(This syllabus is not part of the opinion of the Court. It has been prepared by the Office of the Clerk for the convenience of the reader. It has been neither reviewed nor approved by the Supreme Court. Please note that, in the interests of brevity, portions of any opinion may not have been summarized).

    CLIFFORD, J., writing for a majority of the Court.

    A jury convicted John Martini of murder and other crimes for the kidnapping and murder of Irving Flax. At the penalty phase, the jury found the aggravating factors (escaping detection and murder during a kidnapping) to outweigh the mitigating factors (defendant's age and the "catch-all" factor). The Supreme Court affirmed Martini's conviction and the sentence of death on February 9, 1993, deferring proportionality review. This decision addresses the issue whether the imposition of the death sentence on Martini was disproportional.

HELD:    Martini's sentence, considering both the defendant and the crime, does not disclose any disproportionality when compared to the sentences in factually-similar cases. The Court therefore affirms Martini's death sentence.

1.    The goals of proportionality review are to ensure the existence of a substantial distinction between capitally-sentenced and life-sentenced defendants; to limit death sentences to cases involving the most aggravated of murders; and to promote consistent and rational application of the death sentence. The burden is on defendant to show the presence of disproportionality or aberration in his sentence. (pp. 9-14)

2.    The Court's first step is to determine the universe of cases to which it will compare defendant's case. It does not apply the current version of N.J.S.A. 2C:11-3e, which limits the universe of comparable cases to only those in which a death sentence has been imposed, because Martini murdered Flax before the effective date of the amendment. The Court therefore does not rule on constitutional challenges to the amendment. The universe of cases thus consists of 298 death-eligible cases identified by the Administrative Office of the Courts. (pp. 14-15)

3.    Martini's suggestion to exclude reversed death sentences from the universe of death-sentenced cases is rejected, just as this suggestion was rejected in Bey IV. Death sentences reversed for procedural error, as opposed to errors affecting the substance of the crimes, remain reliable indicators of deathworthiness. Defendant's case is also included in the pool because doing so aids the search for community values used to determine the presence of disproportionality in death sentencing. (pp. 15-20)

4.    The universe of cases must then be grouped according to their comparative levels of blameworthiness by considering statutory aggravating and mitigating factors, as well as non-statutory factors, that are based on objectively-verified measures of blameworthiness. The Court uses two approaches to evaluate these factors: frequency analysis, in which the sentences in death-eligible cases are compared on a statistical basis to determine whether a case such as Martini's will generally result in a death sentence, and a precedent-seeking review, in which the Court engages in a traditional, case-by-case comparison of death-eligible cases. Because the size of the sample pool used in the frequency approach remains small, the results of frequency analysis are not entirely reliable, and the Court therefore relies more heavily on the results of the precedent-seeking review. (pp. 20-23)

5.    Frequency analysis is separated into three parts: the salient-factors test, in which the Court compares defendant's case to factually-similar cases; the numerical-preponderance-of-aggravating-and-mitigating-factors test, which measures defendant's case against others involving the same number of

aggravating and mitigating factors; and the index-of-outcomes test, which compares defendant's case to cases involving the same level of blameworthiness, considering both statutory and non-statutory factors. The results of these tests indicate no disproportionality in Martini's sentence. (pp. 23-46)

6.    In the precedent-seeking approach, the Court considers the objective statutory and nonstatutory factors that were presented to the juries in the cases used in the frequency analysis. It is a search for impermissible factors in the decision to sentence a given defendant to death to ensure that the defendant has not been singled out for capital punishment unfairly. Precedent-seeking review reveals no disproportionality in Martini's sentence. (pp. 47-97)

7.    Finally, Martini argues that the death penalty violates the Federal and New Jersey Constitution, is racially biased, and is applied inconsistently on a geographical basis. The Court rejected those arguments in prior opinions, and rejects them here for the reasons expressed in those opinions. (pp. 97-99).

     The imposition of the death penalty on defendant is not disproportionate.

    JUSTICE HANDLER, dissenting, is of the view that the looseness of proportionality review demands the continued exposure of the fundamental flaws that beset our current system of death-sentence validation. One of the most prominent defects in the Court's proportionality review is the continued use of reversed-death sentences in the universe of comparable cases. Another major deficiency is the inconsistency and inherent subjectivity of the Court's techniques for determining proportionality -- frequency analysis and precedent-seeking analysis. The Court persists in applying those methods without any clear standards or intelligible guidelines. Beyond the problems peculiar to proportionality review, there remains the great systemic flaw of our capital punishment regime, a regime founded on conflicting and contradictory principles and administered without any degree of consistency.

     CHIEF JUSTICE WILENTZ and JUSTICES POLLOCK, O'HERN, GARIBALDI and STEIN join in JUSTICE CLIFFORD's opinion. JUSTICE HANDLER has filed a separate dissenting opinion.
SUPREME COURT OF NEW JERSEY
A- 16 September Term 1993

STATE OF NEW JERSEY,

    Plaintiff-Respondent,

        v.

JOHN MARTINI, SR.,

    Defendant-Appellant.

        Argued January 19, 1994 -- Decided December 21, 1994

On proportionality review of a death sentence imposed in the Superior Court, Law Division, Bergen County.

Mark H. Friedman and William B. Smith, Assistant Deputy Public Defenders, argued the cause for appellant (Zulima V. Farber, Public Defender, attorney).

Craig V. Zwillman, Deputy Attorney General, argued the cause for respondent (Deborah T. Poritz, Attorney General of New Jersey, attorney).

        The opinion of the Court was delivered by
CLIFFORD, J.

