240 likes | 367 Views
Valence Politics Scenarios for Electoral Choice in 2010: Mixed Logit Models of Party Choice. Harold Clarke David Sanders Marianne Stewart Paul Whiteley. Fearless Forecaster! (Only Does One-tailed Tests). Not So Long Ago. But Now! March 15th YouGov Poll.
E N D
Valence Politics Scenarios for Electoral Choice in 2010: Mixed Logit Models of Party Choice Harold Clarke David Sanders Marianne Stewart Paul Whiteley
Modeling Party Choice: What Mixed Logit Models Can Do • Structure of Choice: Multiple Choices and Varying Choice Sets • Theory: Heterogeneity in effects of predictor variables • Possible Consequences of varying choices for election outcomes, e.g., what happens to Conservative vote shares if UKIP/BNP candidates are/are not running
Mixed Logit Model P(j|vi) = exp(Uji)/Σexp(Uji) Utility of party choice j for voter i: Uji = αji + ΒjΧi + ΦjZji + ΘjiWji where: αji = alternative-specific constant (fixed or varying) Βj = vector of fixed coefficients Χi = fixed individual characteristics Φj = vector of fixed coefficients Θj = vector of varying coefficients Zji & Wji = choice-varying attributes of choices
MXL ModelSpecification & Analysis • analyze valence model of party choice -Political Choice in Britain; Performance Politics and the British Voter; Making Political Choices: Canada and the United States • BES CMS Data – Jan 2009 – Jan 2010 • data available from BES website: http://bes2009-10.org • MXL analyzes using NLOGIT 3.0 (Greene 2003)
Valence Politics Model: Key Predictor Variables • Party Best on (Valence) Issues • Party Identification (Dynamic) • Party Leader Images (Key Heuristic) • Demographic Controls – Age, Gender, Income, Social Class • Party Best – Random Parameter (Log Normal Distribution) with Performance Evaluation Covariates
Table 1. MXL Model of Party Choice, Key Parameters Characteristics of Choices B Party Best Handle Most Important Issue 1.30*** Characteristics of Choosers Liberal ConservativeDemocrat BB Brown Affect -0.68*** -0.45*** Cameron Affect 0.87*** 0.05* Clegg Affect -0.15*** 0.48*** Party Identification: Labour 3.06*** 0.32 Conservative -2.53*** -1.92*** Liberal Democrat -0.45* 2.41*** UKIP/BNP 0.50 0.03 Other Parties -1.94*** -0.01 McFadden R2 =.66, % correctly predicted = 74.5, lambda = .59, N = 13345 *** - p < .001; ** - p < .01; * - p < .05
Table 2. MXL Model of Party Choice,Heterogeneity in Effects of Party Best on Most Important Issue δs.e. Standard Deviation in Party Best Coefficient 0.88*** 0.30 Accounting for Heterogeneity Impact of Economic Evaluation Factor on Party Best Coefficient -0.08* 0.04 Impact of Public Service Delivery Factor on Party Best Coefficient -0.06x 0.04 *** - p < .001; ** - p < .01; * - p < .05; x - p < .10
Labour Wins Campaign: Issues and Brown as Popular as Blair in 2005
Party Best, Most Important Issue,2005 Pre-Campaign and Febrary 2010
Conclusions I • MXL Models Confirm Earlier Analyses of Party Choice in 2001 and 2005 • Party Leader Images, Party Best on Most Important Issue, Partisanship & Issue-Party Proximities are the "Great Beasts" • Model heterogeneity in effects of party best on most important issue, econmic evaluations factor as covariate – negative effect – asymmetric effects as often hypothesized
Conclusions II – Scenarios(With a Big Cet Par!) • Pure Leader Affect Scenarios – Cameron and Brown must reach 5.5 to get/thwart majority government. • Feb CMS Data Scenario: Conservatives 42.7%, Labour 26.3% -> Conservative majority! • Labour Wins Issue Campaign: Conservatives 40.3%, Labour 32.4% -> hanging chads! • Labour Wins Issue Campaign & Brown as popular as Blair in 2005: Conservatives 38.3% and Labour 35.5%-> hung parliament!
BTW, How Does Phil Do? • Despite his rep, Phil actually doesn’t forecast really well • And – global warming notwithstanding (apologies to Al Gore!) - forecasting 6 more weeks of winter on Feb 2nd isn’t really hard!
References Clarke, Harold D., Allan Kornberg and Thomas Scotto. 2009. Making Political Choices. Canada and the United States. Toronto: University of Toronto Press. Clarke, Harold D., David Sanders, Marianne C. Stewart and Paul F. Whiteley. 2004. Political Choice in Britain. Oxford: Oxford University Press. _____.2009. Performance Politics and the British Voter. Cambridge: Cambridge University Press. Glasgow, Garrett. 2001. "Mixed Logit Models for Multiparty Elections." Political Analysis. 9:116-36. Greene, William H. 2002. NLOGIT Version 3.0 Reference Guide. Plainview, N.Y.: Econometric Software, Inc. Hensher, David A., John M. Rose and William H. Greene. 2005. Applied Choice Analysis: A Primer. Cambridge: Cambridge University Press. Train, Kenneth E. Discrete Choice Methods with Simulation. Cambridge: Cambridge University Press.
Lognormal Distribution, X > 0, where µ = 0, σ = 1 for the log of X