APPLICATION: THE SUPREME COURT 1
Extra Credit Google search: URL should be: Choose Initial login for all programs Session name: kld1 You will earn extra credit points on HW4 equivalent to the dollar amounts in round 2 (round 1 is a practice round with a potentially different randomly drawn person in the room).
Three Papers Bonneau et al. 2007, AJPS Examines a simple agenda setter model and compares it to a median voter model. Carruba et al. 2012, AJPS Carefully models the court s decision making process and concludes that opinions are likely to reflect the views of the median justice in the majority coalition. Clark and Lauderdale. 2010. AJPS Create Bayesian estimates for the Supreme court which include the location of the opinions and provides support for the opinion being closest to the views of the median justice in the majority coalition.
Recall: One Chamber, No Override Assume: one chamber, fixed agenda setter. Median voter (M) proposes a bill b. President (P) signs bill or vetoes it. If the president signs, the policy outcome is x = b. If the president vetoes, the policy outcome is x = q. Bonneau et al. 2007 (same model) The author of the opinion. Median justice. 4
Bonneau et al. Agenda Control and The Median Justice A. Model 1. M is the author of the opinion and P is the median justice of the Supreme Court. a. The justice writing the opinion will write an opinion that is closest to his/her ideal point that is within the majority winset of q. * W(q) q median author (M) justice (P) propose * 5
B. Data Bonneau et al. Agenda Control and The Median Justice 1. Ideal points estimated two ways: a. As percentage of times voting in the liberal direction on Spaeth dataset in the year prior to the decision. Have we seen this type of b. Martin-Quinn Bayesian scores. spatial estimate before? 2. The status quo is the midpoint between two justices: one of whom votes for cert and one of whom votes to deny. a. Any problems with that? 6
B. Data Bonneau et al. Agenda Control and The Median Justice 1. Dep Var: = 1 if justice voted with majority; 0 otherwise. 2. Ind Var: Author acceptability: justice coded 1 if the rational opinion of the author (*) is closer to their ideal point than q is to their ideal point; 0 otherwise. 3. Ind Var: Median acceptability: justice coded 1 if the justice is closer to the median than they are to q. Ex: If J 1, J 2, or J 3 are authors, the rational proposal is *. J 1 to J 5 are coded 1 on author acceptability because they are closer to * than to q. J 1 to J 6 are coded 1 on median acceptability because they closer to J 5 than to q. 7
Bonneau et al. Agenda Control and The Median Justice Here we are simply looking at which model fits the data better. The BIC strongly favor the agenda control model (i.e., author proposer model) over simple MVT. Conclude: perhaps the Supreme court is not a simple application of the MVT. 8
Carrubba et al. Who Controls the Content of Supreme Court? A. Vocabulary 1. Disposition: whether to affirm or reverse the lower court decision. i.e., whether plaintiff or defendant wins. 2. Legal rule: the legal reasoning or precedent used to decide the case. What most articles focus on. 3. Expressive benefits: benefits from something other than the outcome. In this case, benefits a judge get from putting his/her name on an opinion they believe in. 9
Carrubba et al. Who Controls the Content of Supreme Court? B. Model: implicitly two steps 1. Unofficial vote on the disposition of the case. Affirm Disposition Cut Point Reverse x 1 x 2 x 3 x 4 x 5 2. Vote among proto-majority on legal rule. 10
Carrubba et al. Who Controls the Content of Supreme Court? C. Assumptions 1. Utility Function: where x i is a justice s ideal point, p is the location of the legal rule, K > 0 is a constant (making all valued sources of utility positive), α > 0 is a justice s value for the disposition of the case, e i are expressive benefits for the legal rule a justice signs, c is the cost of writing a separate opinion. 11
Carrubba et al. Who Controls the Content of Supreme Court? D. Bargaining in the Proto-Majority Affirm Disposition Cut Point Reverse x 1 x 2 x 3 x 4 x 5 Possible coalitions in the proto-majority: {1,2,3}; {2,3,4}; and {1,2,3,4}. Suppose coalition {1,2,3} forms and decides to write at 2. Justice 4 might value writing a legal rule at x 4 sufficiently to write a concurring opinion instead of joining the majority. Equilibrium: If majority opinion, rule adopted is at the ideal point of the median member of the coalition signing or the rule closest to median member signing. If plurality opinion, rule adopted at the ideal point of the opinion writer. 12
Carrubba et al. Who Controls the Content of Supreme Court? Dep Var: decision by a justice in the majority to file a special or general concurrence (note: justices dissenting on case j are excluded). Ind Vars: Based on Martin-Quinn scores. 13
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space Paper estimates ideal points and the location of opinions (i.e., legal rules) simultaneously using Bayesian methods. Locations of the opinions was not available in the previous studies. Assumptions Each opinion has a fixed location along a single dimension (like a bill). Ideal points are at fixed locations. The probability of a positive citation of another opinion decreases as the distance between any two opinions increases. Dissenting opinions are written at the author s ideal points. Used to tether the opinion and ideal point spaces. 14
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space Ideal Points Estimated using standard IRT: where α i is the difficulty parameter (i.e. cutpoint), and β i is the discrimination parameter. Court Opinions Estimated using: where x k is the location opinion k. where 1 indicates case k cited case k positively, 0 indicates case k cited case k negatively. NA indicates case k did not reference case k. NA s will be inferred, not dropped. 15
WINBUGS CODE: With Covariates model{ # logit for ideal points for (i in 1:55){ for(j in 1:398){ y[i,j] ~ dbern(pi[i,j]) logit(pi[i,j]) <- x[i]*beta[j] - alpha[j] } } ## priors for roll call parameters above for(j in 1:398){ beta[j] ~ dnorm(0,.1) alpha[j] ~ dnorm(0,.1) } ## probit for court opinions for(i in 1:398){ for(j in 1:398){ y[i,j] ~ dbern(pi[i,j]) probit(pi[i,j]) <- L(xk[i,j]- xk[i,j])^2 } } Loosely: IF THIS WERE THE WINBUGS CODE, You would have your ideal point logit up there. And your opinion probit down here. BUT THIS CODE IS WRONG! You might see if they have an on-line appendix. ## more priors (including those that connect dissenters to opinions)
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space This figure divides the location of the Clark- Lauderdale opinions by whether Spaeth coded the opinion liberal or conservative. Decision of case liberal according (Spaeth) We can see the Clark- Lauderdale scores pass a face validity test. 17
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space Y- location of the majority opinion. X- varies by column. C1: location of the court median. C2: location of the opinion author. C3: location of the median of the majority coalition. C3 most closely related (i.e. on 45). Hence, it fits best. 18
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space Dep Var location of the majority opinion. Concludes that model based on median of majority coalition fits best, though all of the coefficients are significant in the bivariate regressions, and both the author opinion and coalition median are properly signed and significant in search and seizure cases. 19
Clark and Lauderdale Locating Supreme C. Opinions in Doctrine Space Thoughts? Which model is most convincing? How could you create similar models for your own research? 20