    Defendant, John Martini, kidnapped Irving Flax, a Fair Lawn businessman, and held him for $25,000 ransom. After receiving the ransom money, defendant killed Flax by shooting him three times in the back of the head at extremely close range. A jury convicted defendant of purposeful or knowing murder by his own conduct, felony murder, kidnapping, and two weapons offenses. After the penalty-phase proceedings the trial court sentenced defendant to death for the purposeful or knowing murder and to a life term with a thirty-year parole bar for the felony murder, and merged those sentences. Defendant also received a consecutive life term with a twenty-five-year period of parole ineligibility for the kidnapping, and two concurrent four-year terms for the weapons offenses. This Court affirmed defendant's convictions, except for possession of a handgun without a permit, and his sentences, except for that imposed for kidnapping. State v. Martini, 131 N.J. 176 (1993). We granted defendant's request for proportionality review of his death sentence, see N.J.S.A. 2C:11-3e, and now find no disproportionality.

I    Facts3

II    Proportionality Review9

    A. The Universe of Cases14
    B. Method of Classifying Cases15

III    Comparison of Cases20

    A. The Frequency Approach23
        1. The Salient-Factors Test27
             2. The Numerical-Preponderance-of-Aggravating-and-Mitigating-Factors Test35
        3. The Index-of-Outcomes Test39
        4. Frequency-Approach Conclusion46

    B. The Precedent-Seeking Approach47

        1. Relevant Factors50
             2. Application of Precedent-Seeking
Approach54

            a. Parties' Arguments55

            b. Summaries of Similar Cases58

                i    Non-Stranger Kidnapping59
                ii    Stranger Kidnapping65
                iii    Contract-Murder Principals66
                iv    Contract Killers75
                     v    Other Non-Robbery Pecuniary-Advantage Killers86

            c. Analysis of Defendant's Culpability89
                 d. Comparison of Similar Cases to Defendant's Case93

        3. Other Cases96

IV    Other Arguments97

V    Conclusion99

    The facts are set forth in detail in Martini, supra, 131 N.J. at 191-207. We repeat here only those facts that are relevant to our proportionality review.

    In November 1988, Martini returned to New Jersey from Arizona. His girlfriend, Therese Afdahl, for whom defendant had left his wife of thirty-nine years, accompanied him. Using a credit card borrowed from a friend, defendant rented an apartment in Fairview under the friend's name. Lacking money, defendant sought from another friend, John Doorhy, advice on a quick method of replenishing his dwindling supply of funds. Doorhy, having recently worked at Flax's home and having noticed large amounts of cash and several bankbooks in the house, suggested that

Martini kidnap Flax, and acquainted defendant with the Flax family's morning schedule. In exchange for driving Martini to Flax's house and giving him written directions that defendant would later use to return there, Doorhy accepted defendant's promise of a percentage of the money that defendant anticipated receiving from the kidnapping.

    In preparation for the kidnapping, defendant retrieved from Doorhy's house a revolver that Doorhy had been holding for him, and purchased another revolver in Jersey City. On January 23, 1989, Martini and Afdahl drove to Flax's house. When Flax came out of the house, defendant alighted from his car and called Flax by a nickname that he knew Flax had formerly used, having been acquainted with him some thirty years previously. Flax asked whether defendant had been in the Army. Defendant lied, saying "yes," and suggested that they go in Flax's car to a diner for a cup of coffee.

    Flax agreed. After the two men entered the car, Martini pulled out his recently-purchased revolver, told Flax that he was being kidnapped, and directed him to drive to a Garden State Plaza parking lot in Paramus. Afdahl followed them. After both cars had reached the lot, defendant ordered Flax into defendant's car and drove to the Fairview apartment.

    Martini made Flax place a call to his wife, then bound Flax and told Mrs. Flax that if she wanted to see her husband alive, she would have to give defendant $100,000. Defendant also threatened to kill both her husband and her if she notified the police. Defendant called again at 1 p.m. to see if Mrs. Flax had raised the ransom money. When she said that she could not obtain that much cash, Martini he said that he would call back at 6 p.m. to see if she could raise $25,000. Throughout the call defendant repeatedly threatened to kill both the Flaxes.

    During the afternoon, the police placed taps on Mrs. Flax's telephone. After Mrs. Flax withdrew the $25,000, F.B.I. agents recorded the serial numbers of the bills. At 5:30 p.m., defendant called again, arranged the delivery of the ransom money, and again threatened that someone would come to kill the Flaxes if defendant were arrested. The F.B.I. recorded the conversation. Shortly thereafter, Mrs. Flax received a call from her hysterical husband, begging her to give defendant the money.

    As arranged, Mrs. Flax dropped off the money and Martini picked it up. F.B.I. agents followed him, but defendant, fearful of being followed, drove into the Bronx, managing to lose the agents during the course of an hour's drive in traffic. He returned to the Fairview apartment and retrieved Afdahl and the victim, whom defendant ordered to drive to the Garden State Plaza parking lot, where defendant's car was parked. When they

arrived, defendant shot Flax three times in the back of the head, because, defendant claimed, Flax had opened the driver's door and placed his foot on the ground, and defendant feared that Flax would escape.

    Leaving Flax's body in the car, Martini drove his own car onto the Staten Island Ferry, from which he threw both his gun and his victim's car keys into the New York Harbor. He then drove to the Bronx with Afdahl, disposed of the car, and arranged for a ride back to Fairview from the friend whose credit card he had been using.

    The next day, January 24, 1989, a security guard discovered Flax's body in his car at the Garden State Plaza parking lot. That afternoon, an acquaintance of Martini identified the male voice on the taped telephone conversation as Martini's.

    Alerted by a flyer that defendant and Afdahl were wanted in connection with a double homicide in Arizona, police in Fort Lee saw the two leave a motel and, carrying a black bag, walk to a telephone booth at a gas station, where defendant placed a call. When a taxi arrived, defendant and Afdahl entered it, whereupon police arrested them. A search of the bag revealed $23,760 bearing serial numbers that the F.B.I. had recorded, the borrowed credit card that defendant had been using, the second revolver,

and a key for the motel that they had just left. Police did not observe in defendant any signs of drug-related intoxication.

