dissenting.
Proportionality review of death sentences in New Jersey has become crude and imprecise. It allows for the possibility that the Court may find a sentence proportionate based upon comparison to the sentences of a designated group of defendants; yet that review may have overlooked comparison to a number of individuals most like defendant, perhaps even more culpable. We consciously accept that risk, as we accept that we may sacrifice a perfectly individualistic assessment of a defendant’s death sentence in the interest of uniformity. See State v. Ramseur, 106 N.J. 123, 330, 524 A.2d 188 (1987) (noting that dual aims of uniformity and individualization create an “inherent paradox in the process”); see also id. at 347, 351, 524 A.2d 188 (Handler, J., dissenting) (noting that premises of uniformity and individuality underlying proportionality review are fundamentally inconsistent and irreconcilable).
The present case demonstrates how the risks of classification that we generally accept as necessary elements of the review may be exploited. The procedural framework designed to protect the rights of capital defendants has been manipulated to defendant’s disfavor. John Chew was classified as death-eligible solely on the ground that he murdered for insurance proceeds; a classification which, I maintain, was contrary to the intent of N.J.S.A. 2C:ll-3c. Having sentenced Chew as a member of a group of pecuniary gain *250killers from which he should have been set apart, the Court presently fails to compare Chew to the death-eligible defendants with whom he has the most in common — those who murdered loved ones, relatives, or friends for insurance proceeds or inheritance. Comparison with these defendants, none of whom were death-sentenced, shows that Chew’s sentence is arbitrary.
For this reason, and because I maintain that evidence of the risk of race discrimination in the imposition of this State’s death penalty mandates that we discontinue capital punishment, see State v. Loftin, 157 N.J. 253, 373-412, 724 A.2d 129 (1999) (Handler, J., dissenting) (Loftin II); see also State v. Harvey, 159 N.J. 277, 361-74, 731 A.2d 1121 (1999) (Handler, J., dissenting) (Harvey III); accord State v. Cooper, 159 N.J. 55, 163-64, 731 A.2d 1000 (1999) (Handler, J., dissenting) (Cooper II), defendant’s death sentence must be vacated.
Therefore, I dissent.
I
On January 12, 1995, John Chew slashed the throat of his girlfriend, Theresa Bowman, as she sat in the driver’s seat of his Corvette. State v. Chew, 150 N.J. 30, 42, 695 A.2d 1301 (1997). The police learned that Bowman was having an affair, id. at 44, 695 A.2d 1301, and that she and Chew had agreed to separate on the day she was murdered. Id. at 48, 695 A.2d 1301. The police also discovered that Chew was the beneficiary of a $250,000 life insurance policy on Bowman, and that his plans to kill his girlfriend for the benefit were longstanding. Id. at 43-44, 695 A.2d 1301.
A jury convicted Chew of purposeful or knowing murder by his own conduct, contrary to N.J.S.A. 2C:ll-3a(1) or (2), as well as possession of a weapon for an unlawful purpose, contrary to N.J.S.A. 2C:39-4d. Id. at 49, 695 A.2d 1301. The jury’s finding that Chew “committed the murder as consideration for the receipt, or in expectation of the receipt of anything of pecuniary value,” *251N.J.S.A 2C:11-3c(4)(d), triggered a penalty trial. Id. at 50, 695 A.2d 1301. The jury found no other aggravating factors.
At the penalty trial, defendant presented extensive mitigating evidence, most of which the jury found to support N.J.S.A. 2C:11-3c(5)(h), the catch-all mitigating factor. See ibid. The jury unanimously found that Chew was emotionally and culturally deprived by his parents, who failed to provide him with basic guidance, structure and stability. His delinquent behavior was ignored or even condoned: he was not encouraged academically; excessive absences were tolerated, culminating in his leaving school in the seventh grade; and he and his brothers engaged in thievery and other delinquent acts, much of which was done with their parents’ knowledge and approval. Further, both parents were involved in numerous extramarital affairs; his mother often brought her lovers home with the children present.
Eleven jurors also found that defendant was raised in a violent home, where he witnessed serious acts of violence by both parents against each other. His mother, the jury appreciated, was immature, resentful and ill-equipped to care for her children and meet their basic physical and emotional needs; whereas his father was financially irresponsible, emotionally absent, and sexually promiscuous.
Nine jurors found that defendant had suffered from serious mental and emotional disturbance as a youth, reflected in numerous juvenile institutionalizations. They recognized that at multiple times in his life Chew was effectively abandoned by his parents. Chew’s parents were frequently separated from each other and while the parents alternated custody arrangements the non-custodial parent failed to maintain contact with their children.
Seven jurors found that the Chew family moved so frequently that not only were long lasting friendships non-existent, but any intervention to address the family’s multitude of problems by school or mental health officials was almost impossible. Defendant’s education difficulties were evident as early as the second grade, but little, if any, effort was made to correct or compensate *252for these deficiencies. As a result, Chew developed a fear of school which contributed to poor self esteem and loss of self worth.
Six jurors appreciated defendant’s relationship with his eleven-year old daughter Valerie, whom he loved and who loved him in return. Although Chew’s relationship with his daughter was limited by Chew’s incarceration, these jurors recognized that Chew continued to be an important and necessary part of Valerie’s life and would continue to play a significant role in her upbringing. See ibid.
Five jurors found that defendant’s parents raised him to believe he was worthless — his mother told him she hated him and wished he was dead and his father often denied paternity. Three jurors found that defendant’s parents failed to protect him from harm, ignoring signs of sexual abuse by his maternal aunt and witnessing physical abuse by his maternal grandmother. Ten jurors found an additional, unspecified catch-all mitigating factor.
Despite Chew’s anguished confessions, the penalty jury concluded that the sole aggravating factor — that Chew killed for pecuniary gain — outweighed the mitigating evidence and sentenced Chew to death. Ibid.
