\documentstyle[12pt,epsf] {article} \parskip 10pt \reversemarginpar \topmargin -.4in \oddsidemargin .25in \textheight 8.7in \textwidth 6in \pagestyle{myheadings} \markboth{October 18, 1997}{October 18, 1997} \begin{document} \baselineskip 24pt \parindent 24pt \titlepage \vspace*{12pt} \begin{center} \begin{large} {\bf {\it Mezzanatto} and the Economics of Self Incrimination} ( {\it Cardozo Law Review}(May 1998) 19: 1541-1584) \\ \end{large} \bigskip October 18, 1997 \\ \bigskip Eric Rasmusen \\ \vspace{ .7in } {\it Abstract} \vspace*{-12pt} \end{center} This paper uses the economic approach to address a recent legal question involving self incrimination: what is the effect of allowing a defendant to waive his right to exclude statements he makes during plea bargaining from evidence at trial if plea bargaining fails? This was the issue in the 1995 {\it Mezzanatto} Supreme Court decision. What is the reason for such waivers, and do they increase or decrease the amount of plea bargaining? I suggest that the waivers have two functions in ``cooperation bargaining'' as opposed to ``penalty bargaining'': (a) increasing the incentive of the defendant to provide the full cooperation he promises in return for leniency, and (b) increasing the reliability of the information the defendant provides. \vspace{ 32pt} \begin{small} \noindent Indiana University School of Business, Rm. 456, 1309 E 10th Street, Bloomington, Indiana, 47405-1701. Secretary: 812-855-9219. Direct: 812-855-3356. Fax: 812-855-3354. Email: Erasmuse@indiana.edu. Web: Php.indiana.edu/$\sim$erasmuse. I would like to thank Richard Friedman for suggesting to me that {\it Mezzanatto} could benefit from game theory, and John Maxwell, John Wiley, and participants in the Indiana Business Economics and Public Policy seminar, the Indiana Law and Econ Lunch Bunch, and the January 1997 AALS meetings for helpful comments. \end{small} %--------------------------------------------------------------- \newpage \noindent 1. The Problem and the Law \noindent 1.1 Introduction Law and economics scholars have long debated the issue of the efficiency of the common law generally, but they have had considerable success in discovering efficiency justifications for many of its rules. On the efficiency properties of two major areas of the law, however, law and economics has remained conspicuously silent: constitutional law and criminal procedure.\footnote{See, however, Section 28.3, the very last section of Richard Posner, ECONOMIC ANALYSIS OF LAW (4th ed. 1992). } These areas of law seem to have little to do with the utilitarian goals of efficiency, and much to do with notions of fairness that are independent of deliberate tradeoffs and human happiness. The law of self incrimination involves both of these last frontiers of the economic approach to law: criminal procedure and constitutional law. These are the areas of law most often in the newspapers, and not only the economist, but the average citizen often finds it hard to understand the reasoning behind their rules and decisions. Like the strict 19th century rules of pleading, they seem a lawyer's playground, insulated from justice and rationality. \noindent 1.2. The Facts of {\it Mezzanatto} The question at issue in this article is: {\it Should a defendant be able to waive his right to exclude self-incriminating statements made during plea negotiations as evidence in his trial? } This question reached the Supreme Court in 1995 in {\it United States v. Mezzanatto}. Narcotics agents arrested Gordon Shuster for manufacturing the illegal drug methamphetamine. He immediately agreed to cooperate and phoned Gary Mezzanatto so the police could arrange to buy drugs. Mezzanatto was then arrested. Two months later, Mezzanatto and his lawyer asked to meet the prosecutor to discuss cooperation. The prosecutor agreed, on condition that Mezzanatto waive his right to exclude statements he might make during plea bargaining from being used to impeach any contradictory testimony he might give at trial. Mezzanatto agreed to the waiver, but the prosecutor terminated the meeting after catching Mezzanatto in a lie. Mezzanatto testified voluntarily at his own trial, and the prosecutor confronted him with his earlier statements as part of the cross-examination and called one of the agents who was present as a witness. Mezzanatto was found guilty and sentenced to 170 months in prison, but appealed on the ground that his waiver was not valid. He won in the Ninth Circuit,\footnote{ United Statesv. Mezzanatto, 998 F.2d 1452 (9th Cir 1993).} but lost in the Supreme Court.\footnote{ United States v. Mezzanatto, 513 U.S. 196 (1995). } Seven justices concurred in the majority opinion written by Justice Thomas. Three of them, Justices Ginsburg, O'Connor, and Breyer, noted in a concurring opinion that the only holding necessary to decide {\it Mezzanatto} was the validity of a waiver which allowed the government to use statements made during plea negotiations to impeach the defendant's testimony. By implication, four other members of the majority would uphold broader waivers which allowed the use of such statements in the case-in-chief. Justices Souter and Stevens would not allow any waivers at all. \noindent 1.3. Rule 410 and Rule 11(3)(6) At issue was the interpretation of a statute, not of the Constitution. The Fifth Amendment only applies to compelled testimony, not to voluntary statements. Even under the Court's {\it Miranda} ruling,\footnote{ { Miranda v. Arizona}, 384 U.S. 436 (1966).} it is perfectly legitimate for the state to use the defendant's statements as evidence if it can be shown that those statements were made voluntarily and with full knowledge that the law did not require them. Congress, however, has passed statutes which exclude particular kinds of voluntary statements from use at trial. {\it Mezzanatto} turned on the meaning of two rules, almost identical to each other, from the Federal Rules of Evidence and Federal Rules of Criminal Procedure.\footnote{ U.S.CS Fed Rules Crim Proc R 11(e)(6) (1996) says \begin{quotation} \vspace*{-6pt}\begin{small} (6) Inadmissibility of pleas, plea discussions, and related statements. Except as otherwise provided in this paragraph, evidence of the following is not, in any civil or criminal proceeding, admissible against the defendant who made the plea or was a participant in the plea discussions: (A) a plea of guilty which was later withdrawn; (B) a plea of nolo contendere; (C) any statement made in the course of any proceedings under this rule regarding either of the foregoing pleas; or (D) any statement made in the course of plea discussions with an attorney for the government which do not result in a plea of guilty or which result in a plea of guilty later withdrawn. However, such a statement is admissible (i) in any proceeding wherein another statement made in the course of the same plea or plea discussions has been introduced and the statement ought in fairness be considered contemporaneously with it, or (ii) in a criminal proceeding for perjury or false statement if the statement was made by the defendant under oath, on the record, and in the presence of counsel. \vspace*{-12pt}\end{small} \end{quotation}} %end of footnote here. Federal Rule of Evidence 410 says \begin{quotation} \vspace*{-12pt}\begin{small} Rule 410. Inadmissibility of Pleas, Plea Discussions, and Related Statements Except as otherwise provided in this rule, evidence of the following is not, in any civil or criminal proceeding, admissible against the defendant who made the plea or was a participant in the plea discussions: (1) a plea of guilty which was later withdrawn; (2) a plea of nolo contendere; (3) any statement made in the course of any proceedings under Rule 11 of the Federal Rules of Criminal Procedure or comparable state procedure regarding either of the foregoing pleas; or (4) any statement made in the course of plea discussions with an attorney for the prosecuting authority which do not result in a plea of guilty or which result in a plea of guilty later withdrawn. However, such a statement is admissible (i) in any proceeding wherein another statement made in the course of the same plea or plea discussions has been introduced and the statement ought in fairness be considered contemporaneously with it, or (ii) in a criminal proceeding for perjury or false statement if the statement was made by the defendant under oath, on the record and in the presence of counsel.\footnote{ 28 U.S.CS Appx Fed Rules of Evid R 410 (1995). } \vspace*{-12pt} \end{small} \end{quotation} The House and the Senate disagreed as to whether Rule 410 should allow use of plea bargaining statements for impeachment purposes, with the Justice Department strongly supporting such use. The effective date of Rule 410 was postphoned until the enactment of Federal Rule of Criminal Procedure 11(e)(6) without the impeachment exception, after which Rule 410 was amended to match it.\footnote{ \label{dahlin} Eric L. Dahlin, Note: Will Plea Bargaining Survive United States v. Mezzanatto? 74 Or. L. Rev. 1365, 1367 (1995). United States v. Mezzanatto, Brief for Respondent, Gary Mezzanatto (1994) in text near note 14. Courts upheld Rule 410 against the strained interpretation that use ``against'' the defendant only meant in the case-in-chief, not impeachment; United States v. Lawson, 683 F.2d 688 (2nd Cir 1982). } Courts have ruled that Rule 410 only excludes use of the defendant's statements, not derivative use.\footnote{ United States v. Cusack, 827 F.2d 696 (11th Cir 1987). } The government may use the defendant's statements to uncover other evidence,\footnote{\label{amar} This interpretation has an interesting resemblance to the rule of The King v. Warickshall, 168 Eng. Rep. 234 (K.B. 1783). In that case, a woman accused of theft made a confession which was ruled inadmissible because it was induced by government promises, but use of her statement to discover physical evidence was upheld. The holding was not based on a statute, but on the supposed unreliability of testimony induced by threats or promises relative to physical evidence. For further discussion, see Akhil Amar and Renee Lettow, Fifth Amendment Principles: The Self-Incrimination Clause, 93 Mich. L. Rev. 857 (1995). } or to focus cross-examination of the defendant.\footnote{ Sometimes the defendant asks the government to waive its right to make derivative use of his statements. Such an agreement was upheld in United States v. Palumbo, 897 F.2d 245 (7th Cir 1990). In United States v. Rowley, 975 F.2d 1357 (8th Cir 1992), the proffer letter said that statements by the defendant during plea negotiations could not be used against him except for perjury or false statement prosecutions or to impeach him. The government used the information to focus the subject areas of its cross examination of him, and the defendant objected. The court, following United States v. Havens, 446 U.S. 620, 626 (1980), ruled that this was within the purview of impeachment.} If the defendant's unsworn statements in plea bargaining and sworn statements at trial contradict each other, he may be liable to other sanctions. Either he lied to the prosecutor, obstructing justice (see 18 U.S.C. 1001 ) or he committed perjury at trial (see 18 U.S.C. 1621). That there exist sanctions for those offenses, however, does not mean the prosecutor is allowed to point out the discrepancies at trial.\footnote{The prosecutor can use them in the same case, but not the trial, to enhance the sentence for obstruction of justice under U.S.S.G. @3C1.1. That, in fact, happened to Mr. Mezzanatto, who received an addition of two points (about 35 months) to his sentence. {\it Mezzanatto} Respondent Brief, {\it supra} note \ref{dahlin}, note 25. @1101(d) (3) of the Federal Rules of Evidence exempts sentencing from the rules of evidence. Therefore, United States v. Ruminer, 786 F.2d 381 (10th Cir 1986) held that the plea-statement rules do not apply in sentencing. As Professor Raeder points out elsewhere in this issue of {\it Cardozo Law Review}, this can be a major loophole in Rule 410, since the prosecutor can bring an easy-to-prove charge and then obtain the maximal sentence for that charge by bringing evidence of other offenses into the sentencing hearing--- but he must prove the initial charge under the constraints of Rule 410, and the maximal sentence is limited by the Sentencing Guidelines. } \noindent 1.4. Waivers of Rule 410 and Rule 11(3)(6) Rule 410 leaves unmentioned whether the defendant can waive his right to avoid use of plea bargain discussions as evidence. Many other kinds of rights are waivable according to case law,\footnote{See, for example, { Shutte v. Thompson}, 82 U.S. 151, 159 (1872) (waiver of procedural rights upheld) and { Cohen v. Cowles Media}, 501 U.S. 663 (1991) (newspaper waiver of a First Amendment right upheld); United States v. Wenger 58 F.3d 280, 281 (7th Cir 1995, Easterbrook, J.) and United States v. Schmidt, 47 F.3d 188 (7th Cir 1995) (waivers of the right to appeal upheld).} but Rule 410 makes no mention of whether it can be waived by the defendant. It became common for prosecutors to ask for waivers in plea bargaining and for defendants to grant them.\footnote{ \label{hughes} A standard form used by the United States Attorney's Office for the Eastern District of New York requires the potential cooperator to waive any claim that his statements ``are inadmissible for cross-examination should [he] testify.'' As reported in Graham Hughes, Agreements for Cooperation in Criminal Cases, 45 Vand. L. Rev. 1, 41 (1992). In another circuit, United States v. Stevens, 935 F.2d 1380, 1396 (3rd 1991) says ``Plea agreements, for example, commonly contain a provision stating that proffer information that is disclosed during the course of plea negotiations is inadmissible as substantive evidence of guilt, but is admissible for purposes of impeachment. The standard plea agreement provides: `No statements made or information provided by [the defendant] will be used by the government directly against her, except for the purpose of cross-examination or impeachment should she be a witness in any criminal trial or proceeding and offer testimony materially different from any statements made or information provided during the proffer. . . .' '' } The issue in {\it Mezzanatto} was whether the waivers were valid. Waivers can be written with varying degrees of breadth, but they can be divided into three types. The narrowest waiver only allows use of the defendant's plea bargaining statements to impeach his personal testimony at trial, the issue in {\it Mezzanatto}.