    After being arrested, receiving Miranda warnings, and being allowed to consult with Afdahl, whom he advised to cooperate with the authorities as he intended to do, defendant gave the police written and oral statements and his consent to search his motel rooms and his rented apartment.

    At trial, the State's forensic expert testified that the pattern of blood spattering and other physical evidence indicated that Flax had been shot at a range so close that the victim could not have opened the door and placed his foot on the ground as defendant had claimed. A physician with whom Martini had consulted on December 12, 1988, testified that he had observed no evidence of cocaine use. Mrs. Flax testified without objection to the telephone conversations with both her husband and defendant. Finally, a police officer who had taken statements from defendant read them into the record.

    Defendant presented evidence of his cocaine habit, which, in addition to his ten-year affair with Afdahl, a former prostitute, had broken up his marriage, and, defendant claimed, had diminished his capacity to commit his crimes purposefully or knowingly. His defense expert testified that Martini was "unquestionably" under the influence of cocaine, but the witness

could not give an opinion within a reasonable degree of medical probability whether defendant had acted purposefully or knowingly during the shooting. The State's rebuttal expert testified that defendant had, within a reasonable degree of medical probability, acted purposefully or knowingly during the shooting.

    The jury found defendant guilty on all counts. At the penalty trial for the purposeful or knowing murder, the State sought to prove two aggravating factors: c(4)(f) (murder to escape detection), and c(4)(g) (murder during the course of a kidnapping). Defendant offered five mitigating factors: c(5)(a) (extreme mental or emotional disturbance insufficient to constitute a defense to prosecution), c(5)(c) (age of defendant), c(5)(d) (diminished capacity due to mental disease or defect, or intoxication), c(5)(g) (furnishing substantial assistance to the State in prosecuting another person for murder), and c(5)(h) (the catchall factor). The trial court granted the State's motion to strike mitigating factor c(5)(g). Martini offered evidence of his drug abuse and its effects on his personality, and the State offered rebuttal evidence concerning defendant's character.

    The jury found both aggravating factors beyond a reasonable doubt. No juror found mitigating factors c(5)(a) (extreme mental or emotional disturbance) or c(5)(d) (diminished capacity). Six jurors found mitigating factor c(5)(c) (defendant's age), and six found c(5)(h) (the catchall factor). The jury unanimously found

that the aggravating factors, whether considered together or individually, outweighed the mitigating factors.

    From the judgments of conviction and sentences of death for the murder and prison terms for the other offenses, defendant appealed directly to this Court, R. 2:2-1(a)(3). We affirmed defendant's conviction for murder and the sentence of death, but reversed the conviction for possession of a handgun without a permit for failure to prove an essential element of the offense. 131 N.J. at 191, 320. We further vacated defendant's kidnapping sentence because it violated the statutory term, and remanded for resentencing on that count. Id. at 321.

    The purpose of proportionality review is to determine whether a specific defendant's death sentence is disproportionate. See N.J.S.A. 2C:11-3e. A capital sentence is excessive and thus disproportionate if other defendants with characteristics similar to those of the defendant under review generally receive sentences other than death for committing factually-similar crimes in the same jurisdiction. State v. Bey, 137 N.J. 334, 343 (1994) (Bey IV); State v. Marshall, 130 N.J. 109, 131 (1992) (Marshall II). We have declined to set a numerical standard to determine at what point defendants "generally" receive sentences of death, Marshall II, supra, 130 N.J. at 152-53, because such a standard would introduce unacceptable arbitrariness into proportionality review.

    Proportionality review is a procedural, or "offender-oriented," safeguard -- that is, it focuses on the defendant, not on the crime committed. Marshall II, supra, 130 N.J. at 126-27. In focusing on the offender and not the offense, that form of review presumes that death is proportional to the crime, Pulley v. Harris, 465 U.S. 37, 43, 104 S. Ct. 871, 876, 79 L. Ed.2d 29, 36 (1984), and is therefore different from the traditional Eighth Amendment substantive review. In contrast, substantive review considers the offense, not the offender, to determine whether the punishment is excessive for the crime itself. Coker v. Georgia, 433 U.S. 584, 592, 97 S. Ct. 2861, 2866, 53 L. Ed.2d 982, 989 (1977). For a sentence to be facially constitutional, the magnitude of the punishment must measurably serve the acceptable goals of punishment and may not be "grossly out of proportion" to the degree of harm. Gregg v. Georgia, 428 U.S. 153, 173, 96 S. Ct. 2909, 2925, 49 L. Ed.2d 859, 875 (1976) (finding that death sentence does not per se violate Eighth Amendment); accord Coker, supra, 433 U.S. at 592, 97 S. Ct. at 2866, 53 L. Ed. 2d at 989 (concluding that death sentence is grossly disproportional and excessive for crime of rape); Marshall II, supra, 130 N.J. at 129 (tracing U.S. Supreme Court's development of substantive review); see Enmund v. Florida, 458 U.S. 782, 797, 801, 102 S. Ct. 3368, 3376, 3378-79, 73 L. Ed.2d 1140, 1151, 1154 (1982) (finding that

Eighth Amendment prohibits capital sentence for defendant who aids and abets felony in course of which murder is committed by others but who does not himself kill, attempt to kill, or intend that killing or use of lethal force take place).