II
A sentencing jury must unanimously find at least one statutory aggravating circumstance in order for a defendant convicted of first degree murder to be eligible for a death sentence. See N.J.S.A. 2C:11-3. “In this way the jury’s discretion is channeled. No longer can a jury wantonly and freakishly impose the death sentence; it is always circumscribed by legislative guidelines.” Gregg v. Georgia, 428 U.S. 153, 206, 96 S.Ct. 2909, 2940, 49 L.Ed.2d 859, 893 (1976). Chew’s jury found only one aggravating factor — that defendant “committed the murder as consideration for the receipt, or in expectation of the receipt of anything of pecuniary value.” N.J.S.A. 2C:11-3c(4)(d). N.J.S.A. 2C:11-3c(4)(d), however, was not intended to and should not encompass killings for insurance proceeds. See Chew, supra, 150 N.J. at 88-*253101, 695 A.2d 1301 (Handler, J., dissenting). The c(4)(d) aggravating factor should not have been before the jury as a matter of law. See ibid. Defendant’s eligibility for death was thus entirely dependent upon the jury’s mistaken belief that Chew, if he murdered for insurance proceeds, killed for “pecuniary value” as contemplated by N.J.S.A. 2C:11-3c(4)(d).
The language and construction of the statutory provision — “as consideration for the receipt, or in expectation of the receipt of anything of pecuniary value” — demonstrate that the c(4)(d) aggravating factor was intended to apply solely to contract killers, those who accept payment in return for committing a homicide. “Consideration” imports a mutual understanding between two parties, a meeting of the minds. See id. at 90, 695 A.2d 1301 (Handler, J., dissenting). The clause “expectation of the receipt of anything of pecuniary value” merely extends the requisite consideration to a future payment, as opposed to one past or present. See id. at 91, 695 A.2d 1301 (Handler, J., dissenting); accord State v. Clausell, 121 N.J. 298, 344, 580 A.2d 221 (1990) (holding instruction on c(4)(d) factor requires “telling jury that it must specifically find that defendant either received payment, or expected to receive payment, for having killed [victim]”). As such, statutory aggravating factors c(4)(d) and N.J.S.A 2C:11-3c(4)(e), which applies to a defendant who “procured the commission of the offense by payment or promise of payment of anything of pecuniary value,” are symmetrical complements. Both involve consideration and apply to contract murderers: the former applies to the hired gun, the latter to the hirer. See Chew, supra, 150 N.J. at 93, 695 A.2d 1301 (Handler, J., dissenting).
This interpretation of the c(4)(d) aggravating factor is consistent with the Court’s reasoning in State v. DiFrisco, 142 N.J. 148, 662 A.2d 442 (1995) (DiFrisco III). DiFrisco was convicted of murder-for-hire, for which the sentencing jury found that the c(4)(d) aggravating factor applied. See State v. DiFrisco, 137 N.J. 434, 450, 645 A.2d 734 (1994) (DiFrisco II). For purposes of proportionality review, the Administrative Office of the Courts (AOC) *254classified DiFrisco in category I, which includes defendants who committed a “murder involving a pecuniary motive other than robbery or burglary,” excluding those whose crimes involved sexual assault, arson, kidnapping, multiple victims, or a victim who is a public servant. See DiFrisco Report, tbl. 7 (Nov.- 7, 1994); accord ChewICooper/Harvey Report, tbl. 7 (Dec. 3, 1997) (CCH Report). Category I is comprised of four subcategories: contract killers, subcategory 1-1; contract principals, subcategpry 1-2; other pecuniary gain murderers, subcategory 1-3; and contract killers who were hired by the victim, subcategory 1-4. See id., tbl. 6. DiFrisco, a hired killer, was classified in subcategory 1-1. For proportionality comparison, the Court used a composite of the subcategories, comparing DiFrisco to all defendants in category I.
An unmodified comparison of DiFriseo’s sentence to the composite I category would have included cases AOC-designated as I-3 for “other pecuniary gain” murders, such as murders motivated by the receipt of inheritance or life insurance benefits. The Court, however, declined to compare DiFrisco to defendants who killed for insurance proceeds. Specifically, the Court removed defendant Walter Williams, classified as 1-3, on the ground th'at “the jury rejected the pecuniary-motive aggravating factor at his penalty trial.” DiFrisco III, supra, 142 N.J. at 170, 662 A.2d 442.1 The Court further expressly declined DiFrisco’s request to include defendants Anthony Accetturo, Louis Auriechio, Thomas Ricciardi, Michael Pema, and Michael Tarcetta in the comparison group on the ground that “there appeared] to be no basis for an allegation that any of those defendants were either paid to commit murder or that they paid another to do so____” Id. at 167-68, 662 A.2d 442. And the Court denied DiFrisco’s request to include defendant Patrick Lanzel in the comparison, finding his “motive for *255killing was not pecuniary; therefore [his] comparison with defendant is inappropriate.” Id: at 169, 662 A.2d 442. Lanzel killed his cousin’s parents so that he and his cousin could share the proceeds of their life insurance and inheritance.
Thus, the thrust of the Court’s language in DiFrisco III is quite clear: inheritance killings or killings for insurance proceeds— specifically, non-contract murders — are not “pecuniary” within the meaning of the death penalty statute.
The error in the present case was exacerbated by the trial court’s jury instruction on the c(4)(d) aggravating factor. The court instructed the jury that for the c(4)(d) factor to apply and for defendant to be death-eligible the jurors “must unanimously find, beyond a reasonable doubt, that, at least, one of the purposes John Chew had, for murdering Theresa Bowman, was to obtain the insurance proceeds.” The court continued: “although the receipt of the insurance proceeds does not have to be the defendant’s exclusive purpose for killing Theresa Bowman, the receipt of such proceeds must be much more than just an incidental benefit to the defendant, because of Theresa Bowman’s death. That is, the expectation of pecuniary gain must be the cause of the murder, or one of the causes of the murder, and not a result of it.” By inserting the phrase ‘to obtain insurance proceeds’ in place of the statutory language ‘for pecuniary gain’ without reference to contractual arrangements, the court deprived Chew’s penalty jury of any opportunity to apply the e(4)(d) factor as statutorily defined.
The resulting prejudice is tantamount. Chew would not have been death-eligible but for the jury’s finding that his conduct met that contemplated by the c(4)(d) factor. Had that jury finding been based upon a court instruction consisting of a reading of the statutory language, defendant might have only complained that the statutory language is vague. But here the trial judge’s decision to redefine the statutory provision in terms relevant to the case rendered the jury’s finding of the e(4)(d) aggravating factor nothing more than a finding that Chew committed the *256murder to collect the insurance proceeds. Given that the statutory provision does not apply to insurance proceeds in the absence of a contract to kill, defendant was not death-eligible, the jury’s finding of the c(4)(d) aggravating factor was in error, and vacation of Chew’s death sentence is still required. Accord Chew, supra, 150 N.J. at 99-101, 695 A.2d 1301 (Handler, J., dissenting).