\footnote{Just to make sure, the defendant may also be asked to waive a right that, as just discussed, he probably does not have--to exclude evidence derived from his statements. A standard form used by the United States Attorney's Office for the Eastern District of New York said, ``Any prosecution resulting from a breach of this agreement may be premised upon: (a) any statements made by [the cooperator] to the Office or to other law enforcement agents; (b) any testimony given by [him] before any grand jury or other tribunal, whether before or after the date this agreement is signed . . .; and (c) any leads derived from such statements or testimony... the Office may use information derived directly or indirectly from the meeting for the purpose of obtaining leads to other evidence.'' As reported in Hughes (1992), {\it supra} note \ref{hughes}, at 41, 50. Some of those very words are also cited in United States v. Liranzo et al., 944 F.2d 73, 77 (2nd Cir 1991). } A second, broader waiver would also allow use of the defendant's plea bargaining statements to rebut evidence that the defendant brings up at trial, even if that evidence was not the defendant's own testimony. This was the issue in {\it United States v. Dortch}, where the Seventh Circuit upheld such a waiver.\footnote{The waiver read, ``...no statements or information provided by your client during the `off-the-record' proffer or discussion will be used against your client in any criminal case during the government's case in chief. That is, however, the only limitation on the use the government may make of your client's statements.'' United States v. Dortch, 5 F.3d 1056, 1068 (7th Cir 1993).} A Mr. Suess was charged with cocaine trading. He met with government agents after initialling a ``proffer letter'' waiver, and admitted buying cocaine from Tommie Taylor's partnership. The plea negotiations were fruitless and Suess went to trial. He called Tommie Taylor as a witness, and Taylor said Suess had never bought cocaine from him. This contradicted what Taylor had said during his own plea negotations and he was impeached with that. Next, however, the government introduced evidence of Suess's plea negotiation statements to rebut Taylor's testimony. The Court took the legality of limited waivers of the kind at issue in {\it Mezzanatto} to be obvious, but also upheld waivers for rebuttal.\footnote{ The Seventh Circuit has also upheld a Mezzanatto-style clause, in United States v. Goodapple, 958 F.2d 1402 (7th Cir 1992).} The conflict between the 7th and 9th circuits was one reason the Supreme Court granted cert to {\it Mezzanatto}. The third type of waiver would allow the government to use the defendant's statements for any purpose whatsoever. An example of this third type of waiver is the following excerpt from an agreement under which the true-life organized crime figure portrayed in the movie {\it Goodfellas} agreed to testify against his associates: \begin{quotation} \vspace*{-12pt}\begin{small} In addition, in the event that you do not fully comply with all the other terms of this understanding (immediate full and truthful disclosure, testimony, etc. ), this agreement will be nullified. Should this occur, the Government will be free to prosecute you with regard to any and all violations of the federal criminal law in which you may have participated, and to use against you any and all statements made by you and testimony you have given prior and subsequent to the date of this agreement.\footnote{ Nicholas Pileggi, WISE GUY 283 (1985).} \vspace*{-12pt} \end{small} \end{quotation} The validity of this third kind of waiver is less clear, and has not come before the U.S. Supreme Court, at least in the opinion of the three justices who concurred separately in {\it Mezzanatto}.\footnote{Oddly enough, there is, in a way, a majority for the opinion that the {\it Mezzanatto} holding does apply to waivers for the case-in-chief. Four justices refrained from joining the concurrence that denied such a holding, and two other justices joined a dissent which said that there was no principled difference between impeachment and case-in-chief waivers. I leave this question to the jurisprudes.} In a case such as this one, of course, the practical effect of the waiver may not be great, since if the government wished to punish the defendant for lack of cooperation, expulsion from the witness protection program would make his possible incarceration a positive advantage to him in terms of life expectancy. \noindent 2. Public Policy Arguments on Each Side Scholars have had surprising difficulty in finding justifications for the 5th Amendment's privilege against self-incrimination or the Miranda Rule.\footnote{See Paul Cassell, Miranda's Social Costs: An Empirical Reassessment, 90 Northwestern University L. Rev. 387 (1996); Paul Cassell, {\it Miranda}'s Negligible Effect on Law Enforcement: Some Skeptical Observations, forthcoming, { Harvard Journal of Law and Public Policy}; David Dolinko, Is There a Rationale for the Privilege Against Self-Incrimination?, 33 { U.C.L.A. Law Review} 1063 (1986). Amar \& Lettow (1995), {\it supra} note \ref{amar} at 889. } The debate on the statutory Rule 410, however, has stronger arguments on both sides. I will pass over the strictly legal arguments-- for example, whether by analogy if one type of right can be waived, so can a Rule 410 right- - and limit the discussion here to public policy arguments. These, in fact, link to the legal arguments. For all parties seem to agree that the intent of Congress was to facilitate plea bargaining, and if this is the case, and if silence on the issue of waivers cannot definitely be said to mean either that they are permitted or forbidden, then the issue comes down to whether allowing waivers of Rule 410 really does facilitate plea bargaining.\footnote{The Souter dissent, however, does seem to vacillate as to whether policy considerations are relevant: ``The case raises no issue of policy to be settled by the courts, and if the generally applicable (and generally sound) judicial policy of respecting waivers of rights and privileges should conflict with a reading of the Rules as reasonably construed to accord with the intent of Congress, there is no doubt that congressional intent should prevail.'' United States v. Mezzanatto, 513 U.S. 196, 27 (1995). But ``These explanations show with reasonable clarity that Congress probably made two assumptions when it adopted the Rules: pleas and plea discussions are to be encouraged, and conditions of unrestrained candor are the most effective means of encouragement. ... Whether Congress was right or wrong that unrestrained candor is necessary to promote a reasonable number of plea agreements, Congress assumed that there was such a need and meant to satisfy it by these Rules.'' United States v. Mezzanatto, 513 U.S. 196, 32 (1995). } \noindent 2.1 Arguments for Waivers The law treats plea bargains, like settlements of civil suits, according to the principles of contract law, although the judge need not accept the agreement between prosecutor and defendant.\footnote{ ``Pre-trial agreements, such as cooperation agreements and proffer agreements, are interpreted according to principles of contract law.'' United States v. Liranzo, 944 F.2d 73, 77 (2nd Cir 1991), which provides supporting cites. A general, practitioner-oriented article on plea bargains is Larry D. Thompson \& Phyllis B. Sumner, Structuring Informal Immunity, 8 Crim. Just. 16 (1993). } There are differences-- the ability of either side to back out until the defendant actually pleads guilty breaks with the contract doctrine of offer and acceptance\footnote{See Mabry v. Johnson, 467 U.S. 504 (1984). As Professors Scott and Stuntz say, ``... the bargain is ordinarily not binding until the defendant actually pleads guilty. Thus, if the prosecutor promises to recommend a ten-year sentence in exchange for a plea, and the defendant agrees, either the prosecutor or the defendant may still cancel the deal at any time prior to the defendant's performance.'' Robert Scott and William Stuntz, Plea Bargaining as Contract, 101 Yale LJ 1909 (1992). }-- but the principle that mutual promisors are held to their promises is generally upheld.\footnote{The government's contention that separation of powers and prosecutorial discretion forbade courts from binding it to its plea bargains was roundly rejected in United States v. Pavia, 294 F. Supp. 742 (D.D.C. 1969). The decision was probably to the government's advantage, in general. See United States v. Palumbo, 897 F.2d 245, 246 (7th Cir 1990): ``[t]he system works . . . only if each side keeps its end of the bargain.''} In deciding whether such waivers are in the public interest, one must look at their effect in the immediate case, on other cases in which the defendant might testify, and on the court system as a whole. A major concern in both the Ninth Circuit majority opinion and the dissent was the effect that waivers would have on the amount of plea bargaining.\footnote{ United Statesv. Mezzanatto, 998 F.2d 1452 (9th Cir 1993). }. Plea bargaining reduces the risks and costs of defendants, prosecutors, and courts. Judge Sneed, writing for the majority, feared that allowing prosecutors to refuse to negotiate with defendants who did not sign waivers would have a chilling effect on plea bargaining, resulting in more trials and extra burdens for the courts. Judge Wallace, in dissent, said that there would be no such chilling effect, because prosecutors would not require waivers if they hindered successful plea bargaining. Both sides believed that the effect of waivers on plea bargaining in general was an important determinant of whether waivers should be allowed.\footnote{This issue was present but downplayed in the Supreme Court opinion. In particular, the dissent relied heavily on Congressional intent, and made a point of saying that actual policy effects were irrelevant.} If the plea bargain is a bargain, then the parties are presumed to be competent to make agreements in their own interest. A preliminary bargain in which one side makes a concession to start the plea bargaining itself fits naturally into this analytic framework. As Justice Thomas wrote, \begin{quotation} \vspace*{-6pt}\begin{small} A sounder way to encourage settlement is to permit the interested parties to enter into knowing and voluntary negotiations without any arbitrary limits on their bargaining chips. To use the Ninth Circuit's metaphor, if the prosecutor is interested in `buying' the reliability assurance that accompanies a waiver agreement, then precluding waiver can only stifle the market for plea bargains. A defendant can `maximize` what he has to `sell` only if he is permitted to offer what the prosecutor is most interested in buying.\footnote{ United States v. Mezzanatto, 513 U.S. 196,22 (1995). The market metaphor can be traced back to Frank Easterbrook, Criminal Procedure as a Market System, 12 J. LEGAL STUD. 289 (1983). } \vspace*{-12pt}\end{small} \end{quotation} Making negotiations more flexible ought to increase the number of successful negotiations, to the benefit of both defendants and the government. As Judge Wallace wrote in his 9th Circuit dissent, ``Given the mutual benefits achieved through plea bargaining, should we expect the government continually to require waivers if such requirements significantly reduce the number of plea agreements reached?''\footnote{ United Statesv. Mezzanatto, 998 F.2d 1452,21 (9th Cir 1993).} These arguments apply equally to both the Mezzanatto waiver and to waivers for the case-in-chief. In cases where the waivers are offered and accepted, they facilitate plea bargaining; in cases where they are not used, the possibility of waivers does not impede plea bargaining. \noindent 2.2 Arguments Against Waivers Exactly the same framework is interpreted by those less friendly to the workings of markets to imply that waivers should not be granted. They argue that allowing waivers hurts the market for plea bargains in two ways, by reducing the number of successful plea bargains and by unfairly tilting the table towards the government in those that are successful. Judge Sneed of the 9th Circuit said, \begin{quotation} \vspace*{-6pt}\begin{small} To allow waiver of these rules would be contrary to all that Congress intended to achieve. If these rules were subject to waiver, candid and effective plea bargaining could be severely injured. ... Allowing a waiver of these rules would contravene and thwart the policy - efficient case resolution through plea bargaining - these rules were designed to effectuate.\footnote{ United Statesv. Mezzanatto, 998 F.2d 1452, 8,9 (9th Cir 1993). } \vspace*{-12pt}\end{small} \end{quotation} Similarly, the Souter dissent says, ``These explanations show with reasonable clarity that Congress probably made two assumptions when it adopted the Rules: pleas and plea discussions are to be encouraged, and conditions of unrestrained candor are the most effective means of encouragement.''\footnote{ {\it United States v. Mezzanatto}, 513 U.S. 196, 32 (1995). } If bargaining becomes too risky, small-time criminals will be afraid to cooperate with prosecutors, frustrating not one, but two goals: the conservation of judicial resources via guilty pleas, and the production of evidence against criminals.\footnote{ Dahlin (1995), {\it supra} note \ref{dahlin} at 1383.} Both of these goals are matters of the public interest, not just of the defendant's, so any one defendant's willingness to grant a waiver is not dispositive of whether he should be allowed to grant it, unlike in the case of other, purely personal privileges.