    Proportionality review is a response to Furman v. Georgia, 408 U.S. 238, 92 S. Ct. 2726, 33 L. Ed.2d 346 (1972), which reversed three death sentences, one for murder and two for rape. In Marshall II, supra, this Court cited Justice Stewart's concurring opinion to state that the decision whether to sentence a defendant capitally, if left to the unfettered discretion of a jury, violates the Eighth Amendment because the sentence can be "'wantonly and * * * freakishly imposed.'" 130 N.J. at 125 (quoting Furman, supra, 408 U.S. at 310, 92 S. Ct. at 2762-63, 33 L. Ed. 2d at 390 (Stewart, J. concurring)). Proportionality review, which allows this Court to monitor the results of jury discretion, is neither required by the Eighth Amendment nor the only way through which to make a capital-sentencing scheme constitutional. See Pulley, supra, 465 U.S. at 44-46, 104 S. Ct. at 876-77, 79 L. Ed. 2d at 36-37. However, it permits New Jersey's capital-sentencing scheme to comply with the dictates of Furman and with the Eighth Amendment, which prohibit arbitrary and inconsistent application of the death penalty. See id. at 44, 104 S. Ct. at 876, 70 L. Ed. 2d at 36.

    The goals of that kind of review are to ensure that a substantial distinction exists between capitally-sentenced and life-sentenced defendants; to limit capital sentencing to those cases that are most aggravated and in which death sentencing is the expected result; and to promote a rational, consistent, and fair application of the death sentence. Marshall II, supra, 130 N.J. at 131; David C. Baldus, Death Penalty Proportionality Review Project Final Report to the New Jersey Supreme Court 24-25 (Sept. 24, 1991) (hereinafter Final Report). The burden is on the defendant to show disproportionality by establishing that similar defendants who commit factually-similar crimes generally receive sentences other than death. Bey IV, supra, 137 N.J. at 343, 349. We impose that burden on the defendant, not on the State, because the statute, N.J.S.A. 2C:11-3(e), speaks in terms of proving disproportionality, not proportionality. Id. at 349.

    We believe that the dissent reveals a fundamental misperception of the role of proportionality review. That review is not a "system of death-sentence validation," post at __ (slip op. at 1), but is instead a vehicle to ensure that the penalty-phase jury's decision is not insupportable. That purpose stems from the mandate of the statutory language itself: "the Supreme Court * * * shall determine whether the sentence is disproportionate to the penalty imposed in similar cases, considering both the crime and the defendant." N.J.S.A. 2C:11-3e. Thus, our search is not for proportionality, but rather one

in which our goal is to determine whether the jury's decision to sentence a defendant to death is comparable to decisions reached in the appropriate capital cases in our universe of cases. The question is whether other defendants with similar characteristics generally receive sentences other than death. Marshall II, supra, 130 N.J. at 153.

    The dissent finds "a palpable bias in favor of the proportionality of a death sentence." Ibid. That bias, if present, does not stem from what our colleague describes as a "selective and convenient rationalization." Rather, our imposing on the defendant the burden of showing disproportionality stems from the statutory language itself, discussed above. It is settled law. We held as much in Bey IV, supra, 137 N.J. at 343, 349. The dissent unearths nothing new or "treacherous," post at __ (slip op. at 5), here. On the contrary, it simply attempts to rewrite established proportionality jurisprudence. Therefore, the statement that the "Court is determined to put the burden of proof on the defendant," post at __ (slip op. at 24), although accurate, is hardly a damning accusation.

    Likewise, the dissent's claims concerning first-impression murders are incorrect statements of the law. Being the first murderer in a given category of intentional homicide does not "apparently guarantee[] the proportionality of the sentence." Post at __ (slip op. at 25). Instead, that status requires only

that we weigh our comparison of such a defendant with other capital murderers in light of the differences in the respective categories. That approach guarantees sensitivity of process, not certainty of result.

    A. The Universe of Cases

    The first step in proportionality review is to determine the universe of cases to which we compare a defendant's case. In 1992, the Legislature amended N.J.S.A. 2C:11-3e to limit the comparison group to only those cases in which a death sentence has actually been imposed. L. 1992, c. 5, § 1. However, the Legislature did not state whether it intended the amendment to apply to pending appeals. Bey IV, supra, 137 N.J. at 343-44. In Marshall II and Bey IV, we declined to apply the amendment to those appeals for the reason, among others, that those defendants' appeals were pending before the Legislature enacted the amendment. The same situation pertains here: Martini was sentenced to death on December 12, 1990, and his appeal was pending before the effective date of the amendment. We therefore will not apply N.J.S.A. 2C:11-3e as amended to defendant's proportionality review.

    The Administrative Office of the Courts (AOC) is responsible for maintaining the data base of homicide cases used by this Court for proportionality review. It has developed its

statistics based on the procedure created by Professor David Baldus, the Special Master appointed by this Court to create a model for proportionality review, and on the modifications thereto that this Court outlined in Marshall II. The universe of cases that we employ in this case is compiled in the Martini Report prepared by the AOC. That report includes cases collected from 1983 to June 25, 1993. It contains 298 death-eligible cases, 125 of which went to penalty trial, a rate of forty-two percent. Martini Report tbl. 3. Of the 125 penalty-trial cases, thirty-eight resulted in a death sentence, a rate of thirty percent. Martini Report tbl. 2. The overall death-sentencing rate is thirteen percent (38/298). Martini Report tbl. 1.

    B. Method of Classifying Cases

    After establishing the universe of comparison cases, we must sort those cases in a data base. As in Marshall II and Bey IV, we use two approaches: an a priori determination and an empirical method. The a priori procedure requires us to analyze cases according to features that experience has shown influenced the decision whether to sentence capitally. Bey IV, supra, 137 N.J. at 345; Marshall II, supra, 130 N.J. at 141-42. In the empirical process we assemble life-sentenced and death-sentenced cases to identify the characteristics that determine the patterns of life sentencing versus capital sentencing. Marshall II,

supra, 130 N.J. at 142-44. That approach reveals which factors prosecutors and juries consider determinative. Bey IV, supra, 137 N.J. at 345.

    Any attempt to define in advance all characteristics of a murder that "capture the critical facts of [a] defendant's case," as the dissent urges us to do, post at __ (slip op. at 16), would fail to distinguish between individual defendants. Because each capital case is unique, defining all the important facts that a murder might include is impossible. However, that circumstance hardly amounts to a fatal flaw. Instead of turning disproportionality review into "a selective and convenient rationalization for proportionality," post at __ (slip op. at 15), dealing with each case on its own merits in a manner sensitive to its unique set of facts is the only way to give capital defendants the full review to which they are entitled, given the finality of the death sentence.