It is unsettling to compare the death sentence of a defendant who by law should not have been death-eligible with the sentences of others who were. Nevertheless, given that a majority of this Court affirmed defendant’s conviction and death sentence on direct appeal, see id. at 88, 695 A.2d 1301, Chew is constitutionally entitled to proportionality review. See N.J.S.A 2C:11-3e; State v. Marshall, 130 N.J. 109, 117, 613 A.2d 1059 (1992) (Marshall II).
Ill
The principles of proportionality review are often recited. The applications of those principles are less secure, but for purposes of this review they remain those veteran and familiar: statistical frequency analysis, comprised of the salient-factors and index-of-outcomes tests, and precedent-seeking review, a case-by-case comparison of defendant’s sentence with those received by similarly situated defendants. See ante at 199, 731 A.2d at 1078.
The basic standard by which we have measured disproportionality has been that “[a] death sentence is comparatively excessive if other defendants with similar characteristics generally receive sentences other than death for committing factually similar offenses in the same jurisdiction.” Marshall II, supra, 130 N.J. at 153-54, 613 A.2d 1059 (citation omitted). In other words, a death sentence is not disproportionate .if it is generally received by similarly situated defendants. See DiFrisco III, supra, 142 N.J. at 171, 662 A.2d 442; State v. Martini, 139 N.J. 3, 30, 651 A.2d 949 (1994) (Martini II); but see Cooper II, supra, 159 N.J. at 93-94, 116, 731 A.2d 1000 (emphasizing that sentence is disproportionate if it is “aberrational” or if defendant has been “singled out *257unfairly for capital punishment”); Harvey III, supra, 159 N.J. at 356-57, 731 A.2d 1121 (Handler, J., dissenting) (noting Court has used up to fourteen different standards and substandards for measuring disproportionality); Loftin II, supra, 157 N.J. at 417-18, 724 A.2d 129 (Handler, J., dissenting) (noting Court has used at least six different standards for measuring disproportionality). The application of these standards, I maintain, requires some notion of numerical preponderance. Accord Loftin II, supra, 157 N.J. at 419-20, 724 A.2d 129 (Handler, J., dissenting); DiFrisco III, supra, 142 N.J. at 212, 662 A.2d 442 (Handler, J., dissenting); Martini II, supra, 139 N.J. at 90-91, 651 A.2d 949 (Handler, J., dissenting); State v. Bey, 137 N.J. 334, 408, 645 A.2d 685 (1994) (Handler, J., dissenting) (Bey IV).
The deficiencies attendant to our proportionality review are as customary as the process itself. From the inaugural use of these models, the Court noticed problems; and with subsequent scrutiny, defects once hidden have become recognizable blemishes. See generally Loftin II, supra, 157 N.J. at 414-15, 724 A.2d 129 (Handler, J., dissenting) (discussing failure to identify standard by which to measure general imposition of death penalty and inherent subjectivity of precedent-seeking review); Martini II, supra, 139 N.J. at 82-106, 651 A.2d 949 (Handler, J., dissenting) (also discussing Court’s use of reversed death sentences in proportionality review universe).
In light of these defects, the Court appointed the Honorable David S. Baime as Special Master to evaluate the proportionality review methodology. See Loftin II, supra, 157 N.J. at 263-77, 724 A.2d 129 (Order). The Special Master’s report, submitted April 28, 1999, confirms that aspects of our methodology are faulty and in need of revision. See Honorable David S. Baime, Report to the New Jersey Supreme Court: Proportionality Review Project 6-7 (Apr. 28, 1999) (Special Master Report). As I note elsewhere, I cannot in good faith support the majority’s decision, see ante at 196, 202, 731 A.2d at 1077, 1080, not to consider the Special *258Master’s recommendations, which are significant, before assessing the proportionality of Chew’s death sentence. See Harvey III, supra, 159 N.J. at 354, 731 A.2d 1121 (Handler, J., dissenting); Cooper II, supra, 159 N.J. at 163, 182, 731 A.2d 1000 (Handler, J., dissenting). At the very least, the Special Master Report assures us that there are many problems that contribute to the inadequacy of our proportionality review, and ultimately to the arbitrary nature of defendant’s death sentence. With that in mind, I discuss herein only those that pose particular problems in defendant’s case.
A.
Before undertaking frequency and precedent-seeking analyses of Chew’s death sentence, I note my objection to the Court’s adjustment to Chew’s comparison group. Having sentenced Chew as a member of a group of pecuniary gain killers from which he should have been set apart, the Court now proceeds in proportionality review to exclude Walter Williams, a defendant with whom Chew has the most' in common, from Chew’s AOC-designated comparison group. Further, the Court’s inclusion of contract-murder defendants in Chew’s proportionality review does not square with the Court’s exclusion of non-contract pecuniary-motive defendants in DiFrisco III, where the defendant was a contract killer; the small sample size of the 1-3 category does not eviscerate this intellectual hurdle. I also disagree with the Court’s decision not to compare Chew’s sentence to the sentences of certain defendants outside the I category who killed in furtherance of a pecuniary motive.
1.
Chew was assigned to category 1-3. See CCH Report (using database including all death-eligible cases through July 31, 1997). Subcategory 1-3, which designates defendants in the I category who did not commit a contract killing, see id., tbls. 6-7, consists of three defendants, including Chew, see id., tbl. 7, each of whom *259killed for insurance or inheritance. Of these, Chew is the only one for whom a jury found the e(4)(d) aggravating factor. That finding was error, as this Court’s reasoning in DiFrisco III supports. See discussion supra at 192-96, 731 A.2d at 1074-76.
The Court removes Walter Williams from the 1-3 group in response to the State’s objection because the jury rejected the c(4)(d) (pecuniary-motive) aggravating factor at Williams’s penalty trial.2 See ante at 199, 731 A.2d at 1078. The Court reasons— and Chew improvidently concedes — that Williams and Chew are not, therefore, similarly situated. See id. at 199-200, 731 A.2d at 1078-79. I disagree and object to Williams’s removal. Whereas DiFrisco was a contract killer, from whose comparison group Williams was therefore sensibly removed, Chew and Williams— each a unilateral actor — have much in common. By this Court’s able judgment in DiFrisco III, Chew would be properly excluded from DiFrisco’s review as well. By the same reasoning, Williams must remain in Chew’s comparison group.