\footnote{In Brooklyn Savings Bank v. O'Neil, 324 U.S. 697 (1945), the Supreme Court said that ``a statutory right conferred on a private party, but affecting the public interest, may not be waived or released if such waiver or release contravenes the statutory policy... '' An example is the Speedy Trial Act of 1994, 18 U.S.C. @ 3161, which cannot be waived because the public would lose from a delay in the criminal's imprisonment if he is guilty, even though he and the prosecutor might both find a long delay convenient. } To continue the market analogy, prosecutors will charge what ``traffic may bear,'' and this is a bad thing. I will quote a long excerpt from the Souter dissent to allow a full statement of this position: \begin{quotation} \vspace*{-6pt}\begin{small} The Rules draw no distinction between use of a statement for impeachment and use in the Government's case in chief. If objection can be waived for impeachment use, it can be waived for use as affirmative evidence, and if the government can effectively demand waiver in theformer instance, there is no reason to believe it will not do so just as successfully in the latter. When it does, there is nothing this Court will legitimately be able to do about it. The Court is construing a congressional Rule on the theory that Congress meant to permit its waiver. Once that point is passed, as it is today, there is no legitimate limit on admissibility of a defendant's plea negotiation statements beyond what the Constitution may independently impose or the traffic may bear. Just what the traffic may bear is an open question, but what cannot be denied is that the majority opinion sanctions a demand for waiver of such scope that a defendant who gives it will be unable even to acknowledge his desire to negotiate a guilty plea without furnishing admissible evidence against himself then and there. In such cases, the possibility of trial if no agreement is reached will be reduced to fantasy. The only defendant who will not damage himself by even the most restrained candor will be the one so desperate that he might as well walk into court and enter a naked guilty plea. It defies reason to think that Congress intended to invite such a result, when it adopted a Rule said to promote candid discussion in the interest of encouraging compromise.\footnote{ { United States v. Mezzanatto}, 513 U.S. 196, 37 (Souter dissent, 1995)} \vspace*{-12pt}\end{small} \end{quotation} Justice Souter's fear is that a second defect of the plea bargaining market will nullify the intent of Congress: the prosecutor will have all the bargaining power, and will always force a waiver of Rule 410 on the defendant. The Court's ruling would therefore ``render the Rules largely dead letters.''\footnote{ Ibid at 27. } As Eric Dahlin puts it, ``Because the prosecutor will almost always have a strong power advantage over the defendant, a true `bargain' will not result and most defendants will end up signing waivers, despite their desire to not do so.''\footnote{ Dahlin (1995), {\it supra} note \ref{dahlin} at 1381. } And, indeed, both with respect to the Mezzanatto waiver and to plea bargains generally, it seems that the agreements are unconscionably one-sided. Consider the following excerpt from a plea bargain in a different case. \begin{quotation} \vspace*{-6pt}\begin{small} It is further understood that Mr. Schulz must at all times give complete, truthful and accurate information and testimony and must not commit any further crime whatsoever. Should Mr. Schulz commit any further crimes or should it be judged by this Office that Mr. Schulz has given false, incomplete or misleading testimony or information, or has otherwise violated any provision of this agreement, this agreement shall be null and void and Mr. Schulz shall thereafter be subject to prosecution for any federal criminal violation of which this Office has knowledge, including, but not limited to, perjury and obstruction of justice. Any such prosecutions may be premised upon any information provided by Mr. Schulz, and such information may be used against him.\footnote{United States v. Stirling, 571 F.2d 708, 730 (7th Cir 1978).} \vspace*{-12pt}\end{small} \end{quotation} The agreement allows one of the two parties to be the judge of whether it has been violated, and if that is so judged, the judging party is free to keep all of its benefits from the bargain and make even fuller use of them. The government can say that Schult gave incomplete testimony regardless of what he actually gives, and then use what he said against him. If the market for plea bargains leads to such one-sided contracts, how can they be allowed? \noindent 3. An Analytic Framework for Plea Bargaining \noindent 3.1 Penalty Bargaining and Cooperation Bargaining Before I discuss these arguments from the vantage point of an economist and game theorist, it will be useful to distinguish between two kinds of plea negotiations: ``penalty bargaining'' and ``cooperation bargaining.''\footnote{\label{estrada} I have used the term ``penalty bargaining'' to mean the same thing as the ``charge bargaining'' of Mr. Estrada's oral argument before the Supreme Court. 1994 U.S. TRANS LEXIS 170 at 13. ``Penalty bargaining'' more aptly describes bargaining over sentence reduction for a given charge and can be translated to the civil suit context more easily. } Penalty bargaining occurs when the defendant has no information useful to the government, so each side's gain from bargaining is purely avoiance of the risk and cost of going to trial. Cooperation bargaining occurs when the defendant does have information to trade, and the gain from bargaining includes this information, which could be used in other trials. Penalty bargaining leads to an compromise sentence, somewhere between nonprosecution and the longest sentence for the most heinous conceivable charge. Cooperation bargaining, however, can lead to complete immunity for the defendant, or even an improvement in wealth or safety, since both he and the government can still benefit at the expense of third-party criminals.\footnote{ Myrna Raeder writes elsewhere in this issue of {\it Cardozo Law Review} that the Sentencing Guidelines revolutionized sentencing practices after 1989 and that this invalidates many of the assumptions of the present article. I think not. Both before and after the Guidelines, prosecutors had the discretion to agree to reduce the charges in exchange for a guilty plea, and judges had the discretion to reduce sentences of prisoners who cooperated with the prosecutor. The legal issues involved in enforcement of plea bargains remain the same, and while the Guidelines reduced flexibility in the terms of agreements, they by no means eliminated them. In fact, by increasing the predictability of sentences, the Guidelines may have made plea bargaining easier and more attractive-- the the trial judge's idiosyncrasies matter less. As Professor Raeder notes, the Guidelines specifically allow for departures based on the defendant's ``substantial assistance,'' ``acceptance of responsibility,'' and being a minor participant in the crime. At any rate, note that the model to be used below can be restricted to bargaining between a plea of innocent and a plea of guilty to a lesser charge with a penalty fixed by law. The two categories of penalty and cooperation bargaining, though perhaps not their names, apply to civil suits also, where the amount of the settlement is unimpeded by guidelines. Most civil settlements are to avoid the cost of trial, like penalty bargaining. It may also happen, however, that the defendant will testify against other defendants in exchange for a reduction in his own liability. } The law has not always recognized either the similarities or the differences between the two forms of plea negotiations. Rule 410 has been applied to immunity discussions, an easy extension to plea bargains with null sentences. \footnote{ See United States v. Boltz, 663 F. Supp. 956 (D. Alaska 1987). } Cases in which a defendant discussed reduction in other defendants' sentences in exchange for his cooperation, however, have been excluded, and although these are perhaps excludable from ``plea bargaining'' as a semantic matter, clearly departs from the intent of Rule 410.\footnote{ See U.S. v. Robertson, 582 F2d 1356 (5th Cir 1978); U.S. v. Doe, 655 F2d 920 (9th Cir 1980). } The gains from defendant cooperation are a key element because if defendant cooperation is not an important part of the process, there is no need for the defendant to say anything during the plea negotiations. If bargaining were simply over the length of the sentence, the defendant's lawyer, acting as his agent, could conduct the negotiations, reducing the risk of adverse disclosure. It is when the defendant wishes to show what information he has has available to help the prosecutor in other cases that the defendant must speak for himself and risk admissions that may hurt him if the plea bargain fails. Rule 410 is only important when defendant cooperation is important. As Mr. Estrada of the Justice Dept. said at oral argument, \begin{quotation} \vspace*{-6pt}\begin{small} In the first category, it will be a waste of time for everyone involved even to meet with the defendant, because charge bargaining, which is the usual type of plea bargaining that has traditionally been known to this Court, really usually only involves the defense lawyer calling the prosecutor on the phone and arguing about what his client did and what he thinks an appropriate sentence or charge might be. And in the first class, the charge bargaining class of cases, we never ask for a waiver. We never ask to meet with the client, or hardly ever, because it would be a waste of their time and ours, and we only do it in cases that involve cooperation... \footnote{{\it Mezzanatto} oral argument, {\it supra} note \ref{estrada} at 13.} \vspace*{-12pt}\end{small} \end{quotation} Mr. Estrada was wise to make this point in his argument because the justices, like most of us, tend to think of penalty bargaining when we hear the words ``plea bargaining'', and penalty bargaining is quite different from cooperation bargaining. As one of the Justices said in questioning Mr. Mezzanatto's attorney, \begin{quotation} \vspace*{-6pt}\begin{small} ... I'm interested in the same point, because frankly it had not occurred to me until I heard the Government's argument. He says, look, plea bargains go on all the time, and they're not under this procedure. They're not in a context in which the defendant is present with his counsel making statements to the United States Attorney. That context is reserved for questions when they want the witness' cooperation both by disclosing information that he knows so that they can have further investigation, number 1, and number 2 so that they can use his testimony at trial, so that the plea bargaining you're talking about is just not really affected in most cases by this rule.\footnote{Ibid. at 30.} \vspace*{-12pt}\end{small} \end{quotation} If the gain from cooperation is large, there is a substantial benefit to both prosecutor and defendant from reaching an agreement. The prosecutor gets information useful in other cases, while the defendant gets a reduction in his sentence. Even if there is no reason to waive Rule 410 in all plea bargaining, the gain from cooperation could provide a reason to waive Rule 410 in particular cases where the defendant can provide useful information. The question remains, however, of why waiver is necessary for cooperation bargaining to be effective. When Rule 410 is waived, does the prosecutor gain something which the defendant does not lose? The next section of this article will address that question. \bigskip \noindent { 3.2 The Model. } In thinking about either penalty bargaining or cooperation bargaining, it will be helpful to use a formal model to clarify the different situations and their parameters. Let us assume that the defendant and prosecutor both believe that the penalty if the case goes to trial will be $T$.\footnote{Nothing in the present analysis would be significantly affected if the model were to allow for a probabilistic distribution of possible penalties, including acquittal, at trial.} Let the legal costs of going to trial be $C_d$ and $C_p$ for defendant and prosecutor.\footnote{The defendant will often be represented by a public defender, in which case $C_d =0$ and $C_p$ includes the legal costs of both sides, since the government bears the cost of public defenders. } The expected payoffs are then $T - C_p$ for the prosecutor and $-T - C_d$ for the defendant if they go to trial. The defendant may have the additional bargaining chip of information useful in other cases. Denote the value of this cooperation to the prosecutor by $Z_p$, and the cost to the defendant by $Z_d$. The values $Z_p$ and $Z_d$ are measured in expected years of sentence in the immediate case; if $Z_p=2$, it means that the prosecutor would surrender 2 years of sentence in return for cooperation. If the two parties settle for a guilty plea with a penalty of $X$, the payoffs are $X+ Z_p$ for the prosecutor and $-X-Z_d$ for the defendant. The sum of the payoffs equals $X+ Z_p- X-Z_d =(Z_p -- Z_d)$, compared to a sum of $T - C_p -T - C_d= -C_p -C_d$ if the case goes to trial. The surplus to be split is therefore $C_p + C_d + Z_p-Z_d$. The prosecutor's payoff is \begin{equation}\label{e01} \pi_p = \left| \begin {array}{ll} T -C_p &{\rm at \; trial}\\ X + Z_p & {\rm with \; a \; bargain}\\ \end{array} \right. \end{equation} The prosecutor will agree to a bargain if \begin{equation}\label{e02} X+Z_p \geq T-C_p \\ X \geq T -C_p-Z_p. \end{equation} The defendant's payoff is \begin{equation}\label{e03} \pi_d = \left| \begin {array}{ll} -T -C_d &{\rm at \; trial}\\ -X - Z_d & {\rm with \; a \; bargain}\\ \end{array} \right. \end{equation} The defendant will agree to a bargain if \begin{equation}\label{e04} -X-Z_d \geq -T-C_d \\ X \leq T +C_d-Z_d. \end{equation} For a bargain to be agreed to by both sides, it must be that $\overline{X}$ satisfies both the preceding inequalities, so \begin{equation}\label{e05} T -C_p-Z_p \leq X \leq T +C_d-Z_d. \end{equation} Figure 1 illustrates the possible payoffs of the defendant and prosecutor. If the case goes to trial, the payoffs are the ``threat point'' of $(T - C_p, -T - C_d)$. If a plea bargain is reached, the prosecutor's payoff might be as low as $T - C_p$ or as high as $T + C_d + Z_p-Z_d$, which happens if $X = T + C_d -Z_d$. The prosecutor's payoff cannot go below the lower bound because he would refuse the plea bargain, or above the upper bound because the defendant would refuse it. Similiarly, the defendant's payoff after a plea bargain might be as low as $-T - C_d$ or as high as $-T +C_p + Z_p-Z_d$, which happens if $X = T - C_p -Z_p$. \epsfysize=3in \epsffile{mezz2.eps} \begin{center} {\bf Figure 1: Bargaining Over the Penalty} \end{center} Figure 1 shows these bounds on the plea bargain payoffs by light horizontal and vertical lines. Between the bounds is a continuum of possible payoffs from intermediate sentences within the bargaining range, the heavy line which is he hypotenuse of the triangle. \noindent 4. Cooperation Bargaining \noindent 4.1 The Problem of Commitment For studying cooperation bargaining, let us assume that the plea bargained penalty is exogenously fixed at $X = \overline{X}$; the parties must agree to that penalty, or let the trial takes its course. This is somewhat realistic; the U.S. Sentencing Guidelines do not give perfect flexibility to prosecutors. From expression (\ref{e05}), we will assume that $\overline{X}$ takes the moderate level in expression (\ref{e6}), since otherwise one party or the other will refuse the bargain. \begin{equation}\label{e06} T -C_p-Z_p \leq \overline{X} \leq T +C_d-Z_d. \end{equation} Let us also assume that cooperation is necessary for the prosecutor to be willing to accept the bargain. This means that it is additionally true that \begin{equation}\label{e07} \overline{X} T - C_p; \end{equation} that is, the plea bargain sentence is so low that the prosecutor would prefer to go to trial except for the value of the cooperation he receives.