    Defendant makes a number of "alternative assumptions" that he seeks to persuade this Court to adopt in our review of his sentence. First, he argues that three cases, State v. Ramseur, 106 N.J. 123 (1987), State v. Lodato, 107 N.J. 141 (1987), and State v. Hunt, 115 N.J. 330 (1989), should be coded as life-sentenced cases because of errors at trial. Second, defendant would have us exclude nine other cases from the death-sentenced pool for "deliberative errors." Those cases are State v.

Biegenwald, 106 N.J. 13 (1987); State v. Kise (unreported); State v. Koedatich, 112 N.J. 225 (1988); State v. Zola, 112 N.J. 384 (1988); State v. Gerald, 113 N.J. 40 (1988); State v. Williams, 113 N.J. 393 (1988); State v. Coyle, 119 N.J. 194 (1990); State v. Johnson, 120 N.J. 263 (1990); and State v. Oglesby, 122 N.J. 522 (1991). Third, he excludes his own case from the pool of comparable cases.

    In Bey IV we rejected defendant Bey's argument that Ramseur, Biegenwald, and Coyle should be treated as life-sentenced cases. Those cases, involving death sentences that were reversed, either were not pursued after reversal as penalty-trial cases or resulted in life sentences. Bey argued that those cases therefore could not be considered reliable indicators of deathworthiness. Bey IV, supra, 137 N.J. at 345-46. The AOC coded those as death-sentenced cases, and we concluded that they reflected juror evaluations of deathworthiness. Id. at 347. We observed in Marshall, supra, 130 N.J. at 194 n.10, that penalty trials that result in death are most often reversed for burden-of-proof and Gerald issues. (A Gerald error is one in which a capital sentence followed a conviction for purposely or knowingly causing serious bodily injury that results in death, instead of for purposely or knowingly causing death. See Gerald, supra, 113 N.J. at 69.) Burden-of-proof and Gerald errors, which affect the procedural fairness of the trial, not the substance of the crime, "do not necessarily bear on the jury's determination of

deathworthiness." Bey IV, supra, 137 N.J. at 347; accord Marshall II, 130 N.J. at 169 n.5, 194 n.10. In our proportionality review of Bey's sentence we therefore continued to include Ramseur, Biegenwald, and Coyle in the category of death-sentenced cases. We do so here as well.

    The dissent repeats its criticism of the Court for not accepting the dissent's proposition that reversed death sentences may not be used in proportionality review as death-sentenced cases. That criticism is based on the notion that "the only objective indicator that can establish 'deathworthiness' is the imposition of a death sentence." Post at __ (slip op. at 6). For support, the dissent cites examples of cases involving improper jury instructions that required reversal of the death sentence. Post at __ (slip op. at 7-8). That leads our dissenting colleague to insist again that the Court adopt a "rebuttable presumption that reversed death sentences are invalid determinations of deathworthiness." Post at __ (slip op. at 10). As we stated in Marshall II, supra, 130 N.J. at 169 n.5, Bey IV, supra, 137 N.J. at 347, and again today, errors that do not affect the substance of the crime, as opposed to procedural fairness, do not necessarily affect a jury's determination of deathworthiness.

    We similarly reject defendant's "deliberative error" argument. Martini claims that errors such as (1) the failure to

instruct a jury that it must find beyond a reasonable doubt that the aggravating factors outweigh the mitigating factors before it can impose a death sentence, (2) an improper charge on aggravating factor c(4)(c) (torture), or (3) a charge that a jury must unanimously find a mitigating factor before it can weigh that factor so undermine the reliability of the capital sentence that when a life sentence is given on retrial, those cases may not be coded as death sentenced. Defendant argues that we have never squarely considered cases involving the "fundamental fairness" of such overturned death sentences. We continue to believe that the death sentences in cases overturned for procedural error are acceptably-reliable societal determinations of deathworthiness.

    We also note, as we did in Bey IV, supra, that the State's decision not to reprosecute a defendant capitally is not necessarily a reflection of that defendant's lack of deathworthiness. The State's decision may be based on unrelated issues such as the availability of witnesses or the amount of financial resources that the State is willing to commit to retrial. Accordingly, we reject defendant's "fundamental fairness" argument and continue to code Kise, Koedatich, Zola, Gerald, Williams, Johnson, and Oglesby, as death-sentenced cases. As in Bey IV, supra, we acknowledge that our data are not scientifically infallible, 137 N.J. at 348, because all coding decisions contain some degree of subjectivity. Marshall II,

supra, 130 N.J. at 120 ("[W]e recognize that a value judgment is built into every practical measurement."). However, we remain convinced that "even reversed death sentences are sufficiently valid indicators" of "the conscience of the community" to be used as death-sentenced cases. Bey IV, supra, 137 N.J. at 348.

    Finally, we reject defendant's assumption that he will not be included in the universe of cases comparable to his own. As we did in Bey IV, we will evaluate the data both including and excluding Martini. We acknowledge that including a defendant in his or her own proportionality review will increase the rate of death sentencing to an extent that is inversely proportional to the number of cases that are used in the analysis. However, proportionality review is a search for community values, and the case under review is a partial reflection of the those values that we seek to discover. Using two sets of data, one including defendant's case and one excluding it, will give us the broadest picture of societal standards while alerting us to the bias produced by including defendant's case.

COMPARISON OF CASES

    Having established, first, the universe of cases on which we shall rely, specifically, those contained in the Martini Report, and, second, the criteria for coding those cases as either death

sentenced or life sentenced, we must next group those cases according to their comparative levels of blameworthiness. Bey IV, supra, 137 N.J. at 350. In Marshall II and in Bey IV we determined blameworthiness by considering statutory aggravating and mitigating factors as well as "nonstatutory factors based on 'objectively verified measures of blameworthiness.'" Ibid. (quoting Marshall II, supra, 130 N.J. at 145).