Although we are not in any position to modify Chew’s prosecution and conviction as a death-eligible defendant in the wake of this Court’s affirmance of Chew’s conviction and sentence on direct appeal, see Chew, supra, 150 N.J. at 88, 695 A.2d 1301, we may at least partially remedy the unfairness of the situation by comparing Chew to Walter Williams, who, unlike Chew, was properly found not to have qualified for aggravating factor c(4)(d). We may also consider the impropriety of placing great weight in a comparison of Chew, who apparently murdered for insurance proceeds, to defendants who were party to a contract killing. I would leave Williams (and Rose) in the comparison group for *260frequency analysis, and would emphasize comparison of defendant to Williams and other similarly situated defendants extrinsic to the I category on precedent-seeking review.
2.
In addition to the improper exclusion of Walter Williams from the comparison group, I lament the Court’s failure to compare Chew’s sentence for purposes of precedent-seeking review to the sentences of select other defendants not classified in the I category who killed in furtherance of a pecuniary motive under circumstances similar to defendant.
Defendant presented thirty-four cases extrinsic to the I category that he deemed relevant for our comparison of precedent.3 The Court, abiding by the principle of unique classification, refuses to include any of the cases outside the I category in Chew’s review. See ante at 200-01, 731 A.2d at 1079 (citing DiFrisco III, supra, 142 N.J. at 167, 662 A.2d 442). The Court categorically rejects comparison of Chew’s case to the preponderance of the cases suggested by the defendant because they are robbery murders, which the Court views as “significantly different from *261pecuniary-motive murders.” Ibid. I disagree with the Court on both points.
Although I do not favor unique classification as a principle, I recognize the possibility that in some instances an AOC-elassification may be sufficiently comprehensive to include all death-eligible cases relevant for a precedential review of a defendant’s sentence. That is not so in this case. Defendants who murdered for insurance proceeds or inheritance should not be excluded from Chew’s precedent-seeking review merely because they are classified in a category that trumps the I category for purposes of AOC-classification. Whether a defendant kills one person or more than one person for insurance proceeds or inheritance, the anticipated pecuniary gain surely is the “essential attribute” of the crime. Therefore, insofar as our proportionality review comparisons are “an aid to understanding a defendant’s death sentence in relation to other defendants whose murders share the same ‘essential attribute,’ ” ante at 201 n. 3, 731 A.2d at 1079 n. 3, Patrick Lanzel, classified as A-l for a multiple homicide, who killed both of his cousin’s parents for an expected inheritance, should be added to the review. See discussion infra at 271-72, 731 A.2d at 1118. The case of Darrell Collins, who killed his wife and son for insurance proceeds but for whom the jury found no aggravating factors, should also be compared to defendant’s' on precedent-seeking review. See ibid.
The Court’s categorical rejection of comparison to robbery murders raises a fundamental question regarding the AOC’s salient-factor groupings. Elements of a contract killing arguably make it distinguishable from a murder that occurs during the commission of a robbery. Contrary to the Court’s assertion, however, there is not an objectively significant difference between all robbery murderers and defendants who kill for insurance proceeds or inheritance. Robbery murders involving victims who were acquainted with and targeted by their assailant, in particular, bear an intrinsic similarity to murders for insurance or inheritance, in which the defendant is a beneficiary. The nature of cases in the current death-eligible universe suggests that defen*262dants such as Chew who acted alone and murdered for insurance proceeds or inheritance, now classified by the AOC as 1-3, may have as much if not more in common with certain classes of robbers, particularly those who murdered and robbed acquaintances, now classified as E-6, than they do with conspiratorial contract killers.4 This is most apparent in the case of codefendants Carl Norman and Kevin Smith, who killed a friend to acquire the inheritance the victim had received from a deceased relative a week prior; and in the case of William Cooper, who murdered the victim, a eoworker, for his paycheck. See discussion infra at 272-74, 731 A.2d at 1118-19.
*263Insofar as the Court insists upon adhering to the principle of unique classification, the issue of whether killings for insurance proceeds and inheritance might be better placed among robbery-murders than with contract killings, or perhaps combined with robbery murders between acquaintances, warrants consideration. At present, I would expand precedential review of Chew’s sentence to include comparison to certain defendants classified in the robbery murder category, rather than impose a blanket restriction prohibiting such comparison. Therefore, in addition to all members of the I category, as well as Patrick Lanzel and Darrell Collins, I would focus particular attention upon those defendants whose crimes share characteristics fundamental to Chew’s, such as a pecuniary motive that exceeds mere cash or possessions, significant premeditation, and some acquaintance with the victim— namely, those previously mentioned, William Cooper, Carl Norman and Kevin Smith.
B.
I discuss Chew’s frequency analysis briefly because I am in overall agreement with the Court that the statistical results as such do not provide a basis for finding disproportionality of sentence. As noted previously, see supra at 199, 731 A.2d at 1078, I would include Walter Williams in the 1-3 and Michael Rose in the 1-1 category for purposes of comparison. With or without Rose and Williams, however, Chew’s numbers are not convincing. Pecuniary-gain murderers have a relatively high death-sentencing rate. Moreover, the factors producing Chew’s high culpability level in the index-of-outcomes regressions are derivative of his own' heinous crime.
According to the salient-factors test, ■ looking to the death-sentencing rates and the penalty-phase advancement rates for all pecuniary-motive murderers, with Rose and Williams excluded from the comparison group, the death sentencing rate for I-class *264defendants is fifty percent.5 See CCH Report, tbl. 7 (reprinted ante at 204, 731 A.2d at 1081). Adding Rose and Williams to the calculation, the death sentencing rate statistic drops to forty percent. See infra, Appendix A. These statistics suggest that similarly situated defendants receive sentences other than death only slightly more often than death sentences. Considered alone, the numbers do not provide a basis for a strong argument that defendant’s sentence was disproportionate. The ten percent decrease in death-sentencing rate when Williams and Rose are added to the calculation, however, convincingly demonstrates that “the small sample sizes of the groups in this salient-factors test preclude us from investing great weight in those results.” DiFrisco III, supra, 142 N.J. at 174, 662 A.2d 442.
