Extremely Preliminary Version 1.1 October 27, 2001 IT TAKES THREE TO TANGO: COURTS, UNIONS AND THE POLITICS OF LABOR LAW CONSTITUTIONAL INTERPRETATION ARGENTINA 1935-1998 by Matias Iaryczower, Pablo T. Spiller and Mariano Tommasi* * Department of Economics, UCLA; Haas School of Business, University of California, Berkeley; and Department of Economics, Universidad San Andrés and CEDI. The authors can be reached at miaryc@ucla.edu, spiller@haas.berkeley.edu, and tommasi@udesa.edu.ar. This paper was prepared for the Washington University Conference on Constitutional Politics, St. Louis, MO, November 2-3, 2001. 1 It Takes Three to Tango: Court, Unions and the Politics of Labor Law 1. Introduction Strikes and public demonstrations are visible instruments unions use in improving their bargaining position vis-à-vis employers. But the use of these instruments is not limited to the industrial relations area. Unions can, and in most countries do, use them to influence political decision making. The direct influence of unions' actions on government decisions seems obvious.1 What is not so obvious is the influence union action may have on the interplay between the Court and the polity. In this paper we explore such influence. We present a model of interaction among the polity, unions and the Courts, which provide conditions for union actions to increase the independence of the Court from the executive. In particular, when Courts oppose President’s anti-labor policies, union actions provide a leverage the court may use in voting against the President's desired policy. Our model shows that to understand the behavior of unions, the Courts and the President, we must take into account the optimal strategies of all the players. Our model is related to the literature on the influence of public opinion on judicial behavior,2 with one major difference. While public opinion was seen in that literature as essentially passive, here unions behave strategically vis-à-vis the Court and the President. Their fights against the President take into account the potential behavior of the Court, and the Court learns from that behavior. We test our model analyzing, for the period 1935-1998, the joint determination of Argentina's Supreme Court constitutional decisions on labor issues,3 the President's tendency to legislate on labor issues, and the unions' tendency to strike. Our results provide support to the strategic model of labor law constitutional interpretation. Our empirical results show that, ceteris paribus, the introduction of new labor laws is lower the higher the probability of a pro-constitutional judicial interpretation, and the higher the probability of union strikes. At the same time, the higher the probability of a pro- 1 Cite work on public opinion, etc. See, for example, Caldeira and Gibson (1992), others. 3 For a general analysis of the Argentine Court's decision making, see Iaryczower, Spiller, Tommasi (2001) - hereafter IST (2001). 2 2 It Takes Three to Tango: Court, Unions and the Politics of Labor Law constitutional judicial interpretation, the lower the tendency of unions to strike. Finally, the probability of a pro-constitutional judicial interpretation falls with the strength of the unions. Thus, we find that in this triangular relationship, Presidents use sympathetic Courts in their fights against the unions; unions use sympathetic Courts in their fights against the President, and the Court relies on union strength to increase its independence from the President. Our model explores the interrelations among the three players in an environment of imperfect information, where the Court is uninformed about the strengths of the President and the unions. We study three basic cases: Street Conflict (SC) in which the Court and the President both have an anti-labor stance; Institutional Conflict (IC) in which only the Court has an anti-labor stance;4 and Total Conflict (TC) in which the President favors an anti-labor policy, while the Supreme Court does not. As in IST(2001), the Court’s vote against the President decreases with the President's political support in Congress. Nevertheless, there are environments in which an otherwise lenient court stands against the President. Furthermore, there is an “informative” equilibrium, in which unions only fight a relatively weak President, and the Court goes against the President only after observing that the union fights the President. 2. The model The basic structure of the model is as follows: We assume that the President can costlessly introduce a new piece of legislation.5 Specifically, her options at this point are: a pro-labor norm (relative to the status quo), the status quo itself, and an anti-labor norm (again, relative to the status quo).6 The union can react fighting this legislation or acquiescing. If the President wins this challenge, which happens with probability pF, the norm remains alive. It she loses, the outcome is the status quo. Before the fight between The IC case corresponds to the game analized and tested in IST (2001). We are using the term “President” to indicate the governing coaltion, including both the President and the Party’s coalition in Congress. 6 Since the worst outcome for the President in this game is assumed to be the return to the status quo (with potentially positive costs of fighting for the innovation), and the President’s preferences for or against labor are common knowledge, the initial choice will be between either a pro-labor innovation and inaction, or between an anti-labor innovation and inaction. 4 5 3 It Takes Three to Tango: Court, Unions and the Politics of Labor Law the President and the union is resolved, the Court must decide whether it challenges the norm on constitutional grounds or upholds it As in ITS (2001), the President can initiate an impeachment procedure against those Justices who defy her. Would she initiate impeachment proceedings, she will accomplish this with some probability pI. If the impeachment proceedings succeed, the constitutionality of the norm is restored. The preferences of the President, the unions and the Court are represented by: VP(;P)=UP(N();P) – KI () - KF(), VU() = UU(N()) – KU(), and VJ(;J) = UJ(N();J) – KJ(), where is the strategy profile, N {NPL, SQ, NAL} is the norm that prevails at the end of the game, Ki() represent the fixed costs of particular strategies (KI (), KF() are the President's fixed costs of attempting to impeach a Justice and of fighting the union, KU() is the union's fixed cost of fighting the President, and KJ() represents the court's disutility of being effectively impeached by the President). The President's pro- or antilabor stance is represented by P {PL,AL}. Obviously U=PL, while J {PL,AL} represents the Court's pro- or anti-labor stance. As a matter of convention, we will refer to three basic cases depending on the players' preferences: J = P U as Street Conflict (SC); J P = U as Institutional Conflict (IC), and P J = U as Total Conflict (TC). Our interest lies on how unions affect the Court's ability to make decisions free of the threat of presidential retaliation. Thus, the most relevant case is TC. Nevertheless, we solve SC and IC as benchmarks.7 7 The less interesting, No Conflict, case is when J = P = U, which deserves little attention. 