL SOCIALIZATION
Public Opinion.
Turnout and Elections BC Outreach Zach Swannell Political Science UBC.
BULLSEYE VOCABULARY UNIT 2. Political Culture, Political Socialization, Particiapation Good Luck on your Test!!!!
EVAL 6970: Cost Analysis for Evaluation Dr. Chris L. S. Coryn Nick Saxton Fall 2014.
Public Opinion. Prediction…. What is the role of public opinion in American Democracy?
THE COMPARATIVE STUDY OF ELECTORAL SYSTEMS (CSES) The CSES After Twenty Years: Three Modules Going On Four/Five, What Progress? Jack Vowles Victoria University.
Party and the Tea Party Among Aiken County Voters Bob Botsch Political Science USC Aiken.
Issues, the Economy and Character in Campaigns March 23, 2011.
Happiness and Elections: Preliminary Slides from the ALP February 22, 2009.
American Government and Politics Today Chapter 6 Public Opinion and Political Socialization.
Public Opinion and Political Action Chapter 6. Introduction Some Basics: Demography The science of population changes. Census A valuable tool for understanding.
120 Exchange Street Portland Maine 1 October 2010 Maine Voter Preference Study – Wave III Prepared for: Maine Today Media October.
The Political Economy of Labour Support David Sanders Department of Government University of Essex, UK.
Project Team Daniella Aryeh Shana Hecht Kinga Kowalewska Math 110 Final Report – May 9, 2011 Survey and Analysis for Voting!
The 2005/06 British Election Study David Sanders Paul Whiteley Harold Clarke Marianne Stewart.
Outline 1. Definition 2. When and why to use surveys
Non-Experimental designs: Surveys Psych 231: Research Methods in Psychology.
Public Opinion and Political Action Chapter 6. Introduction Public Opinion – The distribution of the population’s beliefs about politics and policy issues.
Walt Borges University of Texas at Dallas Harold Clarke University of Texas at Dallas, University of Essex
Public Opinion and Political Action Chapter 6. Introduction Public Opinion – The distribution of the population’s beliefs about politics and policy issues.
28 April Crawford School 1 Causality and Causal Inference Semester 1, 2009 POGO8096/8196: Research Methods Crawford School of Economics and Government.
Campaigns and Elections. The Structure of American Elections Structured to limit popular control and minimize chances of factions controlling government.
Outline 1.Definition 2.Three major characteristics 3.When and why to use surveys 4.Two distinct forms of survey research 5.Advantages of survey research.
© 2011 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part, except for use as permitted in a license.
Jane Green and Will Jennings Universities of Manchester and Southampton Valence Politics: How Competence Matters to Voters, Parties and Governments How.
American Government and Politics: Deliberation, Democracy, and Citizenship Chapter Eight Public Opinion and Political Participation.
How are Opinions Influenced by Political Campaigns?
Public Opinion and Political Action Chapter 6. Introduction Public Opinion – The distribution of the population’s beliefs about politics and policy issues.
2 February, Office hours  Georgios Tuesdays 1-2; 4-5 (A236B)  Siim Thursday 2pm-3pm (A232) Friday 10am-11am (A232)
Research in Psychology A Scientific Endeavor. Goals of Psychological Research Description of social behavior Are people who grow up in warm climates different.
Political Science American Government and Politics Chapter 6 Public Opinion.
Receive-Accept-Sample Model an information-processing model GV917.
Multivariate Statistics Latent Growth Curve Modelling. Random effects as latent variables: SEM for repeated measures data Dr Patrick Sturgis University.
Introduction Sample surveys involve chance error. Here we will study how to find the likely size of the chance error in a percentage, for simple random.
Stephen Duginger | Kylee Britzman Department of Political Science | College of Liberal Arts & Sciences | University of Illinois at Urbana-Champaign The.
Britain Says NO: Voting in the 2011 AV Ballot Referendum Paul Whiteley Harold Clarke David Sanders Marianne Stewart.
The British Election Study Harold D. Clarke (Essex and UT Dallas) David Sanders (Essex) Marianne Stewart (Essex and UT Dallas) Paul.
A Comparison of Two Nonprobability Samples with Probability Samples
Nicole R. Buttermore, Frances M. Barlas, & Randall K. Thomas
Jonathan Mellon and Geoffrey Evans
Electoral Choice in Britain, 2010: Emerging Evidence From the BES
The Fundamentals of Political Science Research, 2nd Edition
WHAT IS PUBLIC OPINION? DEFINITIONS
Participation & Voting Behavior
Presentation transcript:

British Election Study Harold Clarke, David Sanders, Marianne Stewart, Paul Whiteley

Face to Face Probability Surveys - The Gold Standard? Probability Samples Yield Lists Not Achieved Samples Non-Response is a growing problem in all probability surveys (61 percent response rate for 2005 BES) respondents are unlikely to be a random subset of the list – ‘top-ups’ do not help They are very expensive (The American National Election Study in 2004 cost $5500 per voter interviewed) They are slow moving – two to three months to get the data

Face to Face Probability Surveys - Additional Issues There is a limited ability to do panel surveys – when these are required for causal inference They have limited ability to do experiments when field experiments are increasingly being utilised in social science research The costs mean that we get one study only – replication is problematic The have limited or no ability to study voters in the inter- election period where a lot of the interesting action takes place They can’t be used to study election campaigns when these are becoming more important for understanding election outcomes.