Chew’s index-of-outcomes ratings are even less helpful to his claim of disproportionality. The results of the regressions assessing only penalty-trial cases demonstrate that when penalty-phase juries have decided sentences for similarly situated defendants, they have generally imposed death. Considering statutory factors only in the penalty-trial universe, Chew falls in culpability level five, in which seventy-six percent of the defendants received the death penalty. See CCH Report, tbls. 23, 24; ante at 207, 731 A.2d at 1083. He fares no better by a statutory and non-statutory factor measure, falling in culpability level four, where nine of the twelve defendants who proceeded to penalty trial were sentenced to death. See CCH Report, tbls. 21, 22; ante at 207, 731 A.2d at 1083.
Measured by the full universe of death-eligible defendants and statutory factors only, defendant is in culpability level three, in which the death-sentencing rate is fifty-two percent. See CCH Report, tbls. 23, 25; ante at 207, 731 A.2d at 1083. Although *265defendant appears less culpable here, the mid-range result does little to offset the high percentage of death-sentenced peers reflected in the other tests.
Only when the statutory and non-statutory variables are considered in the full universe does defendant’s death sentence appear disproportionate. By this regression, defendant’s culpability score falls to nineteen percent and he occupies culpability level one, at which only five percent of the defendants have been death-sentenced. See CCH Report, tbls. 13, 14; ante at 207, 731 A.2d at 1083. This regression reveals a disturbingly low probability of death; yet, in light of the other regressions, the results of this regression do not establish that defendant’s death sentence is disproportionate.
Nevertheless, the low predicted value of the last regression requires this Court to carefully scrutinize defendant’s sentence in precedent-seeking review. In addition, the exceptionally large confidence intervals in this case and the wide variance in defendant’s culpability scores demand that the Court not accord this test substantial weight.6 See ante at 209-10, 731 A.2d at 1084. *266Indeed, the index-of-outcomes regressions are so unreliable that they are, arguably, useless. See Special Master Report, supra, at 107 (recommending that the Court abandon the index-of-outcomes test).
C.
Although the Court’s adjustments to the comparison group have minimal effect on the statistical frequency analysis results, the definition of Chew’s comparison group substantially affects precedent-seeking review of his sentence. Before addressing how a more expansive review accentuates the arbitrary nature of Chew’s death sentence, I pause to reflect upon the precedent-seeking analysis conducted by the Court.
1.
Generally, I agree with the Court that Chew’s culpability is high. The manner in which Chew murdered his girlfriend was brutally cruel. Nevertheless, the attention the Court devotes to Chew is routine. The Court fails to distinguish Chew’s insurance-motivated crime from contract killings. Even if I were to agree with the relevance of a comparison between Chew and contract murder defendants, I would find the Court’s analysis reckless.
Attempting to differentiate Chew from the 1-2 defendants, the Court claims, “[njone of the defendants in this category killed solely for pecuniary gain.” Ante at 218, 731 A.2d at 1089. And explains: “Brand paid to rid his family of an abusive brother, the Engel brothers planned the death of William’s ex-wife out of a warped sense of jealousy and resentment, and Marshall had his wife killed in order to eradicate an obstacle to his extra-marital romance.” Id. at 218, 731 A.2d at 1089. How this distinguishes Chew, whose motivation in part for killing Bowman, as the record amply demonstrates, was because she was leaving him, is a mystery. The Court’s distinction of Marshall from Chew is awk*267wardly moralistic, portraying Marshall as less culpable because he killed not just for money, but to be with his paramour.
The Court then attempts to distinguish Chew from the Engels brothers, “[mjore importantly],” on the ground that the Brand and Engels juries found the c(5)(a) mitigating factor. See id. at 218, 731 A.2d at 1089. This does not gel with the Court’s claim to “recognize that even when a jury rejects a specific mitigating factor, it may nonetheless remain influenced by the evidence presented in support of that mitigating factor.” Id. at 210, 731 A.2d at 1085 (citing Loftin II, supra, 157 N.J. at 336, 724 A.2d 129; DiFrisco III, supra, 142 N.J. at 185, 662 A.2d 442; Bey IV, supra, 137 N.J. at 368, 645 A.2d 685). The Court here disregards Chew’s substantial account of his mentally and emotionally troubled life. The inconsistency continues as Chew is haphazardly declared more culpable for engaging others in crime, see id. at 217, 731 A.2d at 1088, but also “more blameworthy” for having “carried out his bloody plot alone,” id. at 219, 731 A.2d at 1089.
The Court’s attempts to differentiate Chew from the 1-1 defendants are equally senseless. Referring to this Court’s assessment of why Chew deserves to die while other life-sentenced 1-1 defendants (Burroughs, Harris, Irizarry, Melendez, and Pinehom) do not, one may glean that Chew’s death sentence is understandable — proportionate—because Chew was not cooperative with the police, he was relatively mentally stable, and he “planned the murder ... for over a year.”7 Id. at 217-18, 731 A.2d at 1089; see also id. at 220, 731 A.2d at 1090 (“In sum, unlike many of the life-sentenced pecuniary-motive murderers, Chew did not substantially assist the State and was not mentally disturbed when he murdered the victim.”). The notion that these characteristics *268rationalize a distinction between life and death sentences is notably absurd, and illuminates the problematic moral determinations this Court undertakes in precedent-seeking review. Accord Marshall II, supra, 130 N.J. at 273-74, 613 A.2d 1059 (Handler, J., dissenting).
The Court slights the defendant by conducting a whimsical review. If this is what proportionality review has come to — if the root of the analysis is “whatever works” — an opinion simply affirming or reversing defendant’s sentence would suffice. Of course, we strive to accomplish more than this. Accordingly, adjustments are warranted.
2.
I take particular note of the Court’s comparison of Chew to 1-3 defendant Celestine Payne. Payne killed her husband with poison, after her request of a friend to fatally shoot her spouse went unfulfilled. As a result, Payne, the unemployed mother of four, received $49,000 in life insurance proceeds. Subsequently, Payne took out life insurance policies on two of her boarders, and hired Charles Pinchom to kill them both. Pinehom’s attempt was only partially successful, murdering one. Payne pled guilty to two concurrent life terms with thirty-year parole disqualifiers.