    We evaluate those factors through two approaches: frequency analysis and precedent-seeking review. Our purpose in that evaluation, stated above, is to determine whether defendant's sentence is disproportionate in comparison to similar cases. Marshall II, supra, 130 N.J. at 148. As we declared in Bey IV, supra, "Proportionality review seeks to determine only whether a particular death sentence is aberrational, not whether it compares perfectly with other sentences." 137 N.J. at 352.

    The frequency analysis allows us to determine the rate of imposition of death sentences in similar cases. It is designed to reveal how jurors and prosecutors treat similar cases. In precedent-seeking review we compare the defendant's case to factually-similar cases to determine whether the defendant is deathworthy in light of similarly-situated defendants. We then compare the results of the two analyses to ensure that our proportionality review is reliable. However, the size of the sample pools used in the frequency approach remains small.

Therefore, because frequency-analysis results are not entirely reliable statistically, we rely more heavily on the results of the precedent-seeking review. See Bey IV, supra, 137 N.J. at 351.

    The dissent objects to the Court's use of a relatively-small number of cases in our search for disproportionality. In our dissenting colleague's view, the uniqueness of a case under review makes "the project of frequency review * * * a sham." Post at __ (slip op. at 22). We recognize the small sample size and the fact that defendant's case is unique in our universe of capital cases. As we have noted, those considerations preclude us from giving great weight to frequency analyses derived under those conditions. In contrast to the dissent, however, we do not see that frequency review, the model agreed on by the offices of both the Attorney General and the Public Defender, and approved of by both the Administrative Office of the Courts and this Court, "falters." Post at __ (slip op. at 22). Instead, we use the cases most similar or analogous as guides, discounting the guidance they provide by their degree of dissimilarity. Our reliance on frequency analysis will increase as our universe of cases grows. In the meantime, we will use frequency analysis as a tool in our review. The absence of more complete data necessarily forces us to place greater emphasis on precedent-seeking review.

    The dissent's discussion of defendant's low predicted frequencies, post at __ (slip op. at 16-19), overlooks their function and purpose. They are predictors only. They provide not answers but only guidance. A low predicted value, therefore, does not mean that we must automatically overturn a death sentence as disproportionate. Instead, it shows a danger of the presence of disproportionality, requiring us to scrutinize more carefully the other elements of review. See Marshall II, supra, 130 N.J. at 159.

    A. The Frequency Approach

    In Marshall II and Bey IV, we separated the frequency analysis into three parts: the salient-factors test, the numerical-preponderance-of-aggravating-and-mitigating-factors test, and the index-of-outcomes test. Bey IV, supra, 137 N.J. at 350-51, Marshall II, supra, 130 N.J. at 154. Each is a different statistical method of gauging a defendant's relative criminal culpability. Bey IV, supra, 137 N.J. at 351. The basic question in the frequency approach is whether the degree of blameworthiness in the instant case reasonably supports an expectation that such a case will generally result in a death sentence. We use that method to determine whether a defendant is in a category that makes him or her more likely than other types of killers to receive the death penalty.

    As indicated supra at ___ (slip op. at 9-10), we employ no set level to determine at what point death sentences may be considered as "generally" imposed for a given type of murder. However, the lower the frequency of death sentences in a class of murderers, the greater scrutiny we must bring to bear to determine whether any impermissible factor has had a role in determining the sentence. Likewise, the greater the frequency of death sentences in a class of cases, the more certain we are that a given death sentence is proportionate for any member of that class. Bey IV, supra, 137 N.J. at 351; Marshall II, supra, 130 N.J. at 153. Accordingly, we use the frequency approach not as a strict rule but as a measure of consistency in our capital-sentencing regime. Marshall II, supra, 130 N.J. at 153.

    As pointed out above, the rate of penalty trials in death-eligible cases and the rate of death sentencing for cases that advanced to a penalty trial are both fairly low -- forty-two percent (125/298) and thirty percent (38/125) respectively. Martini Report tbls. 2, 3. That yields a total death-sentencing rate of thirteen percent (38/298). However, we are looking for a potential aberration, not a perfect comparison to all other cases. As we stated in Bey IV, supra, "Not every statistical disparity establishes disproportionality." 137 N.J. at 352. Moreover, we must compare defendant's case to "similar cases, considering both the crime and the defendant," N.J.S.A. 2C:11-3e, before we can make any determinations about relative frequency.

    Finally, we heed the admonition of the AOC that at this time we should place little substantive reliance on the statistical models. The small number of cases, especially death-sentenced cases, combined with the large number of factors used as independent variables undermines the reliability of the results of the regression models used.

    Whatever concerns we recognize with frequency analysis, nonetheless, do not mean that sentences for unique murders must automatically be struck down and that only those of run-of-the-mill defendants may be upheld. A capital defendant is not entitled to a perfect universe of identical cases, but instead only the best that we can achieve. See generally Bey IV, supra, 137 N.J. at 352 (stating that proportionality review searches for aberration, not for perfect comparison), 362 (limiting search to identifying irrationality and arbitrariness).

    We pause here for a brief explanation of some technical terms. A regression analysis uses an algebraic model to represent a decision-making process by showing the influence of an independent variable on a dependent variable. Here, the decision-making process represented is the sentencing determination. The independent variable, which, once designated, does not change, represents a factor such as a prior murder or a contemporaneous sexual assault that is believed to influence the result of the decision-making. The dependent variable is

influenced by the presence or absence of an independent variable, and here represents the decision whether to sentence capitally. A multiple regression analysis simply includes more than one independent variable in the algebraic model.