\footnote{If inequality (\ref{e07}) were false, then the rest of the discussion would be vacuous, since the situation would be one in which the motive of penalty bargaining would be sufficient to reach agreement whether or not cooperation were possible.} The other essential element of a game besides the players and payoffs is the order of the moves. Consider the following alternatives: \noindent GAME 1\\ 1. The prosecutor offers sentences of $T$ or $X$ for cooperation.\\ 2. The defendant accepts or not.\\ 3. The prosecutor chooses a sentence.\\ 4. The defendant testifies or not for The prosecutor. \noindent GAME 2.\\ 1. The prosecutor offers sentences of $T$ or $X$ for cooperation.\\ 2. The defendant accepts or not.\\ 3. The defendant testifies or not for The prosecutor. \\ 4. The prosecutor chooses a sentence.\\ These two games present in stylized form one of the great problems of cooperation agreements. If performance is not simultaneous, the second party to perform needs an incentive to carry out his end of the bargain. In Game 1 the defendant will refuse to testify, despite his acceptance, and in Game 2 the prosecutor will choose the sentence of $T$ rather than $X$. In either game, one party or the other, anticipating non-performance will refuse to enter into an agreement. Thus both parties lose the benefit of the bargain because of non- enforceability.\footnote{This is also a problem if waivers are unenforceable, since the prosecutor will be unwilling to make concessions in exchange for the waiver if the defendant can costlessly breach by asking the court to invalidate the waiver. ``If, therefore, it appears that the plaintiff in error did waive his rights under the act of Congress...he ought not to be permitted to raise the objections at all. If he may, he is allowed to avail himself of what is substantially a fraud.'' Shutte v. Thompson, 82 U.S. 151, 159 (1872).} This is the central problem problem of contract law generally, and a function of the state is to enforce contracts by penalizing breach. As discussed earlier, plea agreements are largely but not entirely controlled by contract doctrine, so the games are more complicated than has been described. In Game 2, if the prosecutor breaches, the defendant can require the court to hold him to his bargain. Game 1, however, is more of a problem. The doctrine of double jeopardy makes it difficult for a court to require the defendant to ``give back'' the sentence reduction he obtained by breached promises of cooperation. Consider the timing problems in the case of {\it Ricketts v. Adamson}.\footnote{Ricketts v. Adamson, 483 U.S. 1 (1987). } Adamson pled guilty to second-degree murder of a newspaper reporter in return for his cooperation and a specified prison term of some twenty years. He did testify against two accomplices, who were convicted of first-degree murder, and was sentenced, the sentence having been delayed until after his testimony. The accomplices, however, had their case reversed by the Arizona Supreme Court and remanded for retrial, at which point Adamson refused to cooperate without the further concession of release from custody following the retrial, and the trial court refused to compel him. The state's response was to vacate the second-degree murder conviction, on the grounds that Adamson had violated the agreement, and to charge Adamson with first-degree murder. He moved to quash the information and lost, after which he did offer to testify, but had his offer refused.\footnote{Adamson v. Superior Court of Arizona, 125 Ariz. 579, 611, P2d 932 (1980).} He was convicted and sentenced to death, and the United States Supreme Court upheld the proceedings. What this illustrates is how hard it is to pin down a witness, and the contractual nature of the agreement. The United States Supreme Court opinion is as much about contract law as constitutional law, and the dissent, in particular, focusses on contract doctrines such as anticipatory repudiation. Contract doctrine has considerable effectiveness in pinning down the prosecutor, so it can solve the problem of Game 2. Vacating convictions and repeating trials, however, {\it Adamson}'s solution to the problem of Game 1, is cumbersome, and unavailable if the prosecutor's performance takes the common form of recommending a sentence reduction rather than accepting a guilty plea to a lesser charge. What is needed, therefore, is a way to turn a situation that is naturally Game 1 into Game 2, so that both parties can anticipate the agreement being enforced and so will enter into it. Since the agreement is enforceable against the prosecutor but not the defendant, some way must be found to allow the defendant to perform first. The obvious way, which is in fact widely used, is to delay sentencing, if not trial, of the defendant until after he has given his testimony in other trials. This is legal,\footnote{ ``...[T]he great weight of modern authority, particularly in the federal courts, is that, in guilty-plea cases, the postponement of plea and sentence is unobjectionable.'' Hughes (1992), {\it supra} note \ref{hughes}, at 25.} or the government could even wait until after sentencing to recommend a sentence reduction.\footnote{ Rule 35(b), FED. R. CRIM. P., provides that up to a year after sentencing the court may reduce a sentence on a motion of the government.} Postponement of sentencing, however, has two problems. First, it is difficult to know how long to postphone sentencing. As {\it Adamson} shows, just waiting until the other trial seems to be concluded may not be enough. Second, there remains the problem of the quality of the defendant's performance. The prosecutor can be pinned down to a quite specific promise-- - a particular charge, or a specific sentence. The defendant, however, is providing ``cooperation'' . At its most specific, this might just involve repeating in court under oath the story that the defendant told the prosecutor earlier in the negotiations. More often, however, the prosecutor wants the defendant to answer additional questions, to peform creditably under cross examination, and to be available for cooperation in other matters that arise during investigations, pre-trial preparation, sentencing proceedings, appeals, and so forth. Even if the performance were limited to repeating an earlier story, the defendant might effectively breach by adopting an unbelievable demeanor on the stand. Thus, we are left with the problem of performance quality. It is helpful to view the situation as similar to when a seller wishes to guarantee the quality of his product to a buyer, but is not legally required to keep his promises about quality. Many sellers offer money-back guarantees for this reason, even though a dishonest consumer could take advantage of such a guarantee to return products after use even when the quality is satisfactory. Defendants are in this position. Indeed, it is sometimes the case that defendants who are accomplices are in competition with each other to sell cooperation to the government. This is the classic Prisoner's Dilemma of game theory, in which two accomplices each face a choice between confessing and remaining silent. If they both confess, they both receive heavy sentences, while if they both are silent, they both receive light sentences. Each will confess, however, because if one confesses and the other does not, the one who confesses will receive the lightest sentence possible. Defendants in this situation indeed have weak bargaining power vis a vis the government, but that is not a bad thing for society.\footnote{See chapter 1 of my book, Eric Rasmusen, GAMES AND INFORMATION: AN INTRODUCTION TO GAME THEORY (2nd Ed., 1994) for further explanation of the Prisoner's Dilemma. The Prisoner's Dilemma is also a metaphor for the free market; competing sellers bid the price of their product down in an effort to attract the consumer, even though they could do better if they jointly kept prices high.} This Prisoner' Dilemma is a standard trick of prosecutors, and has been for hundreds of years. Even in eighteenth century England, it was commonplace to set up a competition among accomplices to a crime in which all confessed because one would be chosen to be a ``crown witness'' and be given immunity in exchange for testifying against the others. In such a situation, defendants were positively eager to proffer evidence, even without guarantee that any particular one of them would become the crown witness\footnote{ Langbein (1983), {\it supra} note \ref{langbein} at 88. } Their bargaining power was minimal, but this was not a bad result. Increasing their bargaining power by eliminating the practice would be like forbidding car sellers to reduce their prices to try to attract business fom each other--- a good deal for the sellers, but a bad one for the public. This idea may explain why lack of bargaining power is not something to worry about, but it does not help with the basic problem of guaranteeing the quality of the information the defendant offers. Even if he offers full cooperation for a mere two months' reduction in sentence, the prosecutor will turn him down unless the defendant can guarantee he will actually carry out his promise. How can this be done? \bigskip \noindent {\it 4.2 Waivers As Commitment Devices for the Defendant} We now return to waivers of Rule 410, which may provide solutions to the problem of enforcing otherwise non-credible promises of defendant cooperation. Let us also return to the apparently unconscionable plea agreement from {\it Stirling} that I quoted earlier. I suggest that such agreements are actually one of the solutions we are seeking. This resolves the puzzle of why plea agreements so often seem to give enormous advantages to the prosecutor. It also explains why waivers of Rule 410 became so common in cooperation bargaining: it is because they increased the total benefits from the agreements, to the advantage of both buyer and seller of cooperation. Waivers became common in the plea bargaining marketplace because the waiverless cooperation agreement was a product that didn't sell.\footnote{ Judge Sneed of the 9th Circuit said, ``Given the precision with which these rules are generally phrased, the comparative recentness of their promulgation, and the relative ease with which they are amended, the courts can afford to be hesitant in adding an important feature to an otherwise well-functioning rule.'' United Statesv. Mezzanatto, 998 F.2d 1452, 11 (9th Cir 1993). A market analysis suggests that since before enactment of Rule 410 in 1975 defendant statements were used for impeachment without defendant's bargaining for exclusion, and after 1975 waivers of Rule 410 became common, that it was the old rule which was well-functioning. Congress may have given defendants a right that they did not want to have. (In analogy, consider the position of a potential mortgage borrower who is granted immunity from foreclosure on his house by Congress. Will he be grateful for being made such an unattractive borrower? ) } The {\it Stirling} agreement is not unique in its one-sidedness. The standard agreement in use in the Eastern District of New York is quite similar to the one in {\it Stirling}: \begin{quotation} \vspace*{-6pt}\begin{small} If the Office determines that [the cooperator] has cooperated fully, provided substantial assistance to law enforcement authorities and otherwise complied with the terms of this agreement, the Office will file a motion with the sentencing court setting forth the nature and extent of [his] cooperation. . . . In this connection it is understood that the Office's determination of whether [the cooperator] has cooperated fully and provided substantial assistance, and the Office's assessment of the value, truthfulness, completeness and accuracy of the cooperation, shall be binding upon [him].... Should it be judged by the Office that [the cooperator] has failed to cooperate fully, or has intentionally given false, misleading or incomplete information or testimony . . . [he] shall thereafter be subject to prosecution for any federal criminal violation of which the Office has knowledge, including, but not limited to, perjury and obstruction of justice.\footnote{ Hughes (1992), {\it supra} note \ref{hughes}, at 38.} \vspace*{-12pt}\end{small} \end{quotation} A plea agreement can also condition the amount of recommended punishment on the value of the information given. In {\it United States v. Dailey}, the agreement specified that if the defendant ``fully cooperated'' then the Government would recommend a sentence between 10 and 20 years, choosing the length ``depending principally upon the value to the Government of the defendant's cooperation''. If the Court found, upon hearing evidence, that the defendant did not fully cooperate, the sentence would be 35 years.