The Court emphasizes Payne’s clinically diagnosed depression and schizophrenia as the distinguishing factor between her life sentence and Chew’s death verdict. See ante at 215, 731 A.2d at 1087. In Loftin II, supra, the Court similarly sought to distinguish Loftin’s death sentence from the life sentences of similarly situated defendants on the ground that Loftin failed to present uncontroverted evidence that he suffered from mental disease or defect at the time of the crime. 157 N.J. at 342-43, 345, 724 A.2d 129 (distinguishing Loftin from B-2 death-sentenced defendants and B-3 defendants). Although the Court may not be completely off the mark in thinking that jurors respond sympathetically to evidence of mental disease, using clinical evidence of mental illness as a dispositive .basis for distinction, I fear, is somewhat unfounded *269and quite possibly mistaken. Defendant has presented evidence that defendants who qualify for the N.J.S.A. 2C:11-3c(5)(a) mitigating factor, which recognizes extreme mental or emotional disturbance, are more likely to be prosecuted capitally than defendants who do not present evidence of extreme mental or emotional distress. See CCH Report, tbls. 7, 10 (reprinted ante at 224, 731 A.2d at 1092); see also ante at 225, 731 A.2d at 1093 (recognizing “a positive correlation between the mental disturbance factor and capital prosecution”). The statistics also show that, although defendants with the c(5)(a) factor are slightly less likely to be sentenced to death at a penalty trial, they are nevertheless sentenced to death at a higher rate than death-eligible defendants generally. See CCH Report, tbls. 7, 10. Even if these results are somewhat anomalous, they present the possibility that evidence of mental disturbance does not always help a defendant before a sentencing jury.
The Court’s assessment regarding mental distress evidence is conclusory. Although this Court’s proportionality review borders on mechanical, we must give our jurors more credit. We must assume that jurors’ deliberations are more complex and emotions more proximate than the Court’s analysis reflects. We should not dismiss a jury’s awareness of the possibility that many individuals with psychiatric impairments seek neither clinical diagnosis nor treatment — indeed, such avoidance may be symptomatic of the affliction. If it is more than hyperbole that the first step to recovery is acknowledging the problem, Celestine Payne may in a juror’s view be a step beyond Chew along that path.- In the same regard, jurors might have concluded that the medication Payne was taking was helping her condition. The Court, therefore, is impulsive to rely so confidently on the presentation of evidence of mental distress as a dispositive factor.
Recognizing the impetuous nature of the Court’s interpretation of extreme emotional or mental distress does little to alter the fact *270that Chew presented little direct evidence of present mental problems. Yet Chew offered a substantial account of his troubled upbringing. See supra at 251-52, 731 A.2d at 1106-07. At least one juror found all of the evidence of mental distress presented by Chew to contribute to the catch-all mitigating factor. See ibid. To some extent, as the previous discussion suggests, the mental problems rooted in Chew’s childhood and adolescence are implicit in his homicidal act and reflected in the means he used to deal with his personal problems.
Chew and Payne, in fact, have quite a lot in common. Both killed for insurance proceeds; both killed a significant other; both killed helpless victims; and both planned the murders in advance. But Chew is less culpable than Payne in two respects: whereas Payne killed her husband, and paid Pinchom to kill two boarders, Chew killed only Bowman and acted alone. And, whereas Payne killed only for insurance proceeds, Chew’s homicidal tendencies were in part motivated by the fact that Bowman was going to leave him that very day. Were Chew’s sentence compared only to Payne’s, it would certainly seem unfair.
3.
Comparison to Payne’s case alone cannot support a finding of disproportionality. There are, however, a number of defendants similarly situated to Chew, some even more culpable, who received life sentences. Comparison to these cases, a number of which are excluded from the Court’s review, demonstrates that Chew’s sentence is arbitrary and, therefore, both inhumane and unconstitutional, even if by the Court’s limited proportionality review analysis it is not clearly disproportionate.
When the Court undertakes to compare Chew to “the only other killer in the 1-3 category,” ante at 215, 731 A.2d at 1087, it blithely overlooks Walter Williams, a defendant most like Chew, excluded from the analysis because the jury failed to find the c(4)(d) factor. Williams is not the only defendant on record who killed a loved *271one or relative to collect insurance proceeds or an inheritance that is excluded from Chew’s review. Defendants Patrick Lanzel, classified as A-l, and Darrell Collins are left out of the Court’s comparison as well. What makes this acute class of offenders (Payne, Williams, Lanzel, Collins, and Chew) noteworthy is that, with the exception of Chew and Williams, all killed multiple victims, and, more remarkably, all with the exception of Chew were sentenced to life.8
Walter Williams, a thirty-six year old police officer, poisoned his first wife with potassium cyanide to cover up an affair that had developed into a bigamous marriage. Williams also expressly sought to receive the first wife’s estate. Williams had no prior record, and presented no clinical evidence of mental disturbance, but claimed to have flashbacks related to his experience in Vietnam. He was sentenced to life in prison with a thirty-year parole disqualifier.9
Patrick Lanzel conspired with his cousin to kill her father, his uncle, and to split the expected inheritance. After attempts to get the uncle to overdose on his own medication and to kill the uncle with rat poison failed, defendant beat his uncle to death with an iron rod. Defendant then went upstairs and murdered his aunt in the same way. He also beat a younger cousin, who survived. Defendant was eighteen at the time and ultimately confessed. He had partially completed college, had no criminal record, and had worked as a securities broker and a caddy. He did not have a *272history of mental disease, except that he claimed to have had suicidal tendencies in the months before the murder. Lanzel was sentenced to two consecutive fifty-year terms with thirty-year parole disqualifiers.
Darrell Collins stabbed his wife multiple times, cut her throat ear to ear, and suffocated his infant son. He had no history of mental illness or drug abuse, and denied having committed the murders. The sentencing judge recognized that Collins murdered his wife and child to collect the benefits of their life insurance policies, an aggregate of $105,000. The jury found Collins guilty of the non-capital murder of both victims. As with Williams, the jury did not find the c(4)(d) aggravating factor for pecuniary gain. Because no other aggravating factors were found, Collins’s case did not proceed to penalty trial. Collins was sentenced to one life term with a thirty-year parole disqualifier on one count, and a consecutive thirty-year term on the second.