4 It Takes Three to Tango: Court, Unions and the Politics of Labor Law We specify now the game's information structure. Information is assumed to come from two independent sources. A publicly observed signal, M, represents the nominal majority of the President in Congress. The probability of impeachment, pI(M), is non decreasing in M. pI(M) is realized at the beginning of the game, and is hence common knowledge. The probability that the President can successfully resist a challenge by the unions, pF, is unobserved by the Court. We assume pF depends on the realization of a random variable {L, H}, which is observed by both the President and the unions, but is unobservable to the Courts.8 The values that takes represents whether the president is "tough" or "weak," thus, pFH = pF (H) > pFL = pF (L).9 The extensive form of the game with this informational structure is represented in Figure 1 below. The game begins with Nature choosing M and . The President, whether Tough ( = H) or Weak ( = L), then chooses to introduce or not the norm. If she does not, the game ends, and the Status Quo prevail. If she does, the unions (for each value of ), must decide whether to fight or not.10 The Court’s strategy must then specify whether to challenge the norm or not in both the event that the unions fight or do not. While the Court does not observe , it has prior beliefs 0. Following the Court’s decision, the President has to decide whether to attempt to impeach the Justices or not. Outcomes follow. Note that since we are assuming that a defeat of the President whether by the Courts or by the streets can reverse her decision, in the event that the union fights there are four possible events, only one of which – successful impeachment and successful fight - will see the President triumphant. 8 This informational asymmetry can have multiple sources, one of which can be the degree of party discipline in both political parties and the unions. 9 A more general formulation would have assumed that p = p ( ,). I I M 10 As a matter of convention, we will refer to a union being “High” when =H, while a union is "Low" when =L. 5 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Figure 1 N ThetaH ThetaL <Mu0> <(1-Mu0)> PH Innov PL No Innov No Innov Innov UH UL No Fight No Fight <Mu1> Fight JNF <(1-Mu1)> Const Const Unc PH PL Imp Imp NI NI N Can Impeach J Fight Unc N Can't Impeach J Can't Impeach J <Mu2> Can Impeach J <(1-Mu2)> JF U Wins Const Const Unc N N PH PL P Wins P Wins Imp & Fight Imp & Fight NI NI N Presi Wins U Can Impeach J U Wins Unc N Presi Looses U Can't Impeach J Presi Wins U Presi Looses U Can't Impeach J Can Impeach J Presi Wins U Can Impeach J Presi Looses U Can't Impeach J Presi Wins U Presi Looses U Can't Impeach J Can Impeach J Formally, let “I” denote a successful impeachment attempt, “~I” an unsuccessful impeachment attempt, “F” a "street" victory for the President against the unions, and “~F” a defeat of the President at the hands of the unions. Then the set of possible states of nature for the informed players (President and unions) is Inf = {(I,F), (I, ~F),( ~I, F),( ~I, ~F)}, and that of the uninformed player (the Court) is Uninf = {(I,F, H), (I, ~F, H),( ~I, 6 It Takes Three to Tango: Court, Unions and the Politics of Labor Law F, H), ( ~I, ~F, H), (I,F, L), (I, ~F, L),( ~I, F, L), (~I, ~F, L)}. The Court has a probability distribution over Inf, p(;M) = p(/;M).q(). We will denote the Court’s prior over as 0 = q(H) and (1-0)= q(L). After observing the play, but not yet the outcome, by the union and the President, the Court will update its prior. We solve for the sequential equilibria of the game. In a sequential equilibrium the following conditions hold: (a) behavioral strategies satisfy sequential rationality, (b) Bayes’ rules is used when possible to update beliefs – in information sets reached with positive probability in equilibrium -, and (c) for information sets that are not reached with positive probability in equilibrium, beliefs are the limit of the beliefs – computed using Bayes’ rule – in a sequence of assessments [t,t] of totally mixed strategies converging to the equilibrium. Using Bayes’ rule, q(/)=Pr(/).q() / Pr(/).q(). Since we are not allowing correlated strategies, we have that 1 = q(H/Fight by the unions is observed) and 2 = q(H/Fight by the unions is NOT observed) satisfy 1 = 0.DPH.FUH. [0.DPH.FUH+(1-0).DPL.FUL]-1 and 2 = 0.DPH.(1-FUH). [0.DPH.(1-FUH)+(1-0). DPL.(1-FUL)]-1 . Here DPH denotes the probability that the equilibrium strategy of the “Tough” President assigns to “innovate”, DPL denotes the probability that the equilibrium strategy of the “Weak” President assigns to “innovate”, FUH denotes the probability that the equilibrium strategy of the union that observes a “Tough” President assigns to “fight”, and FUL denotes the probability that the equilibrium strategy of the union that observes a “Weak” President assigns to “fight”. We will also let CJF and CJF denote the probabilities that the strategy of the SC attach to voting Constitutional after observing Fight and Not Fight by the unions. 7 It Takes Three to Tango: Court, Unions and the Politics of Labor Law At the expense of tiring the reader with notation,11 we introduce some, we think, useful definitions. Define the union’s “relative” cost of fighting the President as HU KU.[UU(SQ)- UU(NAL)] -1; the ratio of the fixed cost KU and the difference in payoffs of keeping the Status Quo and having the anti-labor norm NAL. Similarly, the President’s “relative” cost of fighting the union is defined as F KF.