    Because the results produced by the regression models are of uncertain reliability, we use the predicted probability of death sentences that those models generate only for purposes of comparison and guidance. We do not accord them final or determinative weight. See Memorandum from John P. McCarthy, Jr., Assistant Director, AOC, to Stephen W. Townsend, Clerk of the Supreme Court, State v. John Martini: Proportionality Review 3 (July 28, 1993) (on file with AOC).

    Nonetheless, a main area of disagreement between the Court and the dissent involves the presence of standards. Our dissenting colleague laments our refusal "to set any standard by which to distinguish a high from a low predicted frequency of death," post at __ (slip op. at 15), as well as the lack of "any established category that captures the critical facts of defendant's case." Post at __ (slip op. at 16).

    We avoid setting numerical standards to determine disproportionality because such an absolute, numerical system suffers from an inherent failure to distinguish between defendants. For example, if we set at thirty percent the

acceptability of a certain frequency criterion as indicating no disproportionality, we would then be bound to hold that the sentence of one defendant who, at thirty percent, met that characteristic was not disproportionate. Likewise, we would be bound to find disproportionality in the case of another defendant who achieved only twenty-nine percent in that category. Yet overall, the latter defendant might be more deathworthy under our statutory scheme, and of the two, might actually be the one properly sentenced.

    Even with the foregoing problems, however, the statistical approach receives our attention because it permits us to distinguish cases by culpability; because it allows us to determine a community consensus, in contrast to the individual assessment of the case-by-case approach; and because it creates a basis for evaluating the fairness of the entire sentencing system. Unlike the precedent-seeking approach, the statistical method provides a means for deciding whether the cases used for comparison are themselves disproportional. Therefore, we use both approaches as complementary techniques.

        1. The Salient-Factors Test

    The salient-factors test allows us to measure the relative frequency of a defendant's sentence by comparing it to sentences in factually-similar cases. Its purpose is to help us determine

whether the death sentence is imposed in a category of comparable cases often enough to create confidence in the existence of a societal consensus that death is the appropriate remedy. We group cases initially around specific statutory aggravating factors and then subdivide that group according to circumstances that serve either to aggravate or to mitigate the blameworthiness of the defendants in those cases. We view that method as the most persuasive of the frequency tests, Bey IV, supra, 137 N.J. at 353; Marshall II, supra, 130 N.J. at 168, because of its close link to statutory factors and its sensitivity to nuance. Final Report 82-83.

    For Martini, similar cases are those involving kidnapping with particular violence and terror. Martini Report tbl. 6, group H(2). Of the six death-eligible cases in that group, three went to penalty trial, and defendant's was the only case resulting in a death sentence. That creates death-sentencing rates of thirty-three percent for penalty-trial cases and seventeen percent for all death-eligible cases. Martini Report tbl. 7, group H(2). Those ratios are somewhat higher than the overall sentencing rates of thirty percent for penalty-trial cases and thirteen percent for all death-eligible cases. See supra at ___ (slip op. at 15). The figures for the similar cases are as follows:

  Death-Sentencing Rate at Penalty Trial   Death-Sentencing Rate for All Death-Eligible Cases   Proportion of Cases Advancing to Penalty Trial   Including Martini   .33 (1/3)   .17 (1/6)   .50 (3/6)   Excluding Martini   .00 (0/2)   .00 (0/5)   .40 (2/5)  

    Defendant argues that his sentence is excessive because he is the only defendant in category H(2) to receive the death sentence. The State contends that defendant does not fit well in category H(2) because he is the only murderer to have kidnapped for ransom. Moreover, the State adds, a sample size of six cases is too small to create statistical reliability. The AOC concurs with the State's view. It notes that its sample size is too small to generate reliable figures. Additionally, according to the AOC, the five other cases are factually dissimilar, because, as pointed out, no other case in that group involved a ransom or terrorizing non-decedent victims. The AOC suggests that Martini's case may be compared to other murders involving a pecuniary motive.

    We agree with the State and the AOC that the H(2) comparison is not instructive. Unlike Martini's case, none of the other

five cases involves ransom or an extended period of terrorizing and threatening the lives of the victim's family. The rates of death-sentencing, nevertheless, are high enough to show that to the extent that Martini may be compared to others who murder in the course of kidnapping, his sentence is not disproportionate. Although defendant is the only one to be sentenced to death in category H(2), his rates, thirty-three percent for penalty-trial cases and seventeen percent for all death-eligible cases, are comparable to Marshall's. For Marshall, placed in group I(2), contract-murder principals, the percentage of death sentencing was thirty-three for penalty-trial cases and twenty-five for all death-eligible cases. Marshall II, supra, 130 N.J. at 168; Martini Report tbl. 7, group H(2), I(2); Marshall Report, tbl. 7, group I(2). Like Martini, Marshall was the only person in his category to receive the death sentence, yet we found no disproportionality. Similarly, Marko Bey was the first multiple murderer whose death sentence we upheld on proportionality review. Being the first murderer in a category does not support a conclusion of disproportionality. See Bey IV, supra, 137 N.J. at 349-50; Marshall II, supra, 130 N.J. at 166.

    Aside from the feature of kidnapping of a non-stranger with particular violence or terror, defendant's case may be profitably compared to kidnapping of a stranger with particular violence or terror, to contract killers, to principals in contract killings, and to other pecuniary-advantage killers. We may compare Martini

to kidnappers of strangers, because although Martini had singled out Flax in advance, the two were, practically speaking, strangers. Thirty years before the kidnapping, defendant and Flax had had an acquaintanceship, but apparently they had not seen each other since. From the victim's perspective, defendant was a virtual stranger. The numbers in the stranger kidnapping are somewhat higher, but with a sample size (including Martini) of two, the stranger-kidnapping rates' statistical reliability is very low. The figures are:

  Death-Sentencing Rate at Penalty Trial   Death Sentencing Rate for All Death-Eligible Cases   Proportion of Cases Advancing to Penalty Trial   Including Martini   .50 (1/2)   .50 (1/2)   1.0 (2/2)   Excluding Martini   .00 (0/1)   .00 (0/1)   1.0 (1/1)  

To the extent that we give weight to those rates, they support a finding of no disproportionality.