Comparison of defendant’s case to these cases is troubling. Williams, Collins, and Lanzel all killed loved ones .or relatives to profit off an insurance policy or inheritance. All premeditated their murders. None presented evidence of clinical mental illness. And Lanzel and Collins killed more than one person. In short, Williams, Collins, and Lanzel were each more culpable than Chew, yet, like Celestine Payne, were not sentenced to death.
In addition, the crimes of Carl Norman, Kevin Smith and William Cooper, all classified in the robbery murder category, are too similar to Chew’s to be ignored. Carl Norman and Kevin Smith decided to rob a friend upon learning that he had inherited approximately $40,000. One week later, they invited the victim to Norman’s home under the pretext of watching a football game. Norman then informed Smith that they were going to kill the victim as well. Two hours after the victim arrived, Norman attacked, beating the head of the victim with a wooden statue. He and Smith then strangled the victim with an electrical cord and *273suffocated him with a plastic bag. They took $227 from the victim’s body. Norman had prior convictions for burglary, grand theft, robbery and aggravated assault. Smith had multiple prior convictions for resisting arrest, assault, drug possession and burglary. Neither had mental health problems. Norman pled guilty to felony murder and was sentenced to thirty-one years in prison with a thirty year parole disqualifier. Smith pled guilty to aggravated manslaughter and robbery, for which he was sentenced to consecutive terms of thirty years and ten years, respectively.
William Cooper, a construction worker, murdered a coworker, first by beating him with a sledgehammer and then by attempting to decapitate the victim with a circular saw and electric drill. Defendant’s apparent intent was to rob the victim of a $360 paycheck. Cooper had previously been convicted for weapons and assault offenses. He used drugs regularly, but claimed to be in good mental health. He was convicted of murder and was sentenced to life in prison with a minimum parole ineligibility of thirty years.
These crimes were committed to obtain some wealth that would not automatically follow from a robbery. Norman and Smith could not have expected that their victim would carry his $40,000 inheritance to a social event, and Cooper would have had to tender the victim’s cheek in order to receive any money. Accordingly, these robbery murders are kin to murders for insurance proceeds such as Chew’s. As in the cases of Williams, Collins, Lanzel and Payne, the murders were premeditated, and the defendants knew their victims. Further, the defendants did not present evidence of clinical mental illness; they did not even claim to be otherwise mentally or emotionally disturbed.
The bottom line is that Norman, Smith and Cooper committed murders with the same “essential attribute” as Chew’s and are relatively as culpable, yet were not sentenced to death — indeed, Norman and Smith pled guilty to less-than-life terms. Given this *274new perspective, and recalling that Williams, Payne, Lanzel and Collins were life-sentenced, Chew’s death sentence, which is not clearly disproportionate by frequency analysis or aberrational according to the Court’s manner of precedent-seeking review, appears arbitrary nevertheless.
When a death sentence can be theoretically proportionate and simultaneously arbitrary, we have as a matter of common sense noticeably strayed from the.purpose of our endeavor. We. ought then to reconsider the purpose of proportionality review. To say that a death sentence is proportionate evokes an infrastructure established by this Court to serve a larger purpose, answer a larger question, and identify a larger problem. Proportionality review, as we have stated, seeks to discern if similarly situated defendants generally receive sentences other than death. See Martini II, supra, 139 N.J. at 20, 651 A.2d 949. The purpose of this inquiry is to assure that “the death penalty is being administered in a rational, non-arbitrary, and even-handed manner, fairly and with reasonable consistency.” Marshall II, supra, 130 N.J. at 131, 613 A.2d 1059. It should follow naturally that an arbitrary sentence is also disproportionate. If our assessment of proportionality leads us to a different conclusion, one that conflicts with the larger issue, then that assessment must be wrong.
In such circumstances, we must not be confined by semantics or administrative rigor. Although we have accepted delineated restrictions for the sake of statistical comparison, we need not impose the same limitations in our review of precedent. “[I]t is crucial that we look beyond [the AOC’s comparison groups] in precedent seeking review in order to gain as much insight as possible into a defendant’s sentence.” See Harvey III, supra, 159 N.J. at 360, 731 A.2d 1121 (Handler, J., dissenting). Disregarding classification, we should derive the most insight from those few cases that share most with defendant’s, maintaining a course strict in purpose but less rigid in scope.
*275In this case, the Court finds defendant’s death sentence proportionate. The statistics may so show, and Chew’s death sentence, when compared to a plethora of other defendants designated by the majority, may not stick out as an error. We ought nevertheless to be disturbed by the fact that death sentences were not imposed upon those defendants whose crimes and circumstances most closely mirrored Chew’s. The fact that the defendants who committed the most similar and more culpable crimes in very like circumstances received life sentences or less is sufficient evidence upon which to conclude that the death penalty has been arbitrarily imposed.
IV
In conclusion, we may recall that Chew’s death-eligibility rested on the improper application of the c(4)(d) aggravating factor and note that Chew is the only defendant on death row who murdered for insurance proceeds. Far from heralding that it is somehow less culpable to kill for insurance proceeds than to hire someone to do it for you or to do it for someone else,10 I merely protest that the fine distinction between these acts, so recognized by the Legislature, see discussion supra at 192-96, 731 A.2d at 1074-76, can be the difference between life and death. By broadening the scope of precedent-seeking review in this case, the Court might recognize that Chew’s sentence, if not disproportionate, is arbitrary. Matters of life and death may well be arbitrary in the natural world; they should not be in a court of law.
Chew’s death sentence should be vacated.
*276[[Image here]]
*277For affirmance — Chief Justice PORITZ and Justices POLLOCK, O’HERN, GARIBALDI, STEIN and COLEMAN — 6.
For reversal — Justice HANDLER — 1.
Walter Williams was the only defendant classified in the 1-3 subcategoiy for "other pecuniary motive" killers at the time of DiFrisco’s proportionality review. See DiFrisco III, supra, 142 N.J. at 167 n. 3, 662 A.2d 442. While dismissing Williams, the Court allowed Michael Rose, classified as an 1-1 defendant, to remain in the group despite the fact that Rose's sentencing jury also failed to find the c(4)(d) aggravating factor.