[UP(NAL)- UP(SQ)] -1. The President's "relative" cost of attempting to impeach the Court when she favors an anti-labor policy but the Court favors a pro-labor policy is I KI.[UP(NAL)- UP(SQ)] -1 while her "relative cost" of impeachment when she favors a pro-labor policy but the Court favors an anti-labor policy is given by I‘ KI.[UP(NPL)- UP(SQ)] -1 Finally, denote SC’s relative cost of being impeached when the President favors an antilabor policy but the Court favors a pro-labor policy as HJ KJ.[UJ(SQ)- UJ(NAL)] -1, and HJ‘ KJ.[UJ(SQ)- UJ(NPL)] -1 will denote SC’s relative cost of being impeached when the President favors a pro-labor policy but the Court favors an anti-labor policy. We now start solving the benchmark cases. The No Conflict case is straightforward, and its results are given in the next proposition.12 Proposition 1. In the No Conflict case, the President enacts a pro-labor legislation, which is not challenged by the Court, or fought by the unions; there is no impeachment. 11 12 We promise better notation for the next draft. Proofs will be provided in the appendix in the next draft. 8 It Takes Three to Tango: Court, Unions and the Politics of Labor Law The following proposition completely characterizes the set of equilibria in the Street Conflict (SC) case: Proposition 2. In the Street Conflict case,13 the Supreme Court does not challenge the President, who never impeaches. Fight by the unions is “non-increasing” in HU, and innovations by the President are non-increasing in F. Specifically, (a) if HU < 1 - pFH,14 the union fights both a weak and a tough President. In this case, for F < pFL, the President innovates independently of ; only the tough President innovates if pFL < F < pFH, and the President does not attempt to legislate independently of her type if pFH < F; (b) if 1 - pFH < HU < 1 - pFL,15 the union only fights a weak President. Here a tough President always innovates, and a weak President only does so if F < pFL. Finally, (c) if 1 - pFL < HU, the union never fights. In this case, the President legislates independently of the realization of . Corollary: The equilibrium is completely independent of the President’s majority in Congress M. We now introduce more definitions: Definition 1: We say that there is a Unified Government (UG) whenever I < pI. Hence, given a majority level M, we say that there is a UG if M > M*, where M* satisfies I = pI (M*). A government that is not Unified is Divided. The next proposition completely characterizes the set of equilibria in the Institutional Conflict (IC) case. 13 Recall that "Street Conflict" takes place when the preferences of the president and the court are aligned. 14 That is, the probability of a Tough President losing the "street" fight to the unions is high. 15 In this case, the probability of a Tough President losing the "street" fight is relatively low, but a Weak President losing the fight is relatively high. 9 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Proposition 3. In the Institutional Conflict (IC) case, (a) A constitutional determination by the Supreme Court, CJ, is, on the equilibrium path, non-decreasing in M. (b) Unless there is a Unified Government, the President never impeaches. Hence, SC always challenges an eventual (pro-labor) norm, which is in fact never enacted (indifferent). (c) Under a Unified Government, the SC only acquiesces if pI.(1+HJ’) >1. The President always innovates in this latter case, and impeaches when SC challenges only if HJ‘< pI. Corollary: The equilibrium play is independent from HU, F and pF. Following our inability to simplify matters, we now introduce even more terminology: Definition 2. We say that there is a Combative Unified Government (CUG) whenever (I+F)/pF< pI . Hence, given a majority level M, we say that there is a CUG if M > M*, where M* satisfies (I +F)/ pF = pI (M*). We now start analyzing the Total Conflict case. We first look at the simplest case in which there is no asymmetric information. The following proposition describes the most salient facts about the equilibrium in this case. Proposition 4. Equilibrium behavior in the Total Conflict with Symmetric Information (TCSI) case displays the following properties: (a) The President only innovates in CUGs. (b) All else equal, Supreme Court's constitutional challenges are non-increasing M and the unions' fighting the President is non-increasing in HU. (c) Free Riding by Courts. Let M* denote the value M such that M > M* CJF = 1, and M** denote the value M such that M > M** CJNF =1. Then M*<M**. (d) Free Riding by unions. Let HU* denote the value of HU such that HU < HU*,CJF = CJNF =0 FU =1, and HU** denote the value of HU such that HU < HU**,CJF = CJNF =1 FU =1. Then HU* < HU**. When CJF = 0 and CJNF =1, FUH =FUL =0. 10 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Point (c) indicates that in TC with symmetric information, the smallest President’s majority in Congress required to get the Court to acquiesce is actually higher when the union doesn’t fight that when it does. In this range the Court is leaving the burden of the fight to the union, avoiding any risks. Similarly, point (d) indicates that in the TC with symmetric information, the highest relative cost of fighting for the union that gets the union to fight is actually lower when the union anticipates that the Court will challenge the President independently of its action, than when it anticipates that it will acquiesce independently of its action. In this range, the union is the one avoiding taking any risks, leaving the burden of the fight to the Court. Moreover, by (c) there is an interval [M*,M**] of the President’s party majority in Congress in which the Court would not challenge if the union strikes but challenges if the union does not fight. The union’s response in this interval is to avoid fighting, leaving the risks to the Court. Now we are prepared to tackle the Total Conflict with Asymmetric Information (TCAI) case. Our first proposition for this case states that a Unified Government is not a sufficient condition for the President to go against the unions. Proposition 5. In the Total Conflict with Asymmetric Information, if the Government is not a CUG, then the President does not innovate. Given the extent of free riding in TCSI, the next proposition explores free riding in TCAI. Proposition 6. In the Total Conflict with Asymmetric Information (a) If pI < pF or pI.