    We may compare Martini to contract killers because he killed his victim as the result of his plan to endanger Flax for money. Moreover, the successful completion of his crime depended in part on the death of Flax. Defendant needed to kill Flax to prevent

the victim from identifying him and from later being a witness against him, and he stated that he shot Flax because he thought that Flax would escape. See Martini, supra, 131 N.J. at 279-85.

    Contract killers show an extremely high frequency of receiving a death sentence. Forty-three percent of such death-eligible defendants received a capital sentence, as did sixty percent of those who advanced to a penalty trial. The figures, with and without Martini, are as follows:

  Death-Sentencing Rate at Penalty Trial   Death-Sentencing Rate for All Death-Eligible Cases   Proportion of Cases Advancing to Penalty Trial   Including Martini   .66 (4/6)   .50 (4/8)   .75 (6/8)   Excluding Martini   .60 (3/5)   .43 (3/7)   .71 (5/7)  

    Those rates, strikingly high, show no disproportionality. When placed into group I(1), defendant shows a sixty-six-percent probability of receiving the death sentence at his penalty trial. We note that the sample size, eight including defendant, is small and that defendant is obviously not perfectly comparable to a contract killer. A contract killer generally is not involved in the selection of the victim but kills simply for the money and

reputation. Here, defendant was actively involved in the selection of the victim and killed to ensure the successful completion of his plan. Although defendant is related closely enough to the contract-killer group to allow an analogy, the problems with the comparison merit consideration in the weight we give to its results.

    Defendant may be compared to principals in contract killings because of the similarity of their roles in planning the crime. Without Martini, twenty-five percent of such death-eligible cases resulted in a capital sentence, as did thirty-three percent of those cases that advanced to a penalty trial. The figures, with and without defendant, are:

  Death-Sentencing at Penalty Trail   Death-Sentencing Rate for All Death-Eligible Cases   Proportion of Cases Advancing to Penalty Trial   Including Martini   .50 (2/4)   .40 (2/5)   .80 (4/5)   Excluding Martini   .33 (1/3)   .25 (1/4)   .75 (3/4)  

    Those figures, like the contract-killer results, clearly support a finding of no disproportionality. Placed in group I(2), defendant shows a fifty-percent chance of receiving a death

sentence at the penalty trial. However, the sample size, five with defendant and four without him, is too small to allow us to rely heavily on the percentages generated by the I(2) table.

    The last category is group I(3), other pecuniary-advantage killers. We compare Martini to defendants in that group because he committed his crime for money. With defendant in group I(3), his percentages are high: fifty percent of death-eligible cases receive the death penalty, as do fifty percent of penalty-trial cases.

  Death-Sentencing Rate at Penalty Trial   Death-Sentencing Rate for All Death-Eligible Cases   Proportion of Cases Advancing to Penalty Trial   Including Martini   .50 (1/2)   .50 (1/2)   1.0 (2/2)   Excluding Martini   .00 (0/1)   .00 (0/1)   1.0 (1/1)  

Although a capital-sentencing rate of fifty percent supports a finding of no disproportionality, the sample size of two, including defendant, precludes reliance on those results.

    Accordingly, we conclude that the salient-factors measure, the most persuasive of the statistical measures, supports a

finding of no disproportionality. It shows that defendant's class generally receives the death penalty. Although the small sample sizes of the groups in the salient-factors test preclude us from investing those results with great weight, we will use them as a check against our findings under both the other two statistical tests and under the precedent-seeking approach.

             2. The Numerical-Preponderance-of-Aggravating-and-Mitigating-Factors Test

    This test compares Martini's case to other cases having the same number of aggravating and mitigating factors. It uses those raw numbers to measure blameworthiness, on the assumption that the presence of more aggravating factors and fewer mitigating factors renders a case more blameworthy. The purpose of using the numerical-preponderance test is to provide a means to control for the fundamental problem of the salient-factors test, i.e., that our universe contains too few cases in any one factual category to permit reliable inferences. The obvious problem with the numerical-preponderance approach is that it assumes that juries weigh each factor equally. Its inability to account for the qualitative character of jury deliberations makes this test more problematic than either the salient-factors test or the index-of-outcomes test. Marshall II, supra, 130 N.J. at 171. To alleviate that problem, the numerical-preponderance test attempts to weight the statutory factors to account for qualitative determinations.

    Here, the jury found two aggravating factors: c(4)(f), murder to escape detection, and c(4)(g), murder during the course of a kidnapping. Six of the twelve jurors found mitigating factor c(5)(c), age, and six found mitigating factor c(5)(h), the catchall factor. All jurors rejected mitigating factors c(5)(a), extreme mental or emotional disturbance, and c(5)(d), diminished capacity. As we stated in Bey IV, supra, "In the frequency analysis, we will consider only those mitigating factors that the jury found relevant to the imposition of the death penalty." 137 N.J. at 361. Unlike precedent-seeking review, in which we will consider "objective factors that are clearly present in the record even if the jury did not find them to be relevant," on frequency analysis we will not use additional factors "because of the need to maintain the uniformity of the statistics." Ibid. If we reworked one case, a valid statistical comparison would require that we rework and reweigh all the cases in the universe.

    The Martini Report identifies twenty-three penalty-trial cases that contain two aggravating and two mitigating factors. Thirteen of those cases resulted in death penalties, producing a death-sentencing rate of fifty-seven percent. Martini Report tbl. 8. For the forty-eight death-eligible cases in that class, the percentage is twenty-seven percent (13/48). Id. at tbl. 9. The figures without defendant are slightly lower:

  Death-Sentencing Rate at Penalty Trial   Death-Sentencing Rate for All Death-Eligible Cases   Inc