Latent Growth Models and Their Application to Political Behaviour Research
description
Transcript of Latent Growth Models and Their Application to Political Behaviour Research
Latent Growth Models and Their Application to Political Behaviour Research
Latent Growth Models and Their Application to Political Behaviour ResearchThomas J. ScottoUniversity of EssexHarold D. ClarkeUT-Dallas and University of EssexIn December 2007 and For Primary Season, Election Appeared to be Foreign Policy Based, but this Changes2008 American Election Was Dynamic
Economic Voting Literature Long and SolidBad Economic News Usually = Bad News for IncumbentsShortcoming: Aggregate Time Series Data or Cross-Sectional DataAssumed that an Individuals Increasing Negativity Leads to Increasing Anti-Incumbent VoteWe can now test that assumption using 2008 America as Case StudyNot trivial: Note Disagreements between Macro-Micro literature on Party IdentificationWhy should we care?Cooperative Campaign Analysis Project6 Wave Internet Based Survey of Registered Voters with over 7,000 RespondentsWe use December 2007, January 2008, March 2008, September 2008, and October 2008 pre-election wavesKey Research Questions:Did Rising Concerns with the Economy Lead to Rising Support (Favour) for ObamaHow did the impact of the economy compete with other issues and attributes?How to examine these thingsLatent Curve ModellingProminent in Education and Sociology, Rare in Political ScienceStructural Equation Modelling Analogue to Random Intercept and Slope ModellingWhy Better than Multilevel Modelling?Five Time Points (Small N at Level 1)Fit Statistics and SEM
Methods and Models?Lets see some equations!
Level 1:Level 2:What this looks like in a FigureObama Dec.Obama. Jan.Obama Mar.Obama Sep.Obama Oct.Econ. Ret. DecEcon. Ret. Jan
Econ. Ret. Mar
Econ. Ret. Sept
Econ. Ret. Oct
ObamaObamaecon.retecon.retNice thing about 5 Waves: Relax Linear Growth AssumptionRaceIdeologyPartisanshipReligiositySocial Demographics?How to incorporate?Before we move onwhat do latent intercepts and slopes mean?But WaitThats Just an Unconditional Growth ModelConditional Growth Model FigureParty ID DecObama Dec.Party ID JanObama. Jan.Party ID MarObama Mar.Party IDSeptObama Sep.Party ID OctObama Oct.Party ID DecEcon. Ret. DecEcon. Ret. Jan
Party ID Jan
Party ID Mar
Econ. Ret. Mar
Econ. Ret. Sept
Party IDSeptParty ID Oct
Econ. Ret. Oct
ObamaObamaecon.retecon.ret.New Racism, Religiosity and Ideology FactorsExogenous Socio-Demographic CovariatesHow does this change the Equations?
Level 1 Equation:Level 2 Equation:10MPLUSWLSMV ESTIMATORRather Quick to ConvergeInterpretation of Output and Things to Look forMeans of the latent slopesMeans and Variances of the intercepts and slopesRegression CoefficientsLambdasFit StatisticsHow do we practically estimate an LGM?What we Get?ParameterObama FavorEconomic Retrospectionsmu-alpha00mu-beta0.18(.10)+0.41(.07)***Lambda(Dec07)00Lambda(Jan08)0.39(0.22)+0.91(0.45)*Lambda(Mar08)2.91(0.20)***2.83(0.34)***Lambda(Sept08)8.40(0.43)***7.72(0.75)***Lambda(Oct08)1010psi-alpha alpha4.52(.36)***3.25(.25)***psi-beta beta0.06(.01)***0.02(.004)***psi-alphabet-0.02(.02)-0.07(.02)***Latent PredictorAlpha-Obamaalpha-Econ. Retalpha-Econ Ret.0.08(0.03)**---Ideology.0.30(0.03)***0.20(0.03)***Religiosity -0.14(0.04)***-0.01(0.03)New-Racism1.60(0.07)***0.23(0.04)***Latent PredictorBeta-Obamabeta-Econ. Retalpha-Econ. Ret.0.05(0.01)*** ---beta-Econ. Ret.0.66(0.10)*** ---Ideology0.02(0.01)***0.00(0.00)Religiosity0.004(0.01)0.01(0.01)+Race0.05(0.01)***-0.013(0.004)**+p