The opinion of the Court was delivered by
POLLOCK, J.In unrelated incidents, defendant, Marko Bey, sexually assaulted and murdered two women. Separate juries sentenced defendant to death for each of the murders. Initially we vacated both death sentences. In State v. Bey, 112 N.J. 45, 548 A.2d 846 (1988) (Bey I), which involved the murder of Cheryl Alston, we reversed the murder conviction and held that defendant was not death eligible because he was under the age of eighteen at the time of the murder. On remand, a jury found defendant guilty of purposeful murder and aggravated sexual assault. The trial court sentenced him to an aggregate sentence of life imprisonment plus twenty years, with no parole eligibility for forty years. In State v. Bey, 112 N.J. 123, 548 A.2d 887 (1988) (Bey II), decided the same day as Bey I, we affirmed defendant’s conviction for the murder of Carol Peniston. Because of an incorrect jury charge, however, we reversed the death sentence and remanded the matter for re-sentencing. Once again, the jury returned a death sentence for the Peniston murder, which we affirmed in State v. Bey, 129 N.J. 557, 610 A.2d 814 (1992) (Bey III). In Bey III, we deferred proportionality review of that sentence pending receipt of a more complete record. We now find no disproportionality in the imposition of the death sentence for defendant’s second murder.
TABLE OF CONTENTS
Page
I. Facts....................................................340
II. Proportionality Review....................................343
A The Universe of Cases ................................343
B. Method of Classifying Cases ...........................345
*340III. Comparison of Cases.....................................350
A. The Frequency Approach..............................350
1. The Salient-Factors Test ...........................353
2. The Numerical-Preponderance-of-Aggravating-and-
Mitigating-Factors Test...........................358
3. The Index-of-Outcomes Test........................362
B. The Precedent-Seeking Approach.......................366
1. Relevant Factors...................................366
2. Comparison of Marko Bey’s Case to Similar Cases ... 369
a. The Cases......................................369
b. The Comparison.................................382
3. Other Cases........... 387
IV. Race as an Impermissible Factor..........................388
V. Conclusion...............................................396
-I-
FACTS
The facts surrounding the murder of Carol Peniston are set forth in Bey II, supra, 112 N.J. at 131-33, 548 A.2d 887, and Bey III, supra, 129 N.J. at 568-69, 610 A.2d 814. We therefore include only a brief summary.
On April 26,1983, around 9:20 p.m., Carol Peniston left Neptune High School, where she had attended a computer course, and had driven away in her car. Approximately four hours later, the car was involved in a one-car accident in Newark. Defendant’s fingerprints were on the rearview mirror. Ms. Peniston, who had been divorced and lived alone, neither returned to her apartment nor reported to work the next day.
On May 3, Asbury Park police discovered Ms. Peniston’s body in a shed near an industrial building. An autopsy performed on May 4 disclosed that she had been dead for several days. The autopsy further disclosed that she had been beaten, sexually assaulted, and strangled. From a sneaker imprint on her chest and from evidence of fractured ribs and hemorrhaging of the right lung, vertebral column, and right atrium of the heart, the Mon*341mouth County medical examiner concluded that Ms. Peniston’s assailant had stomped on her chest. The ultimate cause of her death, however, was ligature strangulation. Subsequent police investigation revealed that the characteristics of spermatozoa found on the victim’s coat were consistent with those of defendant’s saliva, and that defendant’s sneakers bore an imprint that was similar to the impression on the victim’s chest.
On May 6, defendant, who had turned eighteen only three weeks earlier, was arrested for receiving stolen property, Ms. Peniston’s car. After five hours in police custody, defendant confessed to the murder.
Defendant then gave a written statement, in which he admitted that he had accosted Ms. Peniston in front of her apartment building and demanded money from her. The statement continued that when defendant heard someone coming, he grabbed her and led her to the shed. In the ensuing events, he repeatedly struck Ms. Peniston, sexually assaulted her, and took eight dollars, as well as the car keys, from her pocketbook. While on his way to Newark in her car, he had an accident and abandoned the car.
A jury convicted defendant of capital murder and sentenced him to death. The sentence followed from the jury’s finding of two aggravating factors: the murder had “involved torture, depravity of mind, or an aggravated assault to the victim,” N.J.S.A 2C:11-3c(4)(c) (the c(4)(c) factor), and it had been committed in the course of a felony, N.J.S.A. 2C:ll-3c(4)(g) (the c(4)(g) factor). The jury found no mitigating factors. We affirmed the conviction, but reversed the death sentence, primarily because the court had incorrectly charged the jury on the mitigating factors. Bey II, supra, 112 N.J. at 156-64, 166-71, 548 A.2d 887.
On the same day that we reversed and remanded Bey’s death sentence for the murder of Carol Peniston, we also vacated his conviction and death sentence for the prior murder and sexual assault of Cheryl Alston. Bey I, supra, 112 N.J. at 51, 548 A.2d 846. In that decision, we held that defendant was not death *342eligible because he had committed the Alston murder before reaching the age of eighteen. Ibid. On re-trial for the Alston murder, the jury found defendant guilty of purposeful murder and aggravated sexual assault. He received an aggregate sentence of life imprisonment plus twenty years, with forty years of parole ineligibility. The Appellate Division affirmed the conviction, 258 N.J.Super. 451, 610 A.2d 403, and we denied certification, 130 N.J. 19, 611 A.2d 657 (1992).