\footnote{ United States v. Dailey, 759 F.2d 192, 194 (1st Cir 1985). } \begin{quotation} \vspace*{-6pt}\begin{small} The defendant agrees to fully cooperate, as defined in Paragraph 2. If, at the time of sentencing on the Maine indictment, the defendant has fully cooperated with the United States, as defined in Paragraph 2, the Government will recommend a specific term of imprisonment which does not exceed twenty (20) years and, depending principally upon the value to the Government of the defendant's cooperation, the Government, in its sole discretion, may recommend a sentence of ten (10) years; the defendant may argue for a sentence less than the Government's recommendation; in any event, the Court shall impose a sentence no greater than that recommended by the Government. If, at the time of sentencing on the Maine indictment, the Government presents evidence and the Court finds by a preponderance of the evidence that the defendant has not fully cooperated, as defined in Paragraph 2, then the court shall sentence the defendant to a term of imprisonment of thirty-five (35) years. The Court shall not sentence the defendant to pay a fine on either the Maine or Oregon indictment. \vspace*{-12pt}\end{small} \end{quotation} The Government's recommendation if there was full cooperation was ``in its sole discretion.'' Thus, this agreement allows the Court to decide part of the contractual provisions-- whether there was full cooperation-- but one party to the contract can decide another-- how valuable the cooperation was.\footnote{ The issue in the appellate case was whether conditioning the sentence recommendation on the value of the information unduly encouraged perjury. The Court ruled that it did not. } These agreements seem one-sided.They are, in fact, one-sided. But that might be to the benefit of both parties. The prosecutor's office has a reputation to maintain; the defendant does not. The theory of repeated games suggests that the prosecutor can be trusted to keep his promises, but the defendant cannot. Let me try to explain. Consider first the situation of two people trying to make an agreement. Doe has a reputation to maintain, and would suffer greatly in his future dealings if he ever broke an agreement . Roe is involved in this kind of negotiation only this one time, and has no need to maintain a reputation. Suppose the agreement cannot be enforced in court. Doe and Roe would lose the benefit of the bargain, because Doe will not agree to something he knows Roe can violate with impunity. One solution is for Roe to voluntarily put himself at Doe's mercy. If Roe gives a large sum of money to Doe as a hostage for good behavior, and part of the agreement is that Doe will return the money if Roe keeps his side of the bargain, then Doe becomes willing to enter into the agreement. Roe also benefits, because he knows that Doe will keep his word and return the hostage money in order to preserve his reputation for fair dealing. Roe would like best to enter the original agreement and break it, but since Doe will not agree to that, he is happy to settle for adding a hostage clause and keeping his part of the agreement.\footnote{The same argument applies if it is known that Doe is an honest man who keeps his agreements for moral reasons rather than for reasons of reputation, or if Doe is vulnerable to court judgements and Roe is not. } Let us now relate this to criminal prosecutions. The prosecutor is in the position of Doe, and the defendant in that of Roe. The prosecutor's office has a reputation to maintain, and can be better trusted to keep its agreement even if it could benefit in an individual case by a violation. The defendant is more concerned about the immediate case, and also is given the right by law to change his mind about his plea. Thus, the defendant, though eager to obtain a plea agreement, may not be able to credibly promise cooperation. For cooperation, with its benefit $Z_p$ to the prosecutor, to be added to the bargaining surplus requires some way for the defendant to be punished if he fails to keep his end of the bargain. Waiver of Rule 410 is such a way, equivalent to the hostage money in the Doe-Roe example. Having waived Rule 410, the defendant will keep to the bargain, out of fear that the prosecutor will use his statements against him in some future trial-- if not the offense to which he pleads guilty, to a different offense. The defendant desires this, because he wants the prosecutor to believe that he, the defendant, has a strong incentive to cooperate. The vagueness of so many plea agreements may be another sign of reputational asymmetry. Plea agreements commonly state that the prosecution will recommend appropriate leniency to the judge, without specifying precisely what will be recommended. The prosecutor can be trusted, because if he breaks an agreement in one plea bargain, he will face the distrust of future defendants. Thus, the agreement puts the advantage in the hands of the prosecutor. This is entirely dependent on the prosecutor's desire to maintain his office's reputation for carrying out its agreements in order to be able to make agreements with future defendants, or on his integrity and honesty in keeping agreements even when courts would not be able to hold him to them. This is entirely dependent on the prosecutor's desire to maintain his office's reputation for carrying out its agreements in order to be able to make agreements with future defendants, or on his integrity and honesty in keeping agreements even when courts would not be able to hold him to them. That the prosecutor has a reputation to maintain is particularly important because it is not enough for him to contract regarding the formal charge. Evidentiary rules for sentencing hearings are much looser than for trials, and the judge can take into consideration conduct by the defendant which did not result in an indictment, or even conduct which led to an acquittal.\footnote{A standard cite is Williams v. New York, 337 U.S. 241 (1949), in which a state judge overruled a jury suggestion for leniency and sentenced a murderer to death based on criminal and moral offenses for which he was not tried. Justice Black, with the concurrence of the entire Supreme Court except for Justices Rutledge and Murphy, wrote that this was entirely in accord with judges' ``age-old practice of seeking information from out-of-court sources to guide their judgement'' (at 250). The Court has continued to uphold this practice, including recently in BMW v. Gore, 116 S. Ct. 1589 (1996) at note 19 and United States v. Watts, 117 S. Ct. 633 (1997). In {\it United States v. Watts}, the 9th Circuit, contrary to {\it Williams} and to every other circuit, had held that a judge sentencing for one offense could not rely on facts concerning a different offense for which the defendant had been acquitted. The Supreme Court reversed per curiam, with Justice Kennedy dissenting from the denial of oral argument.} Despite the economist's cynicism, it is not implausible to suppose that prosecutors are more likely than criminal defendants to keep their word from motives of conscience, but regardless of that, they do represent permanent offices with reputations to maintain, and hence may have little temptation to cheat in present cases if that would jeopardize future ones. Even if prosecutors were not required by courts to keep their bargains, we would expect them to keep them in all but exceptional cases.\footnote{ \label{langbein} Professor Langbein quotes crown agent Henry Fielding (better known for novels such as {\it Tom Jones} ) as saying in 1751 about the immunity of the crown witness that ``[I]t is true, he hath no positive Title [no entitlement to nonprosecution].. . But the Practice is as I mention, and I do not remember any Instance to the contrary.'' John Langbein, Shaping the Eighteenth..., 50 U. CHI. L. REV 1, 92 (1983). } In this way, we have an explanation for the usefulness of the Rule 410 waiver to both sides, and of why it is necessary to the creation of the cooperation value $Z_p$. When Rule 410 is in place, the defendant will promise to cooperate later, but then back out, so that cooperation never does occur, and nobody obtains $Z_p$ even if a plea bargain has been reached. When Rule 410 is waived, the defendant obtains a lower sentence in return for cooperation, and the prosecutor knows that the defendant will cooperate out of fear that his statements made earlier will be used against him. \bigskip \noindent { 4.3 Waivers as Incentives for Truthfulness} A different explanation for Rule 410 waivers is also based on the quality of cooperation, but focusses on whether the defendant will be a satisfactory witness even if he does testify. Again, the underlying idea is that the purpose of the waiver is not to affect the current case, but to elicit truthful information from the defendant that the prosecutor can use in a separate case. This explanation will be better at explaining the Mezzanatto waver {\it per se}, as opposed to waivers for the case-in-chief or plea agreements that give great discretion to the government side. A common part of plea bargaining is the defendant's offer to incriminate his accomplices. One difficulty is enforcing the defendant's cooperation once an agreement is reached. Another problem, however, is whether the defendant will make a credible witness. If the defendant lies to the prosecutor, and the defense attorney in the other case detects the lie during the discovery process for that case, the defendant's credibility as a witness is damaged. The prosecutor will be interested in making a deal with the defendant only if he thinks the defendant will tell the truth, and tell the truth completely enough to be a credible witness. Credibility is a special problem when the testimony is part of a plea bargain. Ordinarily, a party may not bolster the credibility of his witness until that credibility has been attacked by the other side.\footnote{Federal Rule of Evidence 608(a)(2).} It has been ruled, however, that the Government may introduce plea agreements in direct examination in order to avoid the jury drawing the inference that it was trying to hide a source of witness bias. \footnote{United States v. Edwards, 631 F.2d 1049, 1052 (2nd Cir 1980). ``Admission of this evidence is permitted in order to avoid an inference by the jury that the Government is attempting to keep from the jury the witness' possible bias.'' At least seven other circuits are even more willing to allow introduction of plea agreements, as explained in United States v. Spriggs, 996 F.2d 320, 324 (D.C. Cir 1993). } The prosecutor wants to be able show a good side of the witness's incentives to the jury. Judge Trott suggests that if the prosecutor can point to a plea agreement that is void if the defendant is caught lying, that can (quite rationally) increase the jury's estimate of his veracity. \begin{quotation} \vspace*{-6pt}\begin{small} One aspect of the witness that you can emphasize is his motive to tell the truth. Point out that he can only have a motive to tell the truth because that is what will get him what he wants. Lies will only destroy the deal and cause him to be prosecuted for perjury: He wants to stay out of jail. All he has to do to stay out is tell the truth, not lie. Lies will put him right where he doesn't want to be, in prison. His motive based on the evidence and the record can only be to tell the truth!\footnote{\label{trott} Stephen S. Trott, Words of Warning for Prosecutors Using Criminals as Witnesses,47 Hastings L.J. 1381, 1429 (1996). } \vspace*{-12pt}\end{small} \end{quotation} This is essential, because a witness who lacks credibility is worse than no witness. ``What happened in this case is that their worst witnesses spilled over and poisoned the better witnesses. We were able to create not just reasonable doubt but to prove perjury. And when you prove perjury about witnesses A, B, and C, then the jury automatically distrusts witnesses D, E, and F.''\footnote{Ibid at 1389 (Citing Jim DeFede, The Impossible Victory, Miami New Times, Feb. 29, 1996, at 1). } It is dangerous to use any unreliable witness, but accomplices are the worst, because (a) they have a special temptation to lie to obtain leniency, (b) they are criminal, and hence both lack credibility and, in fact, do lack the habit of truthfulness, and (c) the defendant knows them better than the prosecutor does, and can seize upon their weaknesses. As Judge Trott says, \begin{quotation} \vspace*{-6pt}\begin{small} The defendant knows more about the informer than you do! This advantage may enable the defendant to mount an attack on cross examination, etc., based on facts or circumstances of which you are unaware and about which the informer has not told you. To avoid being caught unprepared, ask the informer what the defendant might bring up to discredit him or his testimony. Take your time on this because you're now probing for information that the informer may not want to tell you.\footnote{Ibid at 1404. } \vspace*{-12pt}\end{small} \end{quotation} It is often difficult for criminals to tell the truth. It is only human nature to want to make oneself look better, and any defendant has a need to improve his image, or he would not be a defendant. Even if his testimony is immunized, vanity is likely to lead to inaccuracy. Yet again, let me quote Judge Trott: \begin{quotation} \vspace*{-6pt}\begin{small} Impress the requirements of absolute honesty and full disclosure on the witness' attorney and ask the attorney to have a private discussion with the witness to try to pound this into the witness' skull. These witnesses invariably hold back information that makes themselves ``look bad.`` It is devastating in front of a jury to find out that the first thing such a witness did was lie to the prosecutor or the case agent!... Mistrust everything he says. Be actively suspicious. Look for corroboration on everything you can; follow up all indications that he may be fudging. Secure information on the witness' background: mental problems, probation reports, prior police reports. Contact prior prosecutors who have either prosecuted the witness or used him in court and read the sentencing memoranda from previous cases. What do the prosecutors think about his credibility? How did the jurors react to him? Was he a helpful witness or was he more trouble than he was worth?