Fowler - two_genes_predict_voter_turnout

Fowler - two_genes_predict_voter_turnout - Two Genes...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Two Genes Predict Voter Turnout James H. Fowler UCSD Christopher T. Dawes UCSD Big Picture: Cooperation The most important unanswered question in evolutionary biology, and more generally in the social sciences, is how cooperative behavior evolved and can be maintained in human or other animal groups and societies. --Robert May 2006, Nature Big Picture: Collective Action [T]he theory of collective action is the central subject of political science…. If political scientists do not have an empirically grounded theory of collective action, then we are handwaving at our central questions. I am afraid that we do a lot of hand-waving! Elinor Ostrom 1997, APSR Cooperation and Participation Evolutionary Game Theory Behavioral economics experiments Fowler 2005, PNAS Fowler 2005, Nature Fowler, Johnson, Smirnov 2005, Nature Dawes, Fowler, Johnson, McElreath, Smirnov 2007, Nature Fowler 2006, JOP Fowler, Kam 2007, JOP Genetic Studies Fowler, Baker, Dawes 2008 (R&R APSR) Fowler, Dawes 2008, JOP Dawes, Fowler 2008 Cesarini, Dawes, Fowler, Johannesson, Lichtenstein, Wallace 2008, PNAS Overview for the Talk Review Twin Study Results Establish hypotheses for two genes that might influence turnout Discuss methods and results of a gene association study for turnout Preview results of a gene association study for partisanship Game Theoretic Models of Turnout Typically assume individuals are self-interested fully optimizing PB - C Equilibrium turnout approaches 0 as population becomes large Palfrey and Rosenthal 1985 Yet millions vote! self-interest or optimizing assumptions are wrong Bendor, Diermeier, Ting 2003; Fowler 2006 And millions abstain inherent variation in cooperation An Empirical Model of Turnout QuickTimeª and a TIFF (LZW) decompressor are needed to see this picture. Plutzer 2004 Empirical Models of Political Participation Theories ignore genes and biology Imply voter turnout is driven purely by environmental factors Recent twin studies suggest genes might be important too Alford, Funk, Hibbing 2005; Hatemi, et al. 2007 Twin Studies Compare the behavior of monozygotic (MZ) twins (identical) dizygotic (DZ) twins (fraternal) share 100% of their genes share 50% of their genes on average Decompose variance A - genetic C - common environment E - unshared environment Criticism of Twin Studies MZ and DZ environments may not be comparable MZ twins may be more strongly affiliated than DZ twins However Studies of twins reared apart validate studies of twins reared together Bouchard 1998 Differences between MZ and DZ twins persist even among twins whose zygosity has been miscategorized by their parents Bouchard and McGue 2003 MZ twins are sometimes in more frequent contact, but evidence suggests that this results from rather than causes greater similarty Posner et al 1996 MZ twins living apart become more similar with age Bouchard and McGue 2003 A Twin Study of Turnout Electronic voter registration records from Los Angeles County 3.8 million voters complete vote histories 8 elections from 2000-2005 (3 primary, 2 statewide, 3 general) Southern California Twin Project (Baker et al. 2007) registry of ~1000 twins living in the Los Angeles area Voting in Los Angeles ρ = 0.71 ρ = 0.50 Los Angeles Voter Turnout Posterior mean A 53% (10%,89%) C 35% (2%,73%) E 12% (3%, 26%) Better model fit when we drop C Independent Sample Replication National Longitudinal Study of Adolescent Health 3 waves of health, social network, and genetic data Oversampled twins (MZ=442, DZ=364) Asked about political, civic behavior in Wave 3 when subjects were in early adulthood Add Health Voter Turnout Posterior mean A 72% (32%,93%) C 20% (1%,57%) E 9% (5%, 15%) Better model fit when we drop C Construct Replication Add Health asked several political participation questions Donate to party or candidate Run for office Contact official Join political organization Attend rally Add Health Political Participation Posterior mean A 60% (11%,91%) C 18% (0%,54%) E 23% (4%, 59%) Better model fit when we drop C Two More Replications Minnesota Twin Family Study 2764 young voters Validated turnout A = 47% (13%, 60%) Twins Days Festival 800 representative adults Self-reported turnout A ~ 40% Some Gene Concepts ~25,000 genes in 46 chromosomes in human DNA Virtually fixed from moment of conception Some genes have different versions (alleles) “Phenotypes” are observable traits and behaviors No endogeneity! May be far downstream from “genotypes” Genes “transcribe” proteins that regulate structure and function Complex social behavior is “polygenic” Mackay 2001, Plomin 2008 There probably isn’t a single “voter gene” Which Genes? Voter turnout linked to prosocial behavior Altruism Edlin, Gelman, Kaplan 2007 Giving in Dictator Game Fowler 2006; Fowler, Kam 2007; Dawes, Fowler 2006 Humanitarianism Jankowski 2002, 2007 For social behavior, Damberg, et al 2001 recommend focusing on genes that influence brain development neurotransmitter synthesis and reception hormone regulation transcription factors Serotonin System in Humans Affects variety of complex social traits Balciuniene et al 2001 Frequently targeted for antidepressants and illegal recreational drugs Craig 2007; Livingston 1996 Serotonin System in Animals In rhesus macaque monkeys impaired serotonin metabolisms cause impulsive and aggressive behavior in response to social stressors Kraemer et al 1989 genetic deficiencies in serotonin system reduce their ability to respond to social stress Suomi et al 2003; Newman et al 2005 In rodents Acute emotional stress hinders serotonin function in several areas of the brain Popova et al 1989; Virkkunen et al 1995 Social Stress, Serotonin, MAOA, and 5HTT Social stress increases serotonin in presynaptic gap 5-hydroxytryptamine transporter (5HTT) encodes integral membrane protein that transports serotonin from synaptic spaces into presynaptic neurons Monoamine oxidase A (MAOA) encodes enzyme that degrades amine neurotransmitters, particularly serotonin QuickTimeª and a TIFF (LZW) decompressor are needed to see this picture. MAOA and 5HTT Genes Regulate transcription, metabolism, and signal transfers between neurons via serotonin system Each stage shown to affect social interactions Craig 2007 Less efficient alleles yield antisocial behaviors Vanukov 1995; Hsu 1996; Lawson 2003; Domsche 2005; Saito 2002; Schmidt 2000; Samochowiec 1999; Contini 2006 Substance abuse, impulsivity, criminality, precocious sexuality, and antisocial personality disorder Rhee & Waldman 2002 Corollary brain activations due to increased sensitivity to negative social experiences Eisenberger 2007 MAOA and 5HTT Genes in Animals MAOA knock-out in rats cause enzymatic activity to come to a complete halt Cases et al 1995 In monkeys, 5HTT is densely concentrated in the output regions of the amygdala, which affects fear recognition O’Rourke et al 2006 MAOA alters the structure of the brain in mice Cases 1996 In mice, social stress increases transcription of both MAOA and 5HTT Filipenko et al 2002 Hypotheses Direct relationship People with more transcriptionally efficient alleles of the MAOA and 5HTT genes are more likely to vote Gene-environment (GxE) interaction MAOA and 5HTT genes interact with social activity to influence turnout Why Focus on Religious Attendance? Consistently strong predictor of turnout Probably because of generalized social belonging (Cassel 1999) Might also be development of civic skills (Verba, Schlozman, Brady 1995) Best measure of general social group activity Other measures political or low incidence Add Health Data ~ 2900 subjects in both genetic sample and Wave III MAOA group 291 and 321 base-pair alleles into a "low" transcription type and 336, 351, and 351 base-pair alleles into a "high" transcription type (Haberstick 2005) Frequency: 41% "low" and 59% "high" 5HTT "long" 528 base-pair allele associated with a much higher basal activity than a "short" 484 base-pair allele (Lesch 1996) Frequency: 43% short and 57% long Add Health Data Voter turnout based on self-report for 2000 election Validated vote better But validated and self-reported turnout both have large and similar genetic components Exclude ineligible (non-citizens, underage) Religious attendance never at least a few times but no more than once a month more than once a month Model Genetic Association Test Mixed-effects logit (Guo & Zhao 2000, Xu & Shete 2006) QuickTimeª and a TIFF (LZW) decompressor are needed to see this picture. Association means allele influences voting behavior “linkage disequilibrium” with allele at another locus false positive due to population stratification (Zkij) QuickTimeª and a TIFF (LZW) decompressor are needed to see this picture. Results MAOA “high” increases turnout ~5% 5HTT “long” increases turnout among attendees ~10% QuickTimeª and a TIFF (LZW) decompressor are needed to see this picture. Other Results Gene-turnout relationship not mediated by partisanship income cognitive ability education Main effect of religious attendance disappears when GxE and controls are in the model Reappears when 5HTT removed (p<0.0001) Analogous to Caspi et al. (2002) Reinterpreting Turnout Empirics Parental turnout strong predictor for turnout in young adults Plutzer 2004 Previously interpreted as