Judicial Auditing MATT SPITZER and ERIC TALLEY 報告人：黃君華 20101020 1
Background • The entire judicial branch of government could be viewed as a delegated decision maker for legislative and executive entities. • There are some problems of delegation. • Causes of institutional conflicts : • 1) imprecision, due to limited information. • 2) Differences of ideologies • Combination of the two problems above 2
Chief goal • Tease out how error costs and agency costs interact with one another within a judicial system, and how they affect the nature of oversight exercised by upper ‐ echelon actors on their lower ‐ level counterparts. • Using a game ‐ theoretic model of judicial review to illustrate how the dual concerns of (1) imprecision and (2) ideological bias affect equilibrium auditing strategies employed by upper ‐ echelon actors. 3
Main Conclusion • It leads to a prediction that strategies turn on which concern tends to predominate. • If imprecision predominates, upper actors will tend to adopt an even ‐ handed (one ‐ sided) policy of intervention. –has few things to do with precise content of the lower courts’ decision. • If the concern over political bias predominates, higher actors will decide whether to review or what intensity to review depending on the conclusion of lower actors. • Finally, a benefic flows from political differences between lower ‐ and upper ‐ echelon courts. 4
Road Map • Section II, game ‐ theoretic model of judicial decision making within a two ‐ tiered hierarchy. • (which might represent various relationships) • Section III provides some applications. 5
II. A Model of Judicial Auditing • An institutional model of hierarchical judicial review— which characterizes the equilibrium strategy of the higher actors in auditing, and its relationship to both ideological differences and measurement error within the lower actors. 6
Assumptions • Rational actor models • Law ‐ creating actors have well ‐ specified preferences regarding outcomes and that they act in a way that is strategically calculated so as to satisfy these preferences. 7
The Players and Regulatory Environment • H : the higher actor; L : the lower actor • H and L interact in a strategic, hierarchical choice mechanism • Outcome : a policy, represented as Υ , and the degree of contractual autonomy accorded to market participants (over the selection of passive restraints for passenger vehicles). • Υ = ‐ ∞ means no autonomy; Υ = ∞ means infinite autonomy • Binary choice faced by both players: the status quo ante the law & a “new” regulation 8
• Y ₀ : status quo, current law; Y ₁ : the “new” law • Y ₀ = ‐ 1; Y ₁ = 1 • the shift from Y ₀ to Y ₁ represents deregulatory change (such as NHTSA rescinded a motor vehicle safety requirement). • L: must review the change and its holding is Y L (overturn or uphold), and Y L = Y ₀ or Y ₁ • H: to or not to review L’s decision; the cost for auditing is between $0 ～ $1 (distributed uniformly) • Final outcome: Y H (if H reviewed Y L ); Y L (if not being reviewed) 9
Judicial Preferences (in this model) • Judicial decision makers’ concerns: • 1. a legal outcome consistent with their own policy preferences • 2. to avoid the administrative costs of reviewing others’ decisions • It follows that L and H attempt to minimize the expected sum of (1) the expected squared distance between the final legal outcome Y F and their respective ideal points; (2) administrative costs bore personally. 10
( Ideal points and the state of the world ) • The “ideal point” in Y ‐ space depends on an existing state of the world . • The state of the world represents any empirical determination relevant to policy choice by judges. • For example, it might embody scientific evidence about the marginal net benefits of increasing the autonomy of participants over the selection of safety restraints. 11
• The relevant characteristics of the state of the world is represented by a one ‐ dimensional random variable X , whose realized value x could be any point on the real line between ‐ ∞ and ∞ . • x = ‐ ∞ denotes extremely small net benefits of autonomy; x = ∞ denotes the opposite. 12
1. Both players begin with identical prior beliefs that X is distributed according to a normal distribution, with mean 0 and variance 1. 1. Figure 1 depicts the classic bell ‐ shaped structure of the standard normal probability density. 13