\footnote{Ibid at 1403, 1406. } \vspace*{-12pt}\end{small} \end{quotation} Even if cooperation simply takes the form of providing information to be used in investigations, its accuracy is important, since it may be used by undercover agents whose lives may be threatened if they rely on inaccurate information. Commonly, however, cooperation takes the form of testimony in other trials, in which case the credibility of the defendant in front of a jury is crucial. Having good information is then not enough; the defendant must also be safe against impeachment in cross examination, something difficult even for truthful witnesses in the face of skilled attorneys. One difficulty is whether the potential witness has an unsavory background which could be used to impeach him. It is worth noting that although a Mezzanatto waiver of Rule 410 for purposes of impeachment is worthless if the defendant would not testify on his own behalf anyway, it has bite precisely when the defendant would make a good government witness--- because it is only if his background and demeanor are credible enough to be useful to the government that his testimony would also be useful to himself.\footnote{The Chicago Jury Study found that in the 1950's, 91\% of defendants without prior criminal records and 74\% of those with prior records chose to testify. H. Kalven \& H. Zeisel, THE AMERICAN JURY 146 (1966). Gary Mezzanatto, with a wife and three children, Vietnam special combat awards, and no criminal record, had all the credentials for a good witness either for himself or the government. {\it Mezzanatto} Respondent Brief, {\it supra} note \ref{dahlin}, in text near note 1. } But that credibility is hard to determine before the cross examination itself. \begin{quotation} \vspace*{-6pt}\begin{small} A prosecutor often may be unable to judge the value of the defendant's information to other investigations, or the value of his testimony in other trials, without extensively questioning the defendant as part of ``plea discussions`` under Rules 11(e)(6) and 410. But the prosecutor may justifiably be unwilling to use the facts proffered by the defendant as the basis for charging or convicting other persons if the defendant wishes to retain the ability to change his story under oath when his own liberty is at stake. It is reasonable for a prosecutor to conclude that a defendant who is willing to cooperate in the prosecution of others only on those terms has little of value to offer to the government.\footnote{\label{usbrief} United States v. Mezzanatto, Brief for the United States (1994).} \vspace*{-12pt}\end{small} \end{quotation} This explains why the government may find a Rule 410 waiver for impeachment useful. Such a waiver gives every incentive for the defendant to be truthful in his plea negotiation interview, and if he cannot be truthful with that incentive, he is unlikely to be useful on the witness stand.\footnote{``If the government lacks effective means of policing the truthfulness of the information proffered by the defendant -- such as the protection supplied by the agreement condemned by the court below -- the defendant's testimony will be subject to ready impeachment when the government attempts to use it in the prosecution of others.'' Ibid, note 8.} This even applies to a defendant who is not willfully lying. Willful or not, if he cannot tell the truth, the prosecutor does not want to buy his testimony. Mr. Mezzanatto's brief notes that Rule 410 waivers will hurt defendants who give inaccurate plea statements because of ``confusion, mistake or faulty memory.'' Such people, however, who cannot tell the truth even when their prison term is at stake, are precisely the sort of witnesses the government does not want on the stand.\footnote{ {\it Mezzanatto} Respondent Brief, {\it supra} note \ref{dahlin}, in text near note 22. } Why will the defendant be more truthful having signed a waiver? Assume that the defendant can either be completely truthful or tell some lies during plea bargaining. If he is truthful, and has signed a waiver, he increases his sentence by $L_d$ if the case goes to trial. If he lies, then $L_d$ can take one of two values. It is zero without a waiver. With a waiver, it is zero if he is not caught lying, but with probability $\gamma$, the prosecutor catches him out, and $L_d= W$. He will refrain from lying if \begin{equation} \label{e1} L_d < (1-\gamma) (0) + \gamma W. \end{equation} Thus, we would expect the prosecutor to ask for waivers when he expects them to deter lying and when he derives value in other cases from truthtelling in this one, which is when $\gamma$, the probability of detecting a lie and $W$, the adverse consequences to the defendant from being caught in a lie, are sufficiently large. If they are, then waiver of Rule 410 will induce the defendant to provide truthful statements that can be used in other trials, and $Z_p$ will be positive. This also tells us when a waiver would not be useful. If the defendant would not testify in his own case under any circumstances, because his credibility is already low or he does not want the jury to hear impeachment evidence of his bad character, then he cannot make the value of his cooperation credible to the prosecutor, and the prosecutor has no incentive to ask a waiver. If he did, however, the defendant would grant it, since if he does not intend to testify, the waiver is costless to him. This argument also suggests a related reason for limited waivers: that the defendant wants to give credibility to his statements for the sake of the plea bargaining itself, even if he has no cooperation to sell. Rule 410 specifically excludes statements made under oath from inadmissibility in future criminal proceedings for perjury. This clause is of advantage to the defendant as well as to the prosecutor, for it allows the defendant to claim more persuasively that he is telling the truth if he is under oath. Limited waivers for impeachment might serve the same purpose, allowing the defendant to increase his credibility by imposing upon himself penalties if he lies. As a result, he may be able to more plausibly demonstate the strength of his case and obtain a better plea bargain. \bigskip \noindent 5. Penalty Bargaining So far I have been discussing cooperation bargaining, which is the main context for waivers of Rule 410. Every cooperation agreement, however, also contains elements of the second kind of plea bargaining: penalty bargaining. In pure penalty bargaining, cooperation is not at issue, but both sides wish to avoid trial and they must decide how to split the gains from settling out of court. The economics of bargaining seems a natural tool for addressing the question of whether waivers increase or decrease the number of successful penalty bargains. Economists have used formal modelling to address the issue of whether plea bargaining increases or decreases social welfare, but the emphasis in those articles is on risk aversion and how the parties interpret each others' offers rather than on the impact of failed plea bargains on what happens at trial.\footnote{ Gene Grossman and Michael Katz, Plea Bargaining and Social Welfare, 73{ American Econ. Rev.} 749 (1983); Jennifer Reinganum, Plea Bargaining and Prosecutorial Discretion, 78 { American Econ. Rev.} 713 (1988). See also the survey by Robert Cooter and Daniel Rubinfeld, Economic Analysis of Legal Disputes and Their Resolution, 27 { J. Econ. Literature} 1067 (1989) } The model that comes closest to addressing the questions that arise when material from failed plea bargaining is introduced as evidence at trial is Daughety \& Reinganum.\footnote{Andrew Daughety \& Jennifer Reinganum , Keeping Society in the Dark: On the Admissibility of Pretrial Negotiations as Evidence in Court, 26 { RAND J. of Econ.}, 203 (1995).} They construct a model of civil litigation with which to answer the question of whether settlement offers ought to be admissible at trial if settlement fails to occur. They show that admitting settlement offers results in less settlement, because the parties have more incentive to bluff with tougher offers.\footnote{The model they use is a descendant of the model used in Jennifer Reinganum \& Louis Wilde, Settlement, Litigation, and the Allocation of Litigation Costs, 8 { RAND J. Econ.} 109 (1986). } Once the offers are admissible, such bluffing has the advantage that the offers can be admitted as evidence at trial even if the bluff fails to work. On the flip side, generous offers have the disadvantage that they hurt the side offering them if they are admitted into evidence. Rigorous analysis confirms what intuition suggests: allowing bargaining discussions to be admissible has a chilling effect, hampering the parties' efforts to come to an agreement because neither party will want to disclose any of their information. In the criminal context, the Daughety \& Reinganum discussion would apply to specific offers to plead guilty to greater or lesser offenses, or to specific government offers to mitigate penalties. Such offers are part of Rule 410, but the analysis is unsatisfactory for dealing with {\it Mezzanatto} for two reasons. First, what is disclosed in the Daughety \& Reinganum model is not direct information--- ``I was there on the night of November 24th''--- but settlement offers that signal a party's opinion of the strength of his case--- ``I will plead guilty to robbery, but not murder.'' The defendant does not have to worry about accidentally making a settlement offer, as he does about factual admissions, and a settlement offer cannot be used to impeach a witness or to rebut his testimony. Second, the model is addressed to the issue of inflexible legal rules, not waivers. Daughety \& Reinganum show that there are more settlements under a rule of inadmissibility of settlement demands. But this does not tell us why inadmissibility should be a mandatory rule, rather than a default rule that could be waived by mutual consent. Given a choice between two inflexible rules, inadmissibility is best in their model. Forbidding waivers seems superfluous, though, because the inefficiency of admissibility means that the party that desires admissibility would not be willing to give up enough to persuade the other side to grant admissibility. For a similar reason, Rule 410 is a sensible default rule. If disclosures during plea bargaining were admissible, there would be less plea bargaining because some defendants would be unwilling to bargain at so great a cost. Prosecutors would voluntarily agree to restrictions like Rule 410 on a case-by-case basis, because they too want to avoid trial. Making Rule 410 the default rule makes these case-by-case agreements unnecessary, and so reduces costs. That explains why Rule 410 is a good default rule, but not why it should be a mandatory rule. Under the reasoning of the previous paragraph, no defendant would want to sign a waiver, and so making waivers invalid should be unnecessary. In general, voluntary transactions benefit both sides of the transaction. If defendant and prosecutor both agree to a waiver, why should the law prevent it? The law allows plea bargaining in general, for example, which is effectively the waiver of a trial in return for a reduced sentence. The puzzle is what benefit the defendant obtains from waiving the Rule 410 exclusion. He can then enter into plea bargaining, which is itself beneficial, but if he is so eager to make a bargain, why does the prosecutor not take advantage of that in the terms of the plea bargain itself, rather than in the preliminaries? The general principle that trade is efficient does, however, have exceptions, which might provide an explanation. If the transaction has negative external effects on third parties, for example, it might be inefficient even if it benefits the two parties directly involved. That will not be a problem here, since the negative effect on third parties would be an increase in the probability of going to trial, which is not something that benefits either defendant or prosecutor. What raises more questions is whether the waivers might serve a strategic purpose for the prosecutor, so that the defendant would like to be restricted not to agree to them, and would gain more from such a restriction than the prosecutor loses. The analysis below will raise a number of both efficiency-enhancing and strategic explanations for waivers. Waivers might be efficient if it is more costly for prosecutors than for defendants to enter into negotiations, so that prosecutors need extra inducement. They may also be efficient if they help defendants make credible their ability to help prosecutors with testimony in other cases. On the other hand, waivers may be a strategic tool for the prosecutor to leverage up his bargaining power, in which case prosecutors might use them even if they reduce the rate of settlement. The discussion will begin with a simple model to show that it is not enough to say simply that the prosecutor has more bargaining power and can therefore extract the waiver as a concession. The next step is to see what is special about information disclosure during plea negotiations, and how prosecutors might use waivers either to balance advantages of the defendant or to exploit the prosecutor's own advantages. The last part of the discussion will be about implications of waivers for the defendant's ability to bargain using his potential testimony in other cases. \noindent 5.1 Use of the Waiver as a Bargaining Tool Let us take the model of earlier the paper and simplify it by assuming that the plea bargaining is purely penalty bargaining, not cooperation bargaining, because the defendant has no information to offer. In that case, $Z_p=Z_d=0$. The exact payoffs, the point on the hypotenuse in Figure 1 which is the actual outcome, depends on the bargaining power of the two sides. Let us use $\beta$ to parameterize the prosecutor's bargaining power, so that the settlement will yield him payoff $T -C_p + \beta( C_d+C_p)$, giving him fraction $\beta$ of the surplus from avoiding trial. If the prosecutor had no bargaining power, then $\beta =0$, and the settlement would be $X=T -C_p$. The defendant would cooperate, but would have bargained for such a low sentence in his own case that the prosecutor would be no better off than if he had gone to trial. If the prosecutor had all the bargaining power, then $\beta =1$, and the settlement would be $X=T +C_d$. The defendant would cooperate, but would have effectively given away his cooperation, and would be no better off than if he had gone to trial. We will assume that $\beta$ is strictly between 0 and 1, so both sides have positive bargaining power. The case of equal bargaining power, with $\beta = .5$, is the Nash bargaining solution conventionally used in economics.