(1+HJ’)>1 there is no equilibrium in which CJF =1 and CJNF =0 . (b) When neither of these conditions hold, for an equilibrium with CJF =1 and CJNF =0 to exist it is necessary that HU be sufficiently small, and that at least the Tough President Government is Combative Unified. 11 It Takes Three to Tango: Court, Unions and the Politics of Labor Law This proposition shows the power of informational asymmetries. While under symmetric information both the union and the court could free ride on each other (the court upholds the President when there is a high probability of the union winning the street fight, and the union not fighting when there is a high probability of the Court going against the President), such results do not hold under informational asymmetries. Indeed, Proposition 6 says that under reasonable conditions, there are no equilibria in which the Court decides to uphold the President if unions fight, while it decides to go against the President if unions do not fight. Since pF refers to the probability that a President can maintain a given piece of legislation against a fight by the union, while pI refers to the probability that she can successfully impeach a Supreme Court Justice, the requirement that pI < pF does not seem too stringent. We will henceforth maintain this as an assumption. Nevertheless, it should be noted that if this condition does not hold, we might have an equilibrium in which Court free rides on the union when at least the Tough President’s Government is Combative Unified. In this case, opposing the Government is too risky for Justices, and the opportunity cost of not doing so is reduced by the relatively high expectation that unions could win their fight against the President “in the Streets”. The relatively small HU is needed to assure that unions will go against the President even if they know they will be fighting alone. The next Proposition shows that – as in the previous cases - Court independence is reduced when they face a more unified Government if they care enough about being impeached, and this is sufficiently “cheap” – or the issue considered sufficiently important – for the President. Furthermore, if even the weak President’s Government is Combative Unified, we find that the Court’s behavior is independent of union’s behavior. Before presenting these results we need one more definition: Definition 3. We say there is impeachment uncertainty when the weak President’s Government is Non Combative Unified, and the tough President’s Government is Combative Unified. 12 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Proposition 7. In the Total Conflict with Asymmetric Information Case: (a) On the equilibrium path, CJ is non-decreasing in M. (b) If HJ’<(1-pI)/pI or no President has a Combative Unified Government, then CJ=0 on the equilibrium path. (c) If HJ’>(1-pI)/pI, and there is impeachment uncertainty, then CJNF =1. (d) If both the weak and tough Presidents have Combative Unified Governments, then CJNF =CJF =1. Proposition 7 shows that in the Total Conflict case, the Court will follow very strong Presidents, independently of what the unions do. However, if there is uncertainty about impeachment and the cost of impeachment is high, then if unions do not fight, the Court will uphold the President. On the other hand, if the President's probability of being very strong is low, then the Court will vote against the President, independently of unions' actions. We next turn to the question of whether there exist an equilibrium in which even if neither the Court nor the union would challenge the President in the IC or SC cases respectively, they would “cooperate” to jointly defy her in the TCAI case. The next proposition establishes the conditions for what we call a “Coordination Inducing Separating Equilibrium” to exist. Proposition 8. In the Total Conflict with Asymmetric Information Case, assume pI.(1+HJ’)>1 and HU>1-pFL (the Court doesn’t challenge the President in the IC case, and the union doesn’t challenge the President in the SC case). (a) If 0 > pFL.[pI.(HJ+pFH)-( pFH-pFL)]-1, and there is impeachment uncertainty, there exists an equilibrium in which FUH<1, FUL=1 and CJNF=1, CJF<1. (b) Furthermore, if HU>1- pI.pFH, this equilibrium has FUH=0, FUL=1, CJNF=1 and CJF=0. Thus, under the conditions of Proposition 8, there is an equilibrium in which the unions fight a Weak President with a higher probability than a Tough President, and the Court 13 It Takes Three to Tango: Court, Unions and the Politics of Labor Law challenges the President with a higher probability when it sees a fight by the union that when it does not. Moreover, if the additional condition that union's relative fight cost are high (HU>1- pI.pFH) holds, this equilibrium is in pure strategies. The union only fights a Weak President providing the Court with a perfect signal of Presidential strength. Thus, the Court only challenges the President when it sees a fight, that is, the Court only challenges a Weak President. Note that the existence of the coordinating inducing separating equilibrium depends upon the fact that there is impeachment uncertainty, so a higher difference between pFH and pFL, and a high relative cost of signaling for the union would contribute to this result, while rather extreme (either high or low) values of M would not. The next result is closely related to the previous proposition. Corollary. In the Total Conflict with Asymmetric Information Case, if pI.(1+HJ’)>1, 0 > pFL.[pI.(HJ+pFH)-( pFH-pFL)]-1,and there is impeachment uncertainty, then CJNF=1, and CJF is decreasing FUH. Finally, we note that in TCAI: Corollary. The union’s equilibrium strategy is dependent upon the Court's equilibrium strategy CJ. To summarize, our model provides conditions under which there is a subtle coordination of actions among the unions and the Court. Unions fight the President, and this fight provides an impetus for the Court to block the President's legislative proposal. Furthermore, in this case, weaker Presidents may decide not to innovate. Thus, a major empirical implication of our model is that the behavior of unions, the Courts and the President must take into account the optimal strategies of all the players. 