Is the Internet an Alternative? Fast responses to internet surveys – typically 95 per cent within a couple of days. This is ideal for campaign studies focusing on short-term dynamics They are very cost effective – in the BES significant savings are obtained in terms of the costs per interview Large N’s are possible – the marginal cost of adding a respondent is small Experiments are possible with large treatment groups No interviewer effects – which can be a problem when interviewers are socially homogenous BUT – they are not random samples. In the BES they are a type of quota sample

BES 2005 CORE FACE-TO-FACE PANEL SURVEY: Wave 1 Pre-election Probability Sample, Face-to-Face N= Primary Sampling Units Wave 2 Post-election Probability Sample, Face-to-Face N=4161 Including top-up, mail-back; 128 Primary Sampling Units Wave 1 Pre-campaign Baseline Survey N=7793 BES 2005 INTERNET CAMPAIGN PANEL SURVEY: Wave 2 Campaign survey 275 interviews per day for 30 days N=6068 Wave 3 Post-election Interview N=5910 Wave 4 One Year Out Interview N=6186 Waves 5, 6, 7 Annual interviews through to 2010

Comparisons of Survey Modes Election Studies are unique in allowing us to compare survey data with actual votes. Two objective measures are highly visible – turnout and vote choice If we compare in-person probability surveys with internet surveys what do we see?

Turnouts in the 2001 and 2005 BES surveys (RDD in 2001; Internet in 2005)

All Types of Survey Over-Report Turnout

How Turnout influences response rates

Vote Choice in the 2005 BES Surveys

What About Modelling the Vote? This involves a variety of different models reflecting theoretical debates in the literature One set of variables relates to social backgrounds of respondents – principally their social class Spatial Issues – issues over which the voters and parties disagree (eg Taxation and spending) Valence Issues – issues over which there is widespread agreement (eg reducing crime) but differences arise over which party can do the best job. Partisanship – the ‘Brand Loyalty’ of the parties Leadership Evaluations – Who do you like and dislike? Campaign effects

Logistic Regressions Panel A Panel B Liberal Other Predictor Variables Labour Conservative Democrat Party Age -.01**.02**.01*.02* Ethnicity -.97*** ** 2.89** Gender ** Social Class -.38*.85*** Party Identification: Conservative -1.09*** 1.56*** Labour.91*** -1.72*** -.88*** -.33 Liberal Democrat -1.51*** ***.89* Other Party -.98** *** Party Leader Affect: Blair.41*** -.50*** -.41*** -.38*** Howard -.13***.58*** * Kennedy -.31***.04.49*** -.01 Party Best on Most Important Issue: Conservative -.91*** 1.42***.08.79* Labour.80*** -.84** -.72*** -.80* Liberal Democrat -.66* -1.21*.84** -.89 Other Party * Party-Issue Proximities: Conservative -.09***.20***.07**.15*** Labour.15*** -.14*** -.12*** -.19*** Liberal Democrat -.12** ***.10 Economic Evaluations *** Party Best on Economy.94*** -1.28*** -.67*** -.42 Iraq Evaluations Emotional Reactions: Economy Iraq * NHS.13* -.24** ** Tactical Voting -.34x *.50 Constant 2.53* *** -5.73** McFadden R 2 = McKelvey R 2 = % Correctly Classified = Lambda =.68.70

Factors Affecting the Probability of Voting Labour in the 2005 Election

The Performance of Rival Labour Voting Models (Logistic Regressions) with In-Person and Internet Surveys McFadden R2 McKelvey R2 AIC BIC A. Models Estimated Using In-Person Survey Data Social Class All Demographics Economic Evaluations Issue Proximities Most Important Issue Party Identification Leader Images Composite Model B. Models Estimated Using Internet Survey Data Social Class All Demographics Economic Evaluations Issue Proximities Most Important Issue Party Identification Leader Images Composite Model

Does Mode Make a Difference? Using the Parameters of the In-Person Model to Predict Labour Voting in the Internet Model in 2005 (and Vice Versa)

Conclusions There are differences in frequency distributions of variables measured with in-person and internet surveys Sometimes the in-person version is more accurate (eg turnout) sometimes not (eg Labour vote choice) The coefficients of models of turnout and party choice are indistinguishable from each other when models are estimated with in-person and internet surveys

Internet Experiment – ‘Feedback to Respondents ’ Key Issue – Are Voter Preferences Exogenous? If they are then survey respondents should not change their preferences simply because others differ eg. Parties and leaders If they are not then voters are likely to change their preferences if they discover that others differ

Feedback to Respondents The BES campaign panel survey, N=7793 Respondents place themselves on proximity scales ‘taxation versus spending’ and ‘crime reduction versus rights of the accused’ For example: __________________________________ Raise taxes and spend more Cut taxes and spend less

Examples of FeedbackTreatments Treatment 1: Control - Respondent’s position only Treatment 2: Respondent and ‘average voter’ Treatment 3: Respondent and ‘party supporters’ Treatment 4: Respondent and ‘party leaders’ Treatment 5: Respondent and ‘party leaders with party labels’

Percentages Wanting to Move

Absolute Distances Moved on the Tax and Spend Scale

Conclusions Respondents change their preferences on issues if given feedback Cues relating to political parties are particularly likely to move them If preferences are not exogenous then spatial models of party competition are wrong Much of neo-classical economic theory is wrong too!

Summary – Advantages of Internet Surveys Huge N’s Quick & Flexible– not tied to slow-moving in-person data collection Clean administration of surveys – avoiding priming and interviewer effects Highly cost-effective Hi-tech experimentation possible (again cost effective) Panels – study dynamics in multi-wave panels easily