By the same logic upon which this Court excludes Williams, the Court also excludes Michaél Rose, whose jury rejected the c(4)(d) aggravating factor as well. This Court included Rose as a comparison case in DiFrisco III, supra, 142 N.I. at 170-71, 662 A.2d 442. I would not exclude Rose from the comparison group here.
Five of the cases suggested for comparison by defendant involved multiple victims, and were therefore AOC-classified in the A category: Patrick Lanzel (A-1); Frank Masini 2(A-1); Ronald Mazique (A-l); Roy Watson (A-l); and Bobby Lee Brown (A-3). See CCH Report, tbl. 7. Two of the cases were classified by the AOC in the B category because the defendants had prior murder convictions: John Fauntenberry (B-l) and Richard Feaster 2(B-1). See ibid. One of the proposed comparison cases, Robert Zeuner, was classified by the AOC as G-2 for residential burglary with forced entry and without particular violence or terror. See ibid. The remaining twenty-six cases suggested by defendant involved robbery murders and were AOC-classified in the E category: Dwayne Caviness (E-l); Albert Fains (E-l); Carlton Felder (E-l); Matthew Ploppert (E-l); Anthony Szadorski (E-l); Wanda Campbell (E-2); James Fitzpatrick (E-2); Darren Grant (E-2); Michael Jones (E-2); Ira Musgrove (E-2); Richard Redden (E-2); Abdel Saleh (E-2); George Shaffer (E-2); Larry Durden (E — 3); Emmanuel Charles (E-5); Richard Feaster l(E-5); David Mark Russo (E-5); Corey Washington (E-5); Charles Williams (E-5); Richard Cain (E-6); George Lazorisak (E-6); Vernon Mclver (E-6); Carl Norman (E-6); Kevin Smith (E-6); Ronald Pierce (E-6); and Mark Robinson (E-6). See ibid.
There are twenty-one death-eligible defendants classified in the E-6 category for murders involving robbery between acquaintances: Jhi-Mar Anderson (sentenced to 20 years in prison); Richard Cain (life); Reginald Clark (20 years); William Cooper (life); Larry Daniels (life); Salvatori Ferrari (life); Clifford Graf, Jr. (life); Michael Grant (life); Louis Harris (15 years); Frederick Hedgespeth (life); Richard Jefferson (life); Nathaniel Johnson (life); George Jones (30 years); Bruce King (30 years); George Lazorisak (life); Vernon Mclver (life); Carl Norman (31 years); Ronald Pierce (life); Mark Robinson (27 years); Kevin Smith (30 years); and Roy Sullivan (life). CCH Report, tbl. 7A; see id., Narratives. Only two of these defendants, George Lazorisak and Ronald Pierce, reached a penalty trial. See id., Narratives.
Of the E-6 homicides, many involved circumstances that are distinguishable from Chew's homicide offense. Some of the murders resulted from defendants' briefly premeditated decisions to take possessions or money from people with whom they were minimally acquainted. See, e.g., Larry Daniels (murdered boy he knew from neighborhood to get victim’s sneakers and sweatpants); George Lazorisak (murdered man he picked up same night at nightclub primarily to obtain victim's car); Vernon Mclver (male prostitute murdered client met night before for victim’s money and car); and Ronald Pierce (slit throat of man he and codefendant met same night at bar seeking victim’s cocaine). Others are heat-of-the-moment murders of friends or relatives who refused to give money to the defendant. See, e.g., Reginald Clark (murdered aunt after she refused to give him $20); Salvatore Ferrari (murdered mother when she refused to give him money); Louis Harris (murdered drinking acquaintance who refused to give him money); Frederick Hedgespeth (murdered acquaintance who refused to give $20); Nathaniel Johnson (stabbed grandmother twice in chest during argument over money); Roy Sullivan (killed elderly friend who refused to give him money to buy drugs). Some of the crimes involved relationships which were predominantly drug-related. See, e.g., Jhi-Mar Anderson; Michael Grant; Richard Jefferson; and Bruce King.
I refer only to those statistics that exclude defendant from the comparison group. As I have noted previously, “gauging the proportionality of a defendant’s sentence by comparing it to a group of which he is a member, skews the analysis.” Loftin II, supra, 157 N.J. at 420 n. 19, 724 A.2d 129 (Handler, J., dissenting).
The Court continues indiscriminately to compare defendants with those whose sentences were the subject of prior proportionality review. See ante at 207-10, 731 A.2d at 1083-84. I would restrict comparisons between Chew and prior proportionality review defendants to circumstances in which they share culpability levels. See Loftin II, supra, 157 N.J. at 423-25, 724 A.2d 129 (Handler, J., dissenting). Because culpability levels are not necessarily synonymous across databases, see ibid.; see also Special Mater Report, supra, at 93-94 (discussing and charting "instability” of culpability scores across cases), in undertaking such a limited comparison, Chew's numbers would properly be compared with numbers for prior proportionality review defendants gleaned from the same regression.
However, for the purposes for which the Court uses the index-of-outcome test results — namely, to conduct a comparison of Chew's culpability ratings with those of prior proportionality review defendants regardless of culpability level— comparison based upon the data this Court used in a given defendant's proportionality review must would provide a more accurate measure. For example, the data used to measure Martini's culpability must be derived from the Martini Report. Therefore, although I agree with the Court's choice to apply the index-*266of-outcome regression results from the present database, I do not agree with its reason for doing so or the comparison for which they use the data.
The claim that Chew "planned" the murder for over a year, see ante at 217-18, 731 A.2d at 1089, is an overstatement. The most that can confidently be drawn from the record in this case is that Chew had considered killing Bowman for over a year.
In Marshall II, supra, the Court compared Marshall to Walter Williams, Darrell Collins and Thomas Johnston, all of whom killed their wives in a premeditated fashion. 130 N.J. at 175-76, 178-79, 613 A.2d 1059. The Court included the three defendants in Marshall's precedent-seeking review, not as members of Marshall’s comparison category, the I class, but as "cases involving highly-premeditated, cold-blooded murders of a defenseless wife” or as "first-time murderers of spouses.” See id. at 170, 175, 613 A.2d 1059.
The fact summaries are derived from the AOC Narratives provided in the CCH Report.
To make a determination of whether it is more humane, impassioned, or provoked to kill by your own hand or more cold, sinister, and removed to hire someone else to do it are the types of moral conundrums that place proportionality review precariously close to a resentencing.