\footnote{See John Nash, The Bargaining Problem, 18 {Econometrica} 155 (1950); Ariel Rubinstein, Perfect Equilibrium in a Bargaining Model, 50 {Econometrica} 97 (1982). } A greater value, such as $\beta =.8$, would be appropriate if the defendant is poorly represented, so that the prosecutor gets most of the gains from the bargain. This would result in an outcome like Point A in Figure 1, where the sentence is actually greater than its expected value as a result of a trial, $T$. A plea-bargained sentence greater than the expected value of the sentence is not anomalous; it would result whenever, for example, the prosecutor has the greater bargaining strength, there is no benefit to the prosecutor in other cases from the defendant's cooperation, and the defendant has higher costs of going to trial, i.e., $\beta >.5$, $Z_p=Z_d=0$, and $C_d> C_p$. Despite this, the defendant still gains from the plea bargain, because he has avoided the cost of going to trial. The goal of plea bargaining is not for a litigant to do better than his opponent, or to reduce his opponent's welfare, but to do as well for himself as possible. A plea bargain that results in a sentence higher than the expected outcome at trial can still benefit the defendant because of the added certainty, speed, and trial cost savings. Conversely, a low sentence can still be of benefit to the prosecutor, for the same reasons.\footnote{It may well be that $\beta$, $C_d$, and $C_p$ are related. The prosecutor will have more bargaining power (higher $\beta$) if he has better staff relative to the defendant, which means higher prosecutor trial costs (higher $C_p$) and lower defendant trial costs (lower $C_d$). Thus, the case where $\beta =.8$ and $C_d > C_p$, so that $X$ is much greater than $T$ may be implausible. If the prosecution has an expensive lawyer, he gains from the effect of more skilled bargaining, but loses from the effect of a less plausible threat to incur the expenses of trial. Links between $\beta$, $C_d$, and $C_p$, however, would not affect the analysis in this paper. } This model makes more precise what is meant by ``bargaining power''. I have distinguished between the parameters which determine the threat point and bargaining region, on the one hand, and the single parameter $\beta$ which determines where in the bargaining region the outcome will occur. This is useful because it shows that there is scope for mutually beneficial bargaining even if the situation is desperate for the defendant. If he is certain to be convicted of a serious crime (large $T$), frantically wishes to avoid trial (large $C_d$), and has no information to trade ($Z_p=0$), he still has bargaining leverage. This is because the prosecutor still wishes to avoid trial, because $C_p>0$, and would be willing to make some concessions. It is meaningful to say that the defendant could have a high degree of bargaining power ($\beta$ near zero), because that has a precise meaning that matches the situation: the defendant, although he begins in a very bad situation, can reap most of the gains from bargaining, and reduce high sentence by almost $C_p$, even though the resulting plea bargain sentence would still be long. Bargaining power concerns how much of the surplus each party ends up with, not with whether their starting points are strong or weak.\footnote{Whether a bargainer's starting point is strong or weak might, of course, depend on the outcome of previous bargaining, but that is not the case in the present context. If the defendant faces a long sentence at trial, it is because of the facts of the case and the state of the law, not because of his personal bargaining effectiveness.} In this model, plea bargaining is always successful. Each side would prefer to settle, and so settlement occurs. This is much simpler than asymmetric information models such as Reinganum (1988), but it helps establish a meaning for the concept of bargaining power. If the defendant agrees to a plea bargain, it must benefit him, or he would not agree. Bargaining power concerns only {\it how much} it benefits him. That is still a matter of great concern, but it would be very misguided for the law to eliminate plea bargains on the grounds that defendants gain less from them than prosecutors do. Even if that were true, which is difficult to establish, banning a practice that benefits defendants just because it does not benefit them enough would not be doing them a favor. This point will be important next, when we discuss waivers. \bigskip \noindent {\it 5.2 Information Disclosure During Plea Bargaining} Now let us add the complication that some information is disclosed during plea negotiations without being bargained for. This might occur because each side is disclosing information to try to show the strength of its case, or by accident. If the plea bargain breaks down, this disclosure, if admissible as evidence or useful as background information, could be used to the disclosing side's disadvantage when the case comes to trial. Looking ahead, each side must decide before entering into plea bargaining whether the ultimate outcome will be to its advantage. It is important, first, to realize that the government does incur costs from entering into plea bargaining, particularly in cooperation bargaining, but also in penalty bargaining. For many defendants, the government's expectation of the value of their cooperation is so small that it is not even worth the cost of a meeting to explore the subject. As the government brief in {\it Mezzanatto} puts it, \begin{quotation} \vspace*{-6pt}\begin{small} In addition, a prosecutor's decision to meet with a defendant and his counsel for the time necessary to evaluate proffered cooperation ordinarily entails a significant commitment of prosecutorial resources, not only for the prosecutor himself, but also for investigative agents who must be present during the interview, prison officials who must make the defendant available when he is incarcerated, and foreign language interpreters who must be present when the defendant is not fluent in English. Meetings with a defendant and his counsel to assess proffered cooperation frequently will occupy a day in the schedule of the prosecutor and the law enforcement agents involved, and still more time to check the defendant's story against other leads. Because a prosecutor cannot allocate those resources to every defendant who expresses a possible desire to cooperate with the government, the prosecutor must choose those prospects that are most likely willing to furnish truthful information.\footnote{{\it Mezzanatto} U.S. Brief, {\it supra} note \ref{usbrief}. } \vspace*{-12pt}\end{small} \end{quotation} Besides the expenses in terms of time and resources, the government may wittingly or unwittingly reveal useful information to the defendant in the course of plea negotiations, just as the defendant may reveal it to the government, with the difference that Rule 410 does not protect the government from its disclosures. This is especially important because of the limited amount of discovery available in criminal cases.\footnote{\label{stuntz} ``The Fifth Amendment has been thought to limit most discovery directed against the defendant, and considerations of parity, together with a mix of other concerns, long led most states to deny virtually all discovery directed against the prosecution.'' Sanford Kadish, Stephen Schulhofer \& Monrad Paulsen, CRIMINAL LAW AND ITS PROCESSES 143 ( 4th Ed. 1983). Robert Scott \& William Stuntz, Plea Bargaining as Contract, 101 Yale LJ 1909, 1937 (1992) note that the police report is not discoverable by the defendant and ``... at the time of settlement negotiations, a criminal defendant has much less knowledge of the government's case than, say, a civil defendant has of a civil plaintiff's case against him.'' } The government brief in {\it Mezzanatto} complained of this: \begin{quotation} \vspace*{-6pt}\begin{small} Under the court's holding, a defendant will be free to use his proffer of cooperation merely as an opportunity to ``try out'' a story on the prosecutor. That tactic will permit the defendant to gain the best of all worlds: favorable treatment if the prosecutor is persuaded, and an opportunity to swear to a different version of the events at his trial (without fear of contradiction and aided by any facts he learns during the proffer session) if the prosecutor is not persuaded.\footnote{{\it Mezzanatto} U.S. Brief, {\it supra} note \ref{usbrief}. } \vspace*{-12pt}\end{small} \end{quotation} The government perhaps felt this especially keenly because Mezzanatto did exactly that, having discovered in the plea bargaining interview that the government had information which disproved his initial story. \begin{quotation} \vspace*{-6pt}\begin{small} That is exactly what respondent attempted to do in this case. After failing to persuade the prosecutor, respondent not only testified to a completely different version of his involvement in methamphetamine trafficking, but he also took care to admit that he had been present on Shuster's property on the day before his arrest -- and thereby attempted to neutralize the surveillance evidence about which he learned during the meeting. Indeed, defense counsel argued in summation that the jury should believe respondent's story because that story was confirmed by the government's own surveillance evidence. \footnote{{\it Mezzanatto} U.S. Brief, {\it supra} note \ref{usbrief} at note 9.} \vspace*{-12pt}\end{small} \end{quotation} The government's statements to the defendants can even be used to impeach government witnesses, an additional risk. No article on criminal procedure in the late 1990's would be complete without a reference to the O.J. Simpson trial, and one is actually relevant here. Detective Philip Vanatter talked to government informers Craig and Larry Fiato about going to Simpson's house after Mrs. Simpson was found murdered, and they took the stand for Simpson to impeach Vanatter's testimony. Talking to criminals is a risky business.\footnote{Trott (1996), {\it supra} note \ref{trott} at 1396.} Thus, engaging in plea negotiations is potentially costly to the prosecutor as well as to the defendant. Let us denote the expected decline in the penalty due to prosecutor disclosures during plea negotiations by $L_p$, and the expected increase due to defendant disclosures by $L_d$. Under Rule 410, the defendant's disclosure is inadmissible, so $L_d=0$. Under a waiver of Rule 410, the defendant's disclosure is admissible, so $L_d$ is greater than zero.\footnote{For game theory aficionados: I am using a reduced-form model of asymmetric information here, rather than the usual signalling game of incomplete information. You should not think of the players as updating Bayesian priors when information is revealed. Rather, this is a game in which a move called ``information disclosure'' is assumed to be required for plea bargaining but also changes the payoffs from trial. The model is useful for looking at the decision to enter into plea bargaining, but takes as given that information must be revealed in the plea bargaining process. More generally, $L_d$ and $L_p$ could be viewed as including the transaction costs of bargaining, though in that case $L_d=0$ would not be true for the defendant even under Rule 410 unless he is represented by the public defender and does not pay any cost from the time spent bargaining.} It will be most convenient to analyze the outcome under waiver of Rule 410 first, and then analyze the outcome with Rule 410. \bigskip \noindent {5.3 The Outcome Under Waiver of Rule 410} If both side's disclosures are admissible, the expected penalty rises by $(L_d-L_p)$, which might be a negative number. The changes the settlement under the bargaining process modelled earlier to \begin{equation} \label{e5} T +L_d -L_p -C_p + \beta( C_d+C_p). \end{equation} A settlement of amount (\ref{e5}) might result in lower payoffs to one party or the other than his expected payoff from going to trial. That party would then refuse to enter into plea bargaining. Successful plea bargaining will be blocked by the prosecutor if \begin{equation} \label{e6} (T +L_d -L_p -C_p + \beta( C_d+C_p)) < T - C_p, \end{equation} because he will do better by going to trial, given how much he weakens his bargaining position by his disclosures during plea bargaining. Similarly, the defendant will refuse to enter into plea bargaining if \begin{equation} \label{e7} T +L_d -L_p -C_p + \beta( C_d+C_p) > T + C_d . \end{equation} Rearranging the inequalities, the prosecutor will block plea bargaining if \begin{equation} \label{e9} \beta(C_d+C_p) < L_p - L_d \end{equation} and the defendant will block it if \begin{equation} \label{e10} -(1- \beta) (C_d+C_p) > L_p - L_d . \end{equation} We can conclude that if either $L_p$ or $L_d$ is too large relative to the other, plea bargaining will break down. The exchange of useful information must be roughly equal for both parties to be willing to reveal their information. Figure 2 shows this graphically. Plea bargaining succeeds in areas B and C, where the disclosure losses of the defendant and prosecutor are of comparable size. In area A, bargaining breaks down because the prosecutor fears that the defense will learn too much during the plea bargaining, or his costs of setting up plea discussions are too high. In areas D and E, it breaks down because the defendant fears the prosecutor will learn too much. Each case will have its individual parameters and lie in one of these five regions. Which is the most common kind of case? This author cannot claim expertise in this area, but I hope that this framework may be helpful to those that can. The government claimed in {\it Mezzanatto} that the government did not wish to devote the resources to talk to many defendants, so that area E held many cases, but whether that is true or not is beyond the scope of this paper. \epsfysize=3in \epsffile{mezz3.eps} \begin{center} {\bf Figure 2: Conditions Under Which a Bargain is Made } \end{center} \bigskip \noindent {5.4 The Outcome Under Rule 410} If Rule 410 prevents the prosecutor from making use of the defendant's statements, then $L_d$ drops from the payoffs. The settlement amount becomes \begin{equation} \label{e11} T -L_p -C_p + \beta( C_d+C_p) . \end{equation} Again, our concern is whether the settlement will lie within the range $ [T - C_p,T + C_d]$. Successful plea bargaining will be blocked by the prosecutor if \begin{equation} \label{e12} \beta(C_d+C_p) < L_p \end{equation} and by the defendant if \begin{equation} \label{e13} C_d + \beta(C_d+C_p) > - L_d. \end{equation} This means that the defendant will never block plea bargaining, but the prosecutor is more likely to than when Rule 410 is waived. In Figure 2, plea bargaining succeeds in areas C and D, but fails in areas A, B and E. In areas C and D, the prosecutor's loss from disclosure during plea bargaining is low, and under Rule 410, the defendant's loss is irrelevant. In areas A, B and E, the prosecutor's loss is high enough that he blocks disclosure . The problem under Rule 410 is that the prosecutor fears that his trial position will become worse relative to the defendant's as a result of unsuccessful plea bargaining. If he were to agree to enter plea bargaining negotiations and were to lay out his case as persuasively as he could, the defendant would listen attentively, but then threaten to break off plea bargaining and use this rehearsal to prepare for trial. The defendant might also have disclosed information, but Rule 410 would make the disclosures inadmissible at trial. As a result of this threat, the prosecutor would be forced to agree to a generous sentence reduction to avoid the cost of trial. In the end, he would wish he had never tried plea bargaining in the first place. \noindent {5.5 Bargaining over the Waiver} It is of course not necessary to limit ourselves to pure regimes of Rule 410 or no Rule 410. What if the prosecutor has the option to ask for a waiver of Rule 410 and the defendant has the option to refuse, on a case-by-case basis?