14 It Takes Three to Tango: Court, Unions and the Politics of Labor Law 3. Some Preliminary Evidence. A. Data Description A.1. Supreme Court Cases In this paper we use labor and social security cases that reached the Supreme Court on Constitutional grounds and were decided from 1935 to 1998. Argentina's Supreme Court decides several thousands of cases a year.16,17 Besides the fact that many of these cases are the exact repetition of one another, although with different plaintiffs, their political significance is extremely diverse. Thus, the first issue to address is the scope of the sample. As in IST (2001), we limit the pool of cases, considering only the cases published in extenso in La Ley, the main judicial publication in Argentina. Following Molinelli (1999), to distinguish between important and unimportant cases, our data-set includes only those cases which fulfill three conditions: (1) the case involves the constitutionality of government norms,18 (2) the Court actually decided for or against the constitutionality of the challenged norm,19 and (3) the case was published in extenso in "La Ley".20 16 Those of you familiar with IST (2001) may skip the description of the nature of the Supreme Court data. 17 Since 1991, the Court has been handling between 5000 and 8000 annually. See Molinelli (1999). Differing from its U.S. counterpart, the Argentine Supreme Court does not have the ability to issue certiorari decisions, nor does the stare decisis doctrine formally exist. As a consequence, the Argentine Supreme Court sees a very large number of cases per year (Bidart Campos 1982). But the thousands of cases mask the fact that many are repetitive cases. Since until very recently the court did not have the ability to determine a law as unconstitutional per se, but rather had to deal with the unconstitutionality of its application to a particular case (person), the Court has ruled multiple cases but essentially implemented a single decision multiple times. 18 By norms, we mean laws, Presidential decrees, administrative decisions and resolutions. Cases in which the constitutionality of a lower court decision was questioned (arbitrariedad) and cases in which the constitutionality of the interpretation of a norm by a lower court was questioned but not the norm in itself , were excluded. 19 Cases in which the Supreme Court decided not to pronounce over the constitutionality of the challenged norm, alleging formal or technical reasons, were also excluded. This is in fact a very disparate category, including multiple types of issues, like lack of foundation, improper presentation, “political question”, and so on and so forth. See Molinelli (1999). For this condition to substantially bias the sample, it has to be the case that the Court facing a government decisions it dislikes, but one which it cannot oppose because of the fear of reprisals, chooses to decline to review it based on "technical" reasons. To explore this potential bias we divided the sample in democratic and dictatorship periods. We find that the probability of the Court rejecting to consider a case for technical reasons is the same (around 22%) in both 15 It Takes Three to Tango: Court, Unions and the Politics of Labor Law A.2. The Political Environment The theory presented above suggests that Justices’ votes adjust partially to reflect the President's ideal policy whenever she has the political strength to retaliate. Given an institutional structure like that of Argentina, this will, in turn, depend on the President's degree of control over Congress. The two "dangers" faced by Justices in Argentina over the period of analysis, apart from constitutional reform, were Court enlargement, which until the reform of 1994 could be achieved with a simple law, and impeachment, which requires a supermajority in both Chambers. To capture Presidential control over congress, we create a set of categorical variables that allow us to distinguish the various political scenarios. Democratic governments are classified at the time of each Supreme Court‘s decision as “Unified” or “Divided”, generating two variables for democratic periods, UNIFGOV and DIVGOV. By "unified government" we understand the situation in which the presidential party has an absolute (more than 50%) or relative (plurality) majority in both chambers of Congress. Governments that are not “unified” are "divided.” To reflect the difference between the court enlargement potential and the (tougher) impeachment, we distinguished two cases within the unified government case. UNIFGOV-SIMPLE indicates that while the government can be classified as a unified government, the President does not have the majority required to impeach Supreme Court Justices. On the other hand, UNIFGOV-SUPER indicates that the President not only democratic and dictatorship periods. Thus, we do not believe that this sample selection biases our results. 20 While these criteria may lose some relevant information, since we are focusing on the interaction of the Court with Federal political institutions, this risk is relatively small. The loss of information is mainly bounded to appear in cases that consider provincial norms and low level administrative resolutions - instead of laws or Presidential decrees, both instances where the potential for political conflict is reduced. Additionally, there could be some loss of data in cases where the Court decided the constitutionality of the challenged norm, but for political reasons they were considered "less relevant" by La Ley. Such could be the case with highly politically charged cases during military regimes – although the data set includes several highly charged cases, like that of Jacobo Timmerman, a famous jewish journalist and newspaper owner jailed for opposing the military regime. Helmke (1999 - Table 2) finds that on average, cases that are reported at length in La Ley, tend to reverse the government more often than those which are not reported at length. 