• Given the state of the world, each player’s ideal point is assumed to be governed by a correspondence between a realized state of the world ( x ) and a legal outcome (Y). • For instance, players who value personal liberty would require a strong evidence to convince them to approve constraints. • Assuming that player H possess an ideology as “centrist”, which is represented as y = x , and L’s ideal point is represented by x + θ . • When θ > 0, L’s is more libertarian that H’s. 14
1 . Y denotes legal outcomes. Υ = ∞ means infinite autonomy 2. x denotes the state of the world. x = ∞ means extremely large net benefits of autonomy. 3. H’s ideal point is represented by y = x 4. L’s ideal point is x + θ , and θ = 1, which means L is more unlikely to constrain individual autonomy. 15
Informational Environment • The information used to know the state of the world for players to hear a case can be understood as “a signal”. • A random variable Z represents the signal received by L • Z :(true state of the world )+ (some additional noise); drawn from a normal distribution with a mean equal to the true state of the world ( x ) and variance equal to 1/ ϒ ≥ 0. The precision of L’s signal is ϒ >0 . • ϒ is the precision parameter; ϒ close to 0 means that the signal is uninformative. 16
• 2 stages of H’s decision: 1) whether to review L’s decision; 2) decide the case. • Stage 1—observe L’s final decision, Y L • Stage 2—observe “Z” (the evidences) directly, and H gets a second signal “V”. • V: normally distributed independent of Z, with mean x and and precision σ ≥ 0. • “ σ ”can be interpreted as H’s marginal accuracy advantage over L. • “ σ ” close to 0 means H’s decision not better than L’s decision. 17
• Figure 3 shows the Interaction between L and H. • X is about the benefit brought by autonomy. • Z is about the factual information L receives. • L’s decision is ‐ 1 or 1. (ante status quo or new regulation) • H gets a signal Z or V, and make a decision to uphold or reverse, which is also equal to ‐ 1 or 1. 18
Equilibrium Behavior Concentrate on: • 1. ideological differences between L and H (embodied by a value of θ that differs from 0) • 2. imprecision of L relative to H (embodied by a value of σ that differs from 0) 19
Three circumstances to observe equilibrium behaviors: Case 1 • Case 1: Auditing for Pure Imprecision ( θ =0, σ =1, ϒ =1/2) • L would support the new regulation Y ₁ when she observed a relatively high signal (Z), and conversely support Y ₀ when observing low signal. • H would favor Y ₁ when the combined signals are sufficiently high. 20
1. L’s Model: Z = 0 is cutting point. 2. L’s decision is based on Z to judge the state of the world (x), since L shares the same ideology as H. 3. H’s Model: employ a similar decision ‐ making heuristic, and tend to audit only if the expected value of the additional information ≥ the costs of auditing ($0.18). 4. The probability that H audits either type of holding is 18%. 4. H makes its holding based on Z and V. 21
• The (unconditional) probability that H reverses L is symmetric and equal to approximately 3.3% • When the dominant reason for auditing is imprecision, H tend to adopt a two ‐ sided auditing procedure. • H’s information advantage ( σ ) is positively related to auditing rate: if σ = ½, auditing rate would fall to less 5% and reversal rate to less than 1%; if σ = 2, the rates would be 36% and 6.7% 22
Three circumstances to observe equilibrium behaviors: Case 2 • Case 2: Auditing for Pure Ideology ( θ≠ 0, σ is small relative to ϒ ) (suppose that θ = 1; σ = 0, ϒ = 1/2 ) • L has a more deregulatory ideology than H. • High signal leads to supporting the new regulation; low signal leads to the opposite. • L needs stronger factual information to uphold a new regulation (Y ₁ ). • H would perceive L’s holding as a biased distortion of the signal. 23
1. L’s deciding Model: L would favor the new regulation (Y ₁ ) only if the evidence from the case strongly support regulation (Z < ‐ 2.9) 2. H’s auditing strategy (one ‐ sided strategy): not to review when yL = Y ₀ = ‐ 1 , but more suspicious when yL = Y ₁ = 1 (with an auditing rate of 65%) 3. H makes its own holding based on the same signal Z but with different intensity. 24
Three circumstances to observe equilibrium behaviors: Case 3 • Auditing for Mixed Motives ( θ≠ 0, σ is large relative to ϒ ) • Auditing strategies depend on the incentive the most dominant. • The behavior is moving between Case 1 and Case 2. 25
Recommend
More recommend