\footnote{The exclusion waiver converts from no-admissibility to admissibility. To go the other way, starting from a regime in which Rule 410 did not exist, the prosecutor would have to agree to a special exclusion provision. This, in effect, is what a grant of limited or full immunity from prosecution does. Immunity is a special waiver by the prosecutor of his right to prosecute the defendant or to make use of his disclosures to prosecute him. } Bargaining over whether to have an exclusion waiver can be modelled as a preliminary move. It is different from bargaining over the penalty in that the good being bargained over cannot be split-- there is either a waiver or there is not. If a plea bargaining would be successful even without the waiver, then the prosecutor gains and the defendant loses from the waiver. This happens in area C of Figure 2, where the disclosure losses are small. The effect of the waiver is then to increase the settled penalty from $T +L_d -L_p -C_p + \beta( C_d+C_p)$ to $T -L_p -C_p + \beta( C_d+C_p)$, an increase of $L_d$. Why would the defendant ever agree to a waiver if it ends up hurting him? He would agree if he still benefits from successful settlement overall instead of going to trial, and if the prosecutor can credibly threaten to break off plea bargaining if the waiver is not signed. Whether the prosecutor can credibly threaten this depends on local circumstances, but if he can, then the defendant's concession of signing a waiver acts as an admission ticket to plea bargaining negotiations. The admission ticket has a price, but a low enough price that the defendant is willing to pay it.\footnote{This is an example of how splitting up negotiations into sequential parts can be a way for one side to leverage up its bargaining power. If indivisible parts can be separated out, and the better bargainer wins on indivisible issues, then instead of winning, say sixty percent of the indivisible issues when they are bargained over as a group, the better bargainer can win all of them sequentially. This example is particularly interesting because the waiver issue cannot be bargained over at the same time as the penalty, involving, as it does, the issue of whether to exclude the information revealed during the penalty bargaining. } If plea bargaining would be unsuccessful without the waiver and successful with it, then both prosecutor and defendant gain from the waiver, because both prosecutor and defendant payoffs are bigger from successful settlement than from trial. This happens in area B in Figure 2, which applies when \begin{equation} \label{e14} L_p - L_d> \beta(C_d+C_p) > L_p, \end{equation} that is to say, when the disclosure losses are relatively large, but are roughly equal for the prosecutor and the defendant. If the defendant would not agree to a waiver, under the parameters in area D, the prosecutor would not ask for one. If he did, and was turned down, then he would lose his share of the bargaining surplus. Thus, we would expect waivers not to be requested if the defendant has sufficiently more to lose from information disclosure in plea bargaining than the prosecutor does. It may also happen that plea bargaining is unsuccessful either with or without the waiver, in which case neither side gains or loses from the possibility of waiver. This happens in areas A and E in Figure 2. From the point of view of society, it is in one respect difficult and in another respect easy to tell whether waivers are helpful. What is difficult to know is the effect on the penalties imposed. Presumably $T$ is the optimal settlement, since it depends on the expected judgement of the law and not on the costs and bargaining strengths of the prosecutor and defendant. We cannot say without knowing the costs and bargaining strengths whether waivers move the settled penalty closer to $T$ or further from it. In another respect it is easy to see a way in which waivers are helpful: they increase the probability of settlement, to the benefit of both sides. Thus, if the aim of Congress is to increase the amount of settlement, {\it Mezzanatto} , allowing voluntary waiver of Rule 410 by the defendant would seem to be helpful. \noindent {5.6 The Strategic Advantages of the Prosecutor} The analysis above provides reasons why waivers would help both sides, by making the prosecutor willing to enter into plea bargaining. In some circumstances, the prosecutor's motive for requiring the waiver was purely strategic, to take better advantage of his superior bargaining power, but that did not reduce the amount of successful bargaining-- it only shifted the terms in the prosecutor's favor. We now come to a strategic move which might actually hurt efficiency, because the prosecutor is willing to reduce the amount of successful bargaining if he can improve the terms enough. An important difference between prosecutor and defendant is that the prosecutor represents an office which is involved in many cases, not just one. The office might benefit from a tough policy in the long run, even if this imposed a cost in the short run. This may itself be a reason for the prosecutor to have more bargaining power ($\beta >.5$), because he has a reputation for toughness to preserve. A second effect, however, is that the prosecutor's office can rely on inflexible bureaucratic rules to commit itself to tough policies. We have seen above that if the prosecutor has most of the bargaining power ($\beta >.5$), then his best option is to ask for waivers on a case-by-case basis, asking only when he expects the defendants to grant them. In this respect, bargaining over the waiver is no different from plea bargaining generally; even someone in a strong bargaining position will not willfully overreach and ask for more than could possibly be granted. If, however, the prosecutor has a weak bargaining position in each individual case ($\beta <.5$), then he might take advantage of bureaucratic inflexibility to set a general rule of requiring waivers. If that happens, the prosecutor could benefit overall, from additional agreement in area B (where the prosecutor's disclosure losses are higher), and better bargains in area C (where the losses of both sides are low), even though he loses from having fewer agreements in area D (where the defendant's disclosure losses are high). Use of bureaucratic rules of this kind is analogous to use of standard-form contracts. Alan Schwartz and Louis L. Wilde have shown that a monopolist would rather use his market power to increase price than to introduce inefficient contract terms.\footnote{ Alan Schwartz \& Louis L. Wilde, Imperfect Information in Markets for Contract Terms: The Examples of Warranties and Security Interests, 69 VA. L. REV. 1387 (1983).} Robert Scott and William Stuntz mention that argument in relation to plea bargaining and note that in any case, each plea bargain is different, rather than being like a standard form, so that the arguments against contracts of adhesion do not apply againt plea bargaining.\footnote{Scott \& Stuntz (1992), {\it supra} note \ref{stuntz}. } The waiver agreement, however, is a special case. It can literally be a standard form, not tailored to the particular case. And although a strong prosecutor would prefer to use waiver agreements, case-by-case, a weak prosecutor would prefer to use a blanket rule, so as to commit to a tough position. Thus, it may be that allowing waivers would reduce the amount of successful plea bargaining, but this depends on hard-to-measure parameters. It is not enough that the prosecutor benefits from waivers at the expense of the defendant-- it must be that he benefits enough to make up for increased failure in plea bargaining. If the defendant has more to lose from disclosures during plea bargaining than does the prosecutor, so that area D is important, it seems unlikely that the prosecutor would want to commit to a policy of requiring waivers. Whatever advantage he gained in better bargains in area C would be outweighed by the losses from failed bargaining in area D. Moreover, the better bargains in area C would be large only if $L_d$ is large, but that is precisely the situation in which not area C, but area D is appropriate. \bigskip \noindent { 6. Conclusions } {\it Mezzanatto} is a particularly interesting case for analysis because it requires so many of the tools of economic analysis without seeming to have any economics in it at all. Yet the basic situation is a market transaction, the trade of one good for another. As a result, the insight of price theory that trade benefits both parties is relevant, but we cannot stop with the theory of Adam Smith. Plea bargaining is not a situation of anonymous market competition for a good of known quality, but the negotiation of a relational contract, difficult to enforce and for a good of unknown quality, where two parties are in a bilateral monopoly and each seek to move the terms of transaction to his own advantage. These strategic considerations make the newer insights of game theory valuable. Indeed, it is the strategic complexities which result in the opposite conclusions of different judges who have heard the case. On the one hand are judges like Judge Wallace and Justice Thomas who view the situation from the laissez faire position that increasing the flexibility of the terms of trade must benefit both sides of the transaction. On the other are judges like Judge Sneed and Justice Souter who are deeply suspicious of such flexibility and think that it will result in less and more one-sided trade, because it will be used by prosecutors to increase their gains from bargaining. My conclusion is that the judges with laissez faire instincts have come to the right conclusion with respect to allowing waivers of Rule 410, but that a full analysis requires quite a bit more complexity to deal with the legitimate fears to which strategic bargaining gives rise, and due consideration to the role Rule 410 waivers fulfill in the world of practical criminal prosecution. Proffer letters commonly {\it are} one-sided, but this is not to the disadvantage of defendants. Rather, they are in the desperate postion of having something to sell but without having the ability to guarantee delivery or the quality of the product. Placing themselves at the mercy of the prosecutor provides this guarantee, and allows them to sell their product. The system only works, however, because prosecutors wish to hold up their end of the bargain to preserve their reputations for future negotiations The frequency of Rule 410 waivers in cases where cooperation bargaining takes place and the infrequency where pure penalty bargaining takes place lends support to this theory. Rule 410 is waived not because the prosecutor has the greater bargaining power, but because in certain kinds of cases the defendant needs to establish his willingness to provide cooperation and to show that he is truthful enough that that cooperation will be useful to the prosecutor. Whether or not the defendant has a good bargaining position will determine how much he gets for his cooperation, but even a defendant with a strong bargaining position would want to establish his credibility by use of a Rule 410 waiver, in order to maximize the concessions he can extract from the prosecutor. This theory suggests that not only waivers of Rule 410 for puposes of impeachment but for use in the case-in-chief should be valid, but also suggests why we do not observe the broader waivers in practice. Justice Souter's dissent is correct that prosecutors will charge all that traffic will bear in the plea bargaining market, but wrong in his estimate of how much traffic will bear. In the economic marketplace, we allow bakers to charge all that traffic will bear for bread, but the consequence is not mass starvation. Similarly, it seems that usually defendants will agree to a proffer agreement waiving Rule 410 for impeachment, but not for the case-in- chief. On occasion, we do see such agreements, of ten in exchange for complete immunity if the defendant cooperates fully, but they are not the norm, and when they do occur, it can be justified as a means for the defendant to guarantee the quality of a special informational product he is selling. %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \end{document}