16 It Takes Three to Tango: Court, Unions and the Politics of Labor Law controls a unified government, but also has the supermajority required to successfully impeach Supreme Court Justices. The complement to these three scenarios (DIVGOV, UNIFGOV-SIMPLE and UNIFGOV-SUPER) is DICTATORSHIP, which takes the value 1 whenever the Presidency is occupied by a dictator and 0 when the President is democratic. A.3. Judicial Preferences21 An important part of the empirical exercise is to account for Justices’ preferences over policies, and through it, to measure the importance of appointing "friendly" Justices. "Measuring" preferences is obviously not an easy task. A first approximation would be to create an absolute index over time reflecting more or less pro-labor positions of Judges and Congress derived from voting behavior.22 Nevertheless, the absence of strong national political parties with fairly stable positions in the policy spectrum, and the scarcity of roll-call data, makes this a very difficult task in Argentina.23 Here, instead, we compute the extent of political alignment between the Justice and the sitting President by examining the appointment process. The basic idea is to look at whether the justice was appointed by the sitting president, a past president with a similar stance in labor legislation, or a past president with a different stance in labor legislation, combining this with the appointing president’s control over the senate. To explain the way we compute our political opposition variable (POLOPOS), assume initially that Congress does not participate in appointing the Justices. That is, the President can appoint whomever she wants. In this case, the President would appoint a justice with preferences identical to her own. During this President's tenure, the justice has a 100% political alignment. Thus, our political opposition variable, POLOPOS, will take a value of 0 for that particular Justice, reflecting that the President and the Justice have the same political tendency. Assume now that a new President is elected, and that the Justice is still at the Court. Since we are assuming that the Justice is a perfect clone of the nominating president, the value of POLOPOS assigned to the Justice will depend on 21 22 The discussion of justices' preferences, follows IST (2001). See Bergara, Richman and Spiller (1999) and Segal (1997). 17 It Takes Three to Tango: Court, Unions and the Politics of Labor Law the comparison of the two presidents’ political tendencies. If the new president has the “same political tendency” of the former president, the value of POLOPOS will still be 0. If the new president has an “opposing political tendency”, POLOPOS will take the value 1.24 Prior to its reform in 1994, the Argentinean Constitution established that Supreme Court candidates must be nominated by the President and approved by the Senate by a simple majority. Since 1994, a two-thirds majority in the Senate is required. To get a more accurate description of the Argentinean appointment process, we modify the POLOPOS variable as follows: Whenever the President has the required majority in the Senate, we assume that the President can appoint her most preferred judicial type.25 However, when the President doesn’t have the required majority in the Senate, the equilibrium nomination will reflect a bargaining between the President and the opposition in the upper house. We assume this bargaining game to take the following form. We give the value of 0 to the position of a President with respect to labor policy. An opposed Senate, then, has a value of 1 in the policy opposition spectrum.26 Whenever a vacancy appears, the President has to produce a nomination. If the Senate does not accept this particular candidate, the position remains vacant. In this case, the position of the median voter of the incomplete Court (call this MVI) becomes the status quo, and the payoff that this situation provides to the players becomes their outside value in the bargaining game. The See Spiller and Tommasi (2000). This method allows us to classify Justices’ and Presidents’ preferences along the complete sample (1935-1998) without having to use a common measure for Presidents "located" far in time and political environments. This would be a daunting task given the absence of strong national political parties with fairly stable positions in the policy spectrum. Instead, we only need to compare presidents who “share” Justices, which given the volatility in the Court, substantially simplifies the task. A similar method is what gives continuity to the “nominate” approach. See Poole and Rosenthal (1991). 25 We assume that loyal legislators will accept the President’s nomination without imposing a cost. 26 Since the President lacks a sufficient majority, the President must bargain with the opposition. As mentioned before, given the scarcity of roll calls, it is almost impossible to quantify the "degree" of political opposition of the opposition. Thus, we give it a value of 1 to its political opposition. 23 24 18 It Takes Three to Tango: Court, Unions and the Politics of Labor Law President would like, as in the previous exercise, to nominate a “clone”, but anticipates that this would not be accepted by an opposing Senate, as it would not accept a Justice of a type located further away from its policy ideal than the MVI. Since the President, in turn, will not nominate a justice of a type that is more distant than the MVI, in this simple game an equilibrium appointment is a person of type identical to the MVI. 27 The median across current justices of the POLOPOS variable computed in this way is called MEDIAN. A.4 Unions In the Total Conflict with Asymmetric Case of the previous section the Court’s vote depends on the fighting decision by the union. To measure the model’s “Fight”, FU, we use the number of strikes per year, STRIKES, for which we gathered strikes data from 1935 from various sources (Figure 2). 28 Figure 2 shows a drastic increase in the number of strikes following the military coup that deposed President Juan D. Perón, and the creation of the Peronist Movement. As we showed in IST (2001), the first administration of President Perón marked a defining moment in the relation of the polity to the judiciary. Figure 2 also shows, what many have already mentioned,29 that it also marked a defining moment in the organization of the labor movement, and in the extent of use of strikes. Our method may be inaccurate when multiple appointments are considered at the same time. In this case appointments away from the MVI are feasible, as long as they are balanced (i.e. one to each side of the MVI). Snyder and Weingast (1999) develop a slightly similar model of appointment for NLRB commissioners. 