social influence More likely genetic, since shared environment component is small Could be due in part to MAOA and 5HTT Reinterpreting Turnout Empirics Turnout is habitual -- people typically either always vote or always abstain Fowler 2006b; Gerber, Green, and Shachar 2003; Green and Shachar 2000; Miller and Shanks 1996; Plutzer 2004; Verba and Nie 1972 Previously interpreted as reinforcement learning Large genetic component suggests inherent variability Could be due in part to MAOA and 5HTT Reinterpreting Turnout Theory Consistent with D & E terms PB + D + E > C Riker and Ordeshook 1968; Schuessler 2000 But variation in taste for voting may be heritable Consistent with new models of ethical voters Feddersen and Sandroni 2006 But genetic variation suggests voting behavior may result from evolutionary adaptation instead of rational optimizing behavior Summary Five tests from four different samples of twins suggest significant heritability of political participation “High” MAOA allele increases turnout by about 5% “Long” 5HTT allele increases turnout by about 10% among religious attendees suggests sensitivity to negative social experiences plays a role in effect of religious attendance on turnout Preview: Dopamine, Partisanship and Genes Significant association between the D2 dopamine transporter gene (DRD2) and partisanship voter turnout Partisanship partially mediates the gene’s influence on voter turnout Parting Thoughts Association studies require replication Further work is needed to understand the difference between the MAOA and 5HTT results Study of schizophrenia yielded several dead-ends Nature and Science now require replication for publication Could have different impacts on the brain Could result from multipurpose nature of MAOA Genes are the institutions of the human body Acknowledgements Chris Dawes Laura Baker and the Southern California Twin Project staff Add Health team Matt McGue and the Minnesota Center for Twin and Family Research Twins Days Festival National Science Foundation grant SES-0719404 University of California Institute of Government Affairs Heritability of Cooperation in the Trust Game, PNAS 2008 Trust Game (Berg, et al. 1995) P1 gets $10, chooses how much to send to P2 Each dollar sent is tripled P2 chooses how much to send back Yields measures of trust and trustworthiness Used to study cooperation fMRI (de Quervain 2004) oxytocin experiments (Zak 2007) Bayesian Markov Chain Monte Carlo (MCMC) Evaluate high-dimension integrals using random draws Also allows use of singleton data Performs better than SEM for recovering parameter estimates in simulated data (van den Berg, et al 2006) Bayesian Inference in Genetic Models Recent studies have successfully applied Bayesian methods to binary data (van den Berg, et al. 2006) survival analysis (Do et al. 2000) GxE interaction (Eaves and Erkanli 2003) item response theory (Eaves et al. 2005) longitudinal models (Burton et al. 2005) multivariate ordinal models (van den Berg et al. 2006) D12C Zi+i+ τA+ij i=i E j A j The ACE Mixed-Effects Model Pr(Tijk = 1) = Y (b k t ij - a k ) MZ τ ij = Ai + Ci + E ij Subject j is a member of family i and voting in election k Ψ = a logit function α = election-specific threshold in voter turnout β = election-specific factor loading τ = latent propensity to vote A, A1, A2 = genetic effects C = shared environmental effect E = unshared environmental effect Model Specifics Random effects are normally distributed: A ~ N(0, σA2) A1 ~ N(0, σA2/2) A2 ~ N(0, σA2/2) C ~ N(0, σC2) E ~ N(0, σE2) MZ shared genetic effect DZ shared genetic effect DZ unshared genetic effect common environment effect unshared environment Model Specifics 2222 2222 2/AC hAACcCACess1 =+ =+ =+ sss1sss11 + / + / + () () () 22 Vague priors α , β ~ N(0,1000000) 1/σA , 1/σC ~ Pareto(0.01,0.01) (also tried U(0,100), Γ(0.01,0.01) QuickTimeª and a TIFF (Uncompressed) decompressor are needed to see this picture. To identify the model We fix σE2 = 1 and then use estimates of σA2 and σC2 to derive the proportion of variance generated by each factor For AE model fix σC2 = 0 Model Specifics QuickTimeª and a TIFF (Uncompressed) decompressor are needed to see this picture. Convergence We ran two MCMC chains for each model Brooks-Gelman (1998) test shows ‘potential scale reduction factors’ reduced to 1.1 or less by 10,000th draw Severe autocorrelation requires 2 million draws Sample Thin second half and keep 10,000 thinned draws from each of the two chains for analysis ...
View Full Document

This note was uploaded on 04/10/2008 for the course HUM 161 taught by Professor Bader,gear during the Spring '08 term at Vanderbilt.

Ask a homework question - tutors are online