28 It should be noted that the accuracy and reliability of strike data Argentina is at best questionable. Although the importance of unions in Argentina’s political life generated a vast literature, in few cases authors provide data at all, and where these are provided, they are often contradictory. Since every study focuses on a relatively short period of time, we were unable to find a single source covering the entire period of our sample (1935 – 1998). With these constraints, and showing amazing courage for which we may one day pay dearly, we built the variable STRIKES following the following steps. First, we selected what we thought was the best possible source for each subperiod of our sample. Second, to reduce possible differences in absolute values coming from different methodologies, we fixed the absolute values of what we considered was the best source available, and generated the remaining data using the percent variation in the next best available series. The most comprehensive and reliable source is O’Donnel (19..) which covers the period 1955-1972. For 1935 – 1955 we used the variations in data from Rotondaro (19..), which independently covers the period 1935 – 1968. For the period 1972 – 1998 we used Torre (19..) , Fernandez (19..) and Nueva Mayoria (19..). 27 19 Figure 2 Number of Strikes in Argentina, 1934 - 1998 It Takes Three to Tango: Court, Unions and the Politics of Labor Law 800 700 600 500 400 300 200 100 0 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 Since our model derives a different impact of STRIKES upon Court’s behavior according to the political strength of the President, we also introduce the interactions of STRIKES with the political environment, STKDIC, STKUSUPER, STKUSIMPLE and STKDIV. Assuming that labor legislation maps into a state – including wages and unemployment – which determines workers well being, we expect FU to be a function of wages and unemployment. Assuming concavity of the union’s utility function, a given change in policy would produce a bigger change in utility when wages are low, and unemployment is high. Hence, the relative cost of fighting for the union, HU, would be smaller when wages are low and unemployment is high. According to the model in section 1, we would then expect FU to be decreasing in wages, and increasing in unemployment. We then introduce WAGES, which measures the detrended real wage, and OUTPUT, the detrended GDP. A.5 The President According to our model, the President will only introduce her desired legislation if her majority in Congress is sufficiently high. This creates a natural sample selection problem, as laws do not appear randomly. Correcting for this potential problem is no easy task, as 29 References… 20 It Takes Three to Tango: Court, Unions and the Politics of Labor Law we would need to have information about the laws that the President would have liked to introduce, but did not. We approximate this variable by considering the number of labor laws enacted by each President, QLABLAWS, assuming that there is some connection between the number of laws enacted and the number of policy issues addressed (See Figure 3). Figure 3 also shows that labor issues started to become contentious during the first administration of President Perón and worsened thereafter. Figure 3 National Labor Laws - Argentina, 1900 - 2000 25 20 15 10 5 0 1900 1904 1908 1912 1916 1920 1924 1928 1932 1936 1940 1944 1948 Table 1 provides a summary of the data. 21 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Table 1 Sample Information Variable Obs Mean Std.Dev. Min Max QLABLAWS 314 9.105 6.530 0 25 DICTATOR 314 0.433 0.496 0 1 U.SUPER 314 0.105 0.307 0 1 U.SIMPLE 314 0.322 0.468 0 1 DIVIDED 314 0.140 0.348 0 1 CONST 314 0.637 0.482 0 1 WAGE 314 0.043 0.123 -0.302 0.328 OUTPUT 314 0.040 0.112 -0.207 0.209 POSTPER 314 0.793 0.406 0 1 MEDIAN 314 0.172 0.316 0 1 PREFCOURT 314 0.446 0.453 0 1 STRIKES 314 177.643 152.697 7 755 STKDIC 314 60.446 92.576 0 368 STKUSUPER 314 3.866 17.415 0 206 STKUSIMP 314 73.041 132.564 0 688 STKDIV 314 40.290 133.666 0 755 AGCONST 194 0.572 0.496 0 1 AGFORMAL 194 0.119 0.324 0 1 Conditional on Data for AGCONST and AGFORM QLABLAWS 194 9.577 6.878 0 25 DICTATOR 194 0.521 0.501 0 1 U.SUPER 194 0.124 0.330 0 1 U.SIMPLE 194 0.278 0.449 0 1 DIVIDED 194 0.077 0.268 0 1 CONST 194 0.675 0.469 0 1 WAGE 194 0.037 0.127 -0.216 0.328 OUTPUT 194 0.054 0.112 -0.152 0.209 POSTPER 194 0.732 0.444 0 1 MEDIAN 194 0.147 0.316 0 1 PREFCOURT 194 0.363 0.455 0 1 STRIKES 194 136.026 105.954 7 368 STKDIC 194 72.423 97.426 0 368 STKUSUPER 194 4.861 21.043 0 206 STKUSIMP 194 56.577 106.338 0 331 STKDIV 194 2.165 8.031 0 54 22 It Takes Three to Tango: Court, Unions and the Politics of Labor Law B. Econometrics We specify the econometric model as follows. Consider first the behavior of SC Justices. Assume the net utility the SC can derive from voting “Constitutional” a particular piece of legislation can be written as UJ(cj, -J, x) = ZC.C + C, where ZC is a vector containing characteristics both of the case and the SC. In particular, ZC includes the decision of the President DP and the reaction of the unions FU. Then we can write CJ = 1 if UJ > 0 and CJ = 0 if UJ < 0, so Prob(cj =1 /ZC)= E(cj /ZC)= Prob(C < ZC.C)= F(ZC.C), where F is the distribution of C. Considering in turn the behavior of the unions and the President, we have FU = ZU.U + U, and DP = ZP.P + P, where ZU includes the behavior of Court, CJ, and of the President, DP, while ZP includes CJ and FU. Assuming that the distribution of C is uniform, we have a system of three linear equations in the three endogenous variables QLABLAWS, STRIKES, and CONST, which we estimate using 3SLS. We impose some identification restrictions which arise naturally from each players' incentives. See Table 2 where we present the results of the 3SLS. Table 2 Three Stage Least Squares Regression QLABLAWS DICTATOR U.SUPER U.SIMPLE DIVIDED STRIKES CONST STRIKES WAGE OUTPUT CONST I 30.43 (2.90) 26.97 (3.66) 25.21 (3.56) 22.65 (3.42) -0.02 (0.003) -23.35 (4.63) I -179.84 (98.89) -233.60 (160.52) -647.38 II III IV 24.78 (2.28) 18.66 (3.07) 19.42 (2.83) 14.72 (2.72) -0.01 (0.003) -15.24 (3.51) 12.89 (1.09) 7.16 (1.39) 4.37 (1.45) 1.93 (1.60) 0.01 (0.004) -1.04 (1.14) 12.96 (1.10) 7.29 (1.40) 4.46 (1.45) 2.05 (1.60) 0.01 (0.004) -1.22 (1.15) II III IV 312.41 (47.28) 184.08 (95.44) -37.63 313.09 (46.87) 184.18 (93.98) -39.01 135.51 (62.00) -472.86 (120.18) -191.85 23 It Takes Three to Tango: Court, Unions and the Politics of Labor Law QLABLAWS POSTPER constant CONST DICTATOR U.SUPER U.SIMPLE DIVIDED MEDIAN (142.07) -6.63 (5.56) 667.39 (120.70) (69.69) -2.48 (3.26) (219.31) (21.76) 161.53 (66.71) I II 1.3782 (0.10) 1.1889 (0.08) 1.1994 (0.06) 1.0754 (0.08) -0.0043 (0.04) STKUSUPER STKUSIMP STKDIV QLABLAWS III 1.5308 (0.10) 1.2038 (0.13) 1.2811 (0.09) 0.9983 (0.10) 0.2325 (0.16) 0.5010 (0.15) 0.5520 (0.14) 0.4456 (0.25) -0.0011 (0.0002) -0.0006 (0.0007) -0.0014 (0.0002) -0.0012 (0.0002) -0.0441 (0.01) -(0.05) (0.07) -0.0013 (0.0003) -0.0010 (0.0011) -0.0010 (0.0003) -0.0011 (0.0002) -0.0535 (0.01) 314 314 -(0.23) (0.10) -0.0011 (0.0004) -0.0005 (0.0016) -0.0008 (0.0006) -0.0022 (0.0081) 0.0064 (0.01) 0.6866 (0.05) 0.5494 (0.09) 194 PREFCOURT STKDIC AGCONST AGFORMAL Obs. (17.41) -7.39 (2.07) (167.43) (17.45) 87.95 (23.32) (17.42) -7.27 (2.02) (166.04) (17.35) 88.72 (23.12) IV 0.5344 (0.15) 0.4142 (0.12) 0.4142 (0.11) 0.3823 (0.25) -0.0849 (0.09) -0.0007 (0.0004) 0.0012 (0.0014) 0.0000 (0.0005) -0.0015 (0.0082) -0.0198 (0.01) 0.6758 (0.06) 0.5816 (0.09) 194 The results are consistent with the predictions of the model. The strategies of each of the players depend on the optimal strategies of the other two. Furthermore, the results seem to reject the symmetric information model, as we observe a subtle complementarity between the actions of the Court and the unions, rather than the free riding that would have been predicted by the symmetric information model. The political environment variables in the QLABLAWS equation are all significant and match the predictions of the theory: QLABLAWS increase with the strength of the President as measured by her majority in Congress. The coefficient of STRIKES is negative and statistically 24 It Takes Three to Tango: Court, Unions and the Politics of Labor Law significant, as would be expected. This is also true for the coefficient of CONST. The negative coefficient of CONST, although not directly consistent with our model, can be understood in that our measure of policy innovation (QLABLAWS) does not take into account the fact that in Argentina much policy innovation is done via Presidential decrees rather than legislation. These innovations, if upheld by the Court, would become the new status quo. Thus, if the Court upholds the President, there is less need to introduce new legislation.30 The strike equation is also easy to interpret using our theoretical results. As wages decrease or unemployment increase (output decrease) the relative cost of fighting for the union goes down, producing a higher degree of confrontation by the union in equilibrium. The negative coefficient of CONST in this equation indicates that a challenging Court fosters confrontation by the unions, as would be the case in the Total Conflict with Asymmetric Information case under the assumptions presented in Proposition 9. Finally, as was seen in Figure 2, the post Perón period has a significantly higher level of labor conflict. This can be understood in that a main achievement of the Peronist movement was the organization of labor unions. The constitutional decision making equation addresses directly our main argument. The results are consistent with our theory. The political environment variables are all statistically significant, and are as predicted by the theory: the probability that the Court challenges the President decreases with her majority in Congress.31 The coefficient of MEDIAN – which measures the degree of political opposition of the Court to the President – is negative, as predicted, although not statistically significant. We also introduce in versions II and III a variable, PREFCOURT, which measures the pro-labor stance of the Court. It is defined as MEDIAN when the President is against labor, and 1MEDIAN when the president is pro labor. This variable is also not highly significant. Our model suggests that the impact of strikes differs with the power of the President. 30 We plan to revise the model to add this extra dimension in future work. Also, note that our database of judicial decisions includes both decrees and laws. 31 The point estimate of UNIFIEDSIMPLE and UNIFIEDSUPER are higher than that of DIVIDED, and the difference is statistically significant. 25 It Takes Three to Tango: Court, Unions and the Politics of Labor Law Thus, we introduce various interactions of strikes and presidential power in this equation. The coefficients of the interactions of STRIKES with the political environment variables are negative and statistically significant, with the exception of STKUSUPER. These results are consistent with the theory, since we only expect union’s behavior to influence Court’s vote for intermediate majorities (although see the significant coefficient for the interaction with DICTATOR). Finally, QLABLAWS’ coefficient is negative, indicating that, all else held constant, the more challenging the President is, the higher the probability of the Courts going against the President. In versions III and IV we introduce the Attorney General's determination on the constitutionality of the case. As in IST(2001) we find that the AG's determination is extremely important in the Court's final determination. Although results in versions I and II are not directly comparable with versions III and IV because of the loss of data, we find that the political environment variables remain highly significant. The strike interactions, though, lose their power. 4. Final Comments Although preliminary, our evidence suggests that in Argentina, contrary to popular beliefs, it takes three to tango. To understand the Court's constitutional interpretation of labor law it is not enough to know its preferences and those of the polity.32 The Court takes direct hints from unions' actions. Unions, and the President, knowing this, behave strategically. When the Court is less prone to support the President, the unions' incentives to take to the streets are higher. The Court takes unions' actions as a signal of Presidential weakness, and increases its defiance of the President. Although supported by the data, we do not claim that our model is directly applicable to all jurisdictions. Indeed, a key feature of the model is the underlying willingness to resort to violence by both the unions and the President to achieve their aims. Unions strike for political, not just for industrial relations purposes. Presidents threat to impeach justices for political, not just for judicial misconduct. Countries which for one or another reason have been able to develop better working institutions, will also show more civilized behavior, and the types 32 See, for example, Spiller and Gely (1992) for an approach to labor statutory interpretation by the US Supreme Court. 26 It Takes Three to Tango: Court, Unions and the Politics of Labor Law of signals which we explore in this paper will be transmitted via more subtle ways. But, it will still take three to tango. 27 It Takes Three to Tango: Court, Unions and the Politics of Labor Law References TB provided 28