At the re-sentencing trial for the Peniston murder, the State proffered two aggravating factors: defendant previously had been convicted of a murder, that of Cheryl Alston, N.J.S.A. 2C:11-3c(4)(a) (the c(4)(a) factor), and the Peniston murder had occurred during a sexual assault and robbery, the c(4)(g) factor. Defendant did not contest these aggravating factors, but argued that four mitigating factors outweighed them: “defendant was under the influence of extreme mental or emotional disturbance,” N.J.S.A. 2C:ll-3c(5)(a) (the c(5)(a) factor); defendant’s age at the time of the murder, N.J.S.A. 2C:22-3c(5)(c) (the c(5)(c) factor); “defendant’s capacity to appreciate the wrongfulness of his conduct or to conform his conduct to the requirements of the law was significantly impaired as the result of mental disease or defect or intoxication,” N.J.S.A. 2C:ll-3c(5)(d) (the c(5)(d) factor); and the catch-all factor — “[a]ny other factor which is relevant to the defendant’s character or record or the circumstances of the offense,” N.J.S.A. 2C:ll-3c(5)(h) (the e(5)(h) factor).
The jury unanimously found both aggravating factors. Two jurors found extreme mental or emotional disturbance, c(5)(a), and six jurors found the catch-all factor, c(5)(h). None of the jurors found that either defendant’s age, c(5)(c), or the significant impairment of his moral faculties, c(5)(d), was a mitigating factor. Furthermore, the jury found beyond a reasonable doubt that the two aggravating factors outweighed the two mitigating factors. Consequently, the court sentenced defendant to death. Bey III, supra, 129 N.J. at 576, 610 A.2d 814.
*343-II-
PROPORTIONALITY REVIEW
N.J.S.A. 2C:ll-3e, a section of the Capital Punishment Act (the Act), requires a proportionality. review on a defendant’s request to determine whether the death sentence, considering both the crime and the defendant, is disproportionate to the penalty imposed in similar cases. L. 1985, c. 478. In general, the death penalty must be imposed fairly and with reasonable consistency. The test of disproportionality is 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.’ ” State v. Marshall, 130 N.J. 109, 131, 613 A.2d 1059 (1992) (citing Tichnell v. State, 297 Md. 432, 468 A.2d 1, 17 n. 18 (1983)). Thus, a death sentence is valid unless the defendant establishes that similar defendants who commit factually-similar offenses generally receive sentences other than death.
Before reviewing the proportionality of defendant’s sentence, we first must answer preliminary questions regarding the universe of eases and the method of classifying those cases.
-A-
THE UNIVERSE OF CASES
In Marshall, we defined the relevant universe of cases. 130 N.J. at 137, 613 A.2d 1059. Preliminarily, we must decide whether to follow the May 12, 1992, amendment to the Act, which limits proportionality review to a comparison of similar cases in which the sentence of death actually has been imposed, L. 1992, c. 5, or whether we should continue to consider all cases that are death eligible, including those cases in which the State did not seek the death penalty.
Although the amendment was designed to také effect immediately, the Legislature did not indicate whether it should apply to *344pending appeals. If we were to apply the amendment to pending appeals, we would be obligated to consider whether it is unconstitutional as an ex post facto law. We decided in Marshall that because of the long pendency of that appeal, we would review the sentence under the prior law. 130 N.J. at 119, 613 A.2d 1059. Because we rejected Marshall’s proportionality challenge under the prior law, the amendment would not have affected the outcome in that case. Ibid. We come to the same conclusions here.
Defendant murdered Ms. Peniston on April 26, 1983, sixteen months before the murder of Robert Marshall’s wife. Defendant’s appeal has been pending since his initial death sentence on September 28, 1984, more than eight years before the effective date of the amendment. As in Marshall, we reject defendant’s proportionality challenge under the old law. For these reasons, we decline to address the constitutionality of the amendment. We shall apply the statute in its pre-amendment form. Thus, as in Marshall, the relevant universe of cases consists of those that are death eligible, even if they were not prosecuted as capital cases.
• Marshall summarizes the procedure for identifying the universe of cases. Id. at 137-41, 613 A.2d 1059. Since the Marshall decision, the Administrative Office of the Courts (AOC) has assumed the responsibility from Special Master David C. Baldus for maintaining the data base of cases. In compiling the statistics, the AOC has followed the Special Master’s procedure, as modified by our opinion in Marshall. The universe of cases for Bey consists of 266 death-eligible homicides committed from 1983 to 1992, 117 of which proceeded to the penalty phase. After oral argument, we granted defendant’s motion to supplement the record with data that had been compiled since March 25, 1993, the date of the last revision of the appendices and tables for Bey’s proportionality review (the Bey Report). This information, which constitutes the universe of cases compiled through June 25, 1993, for the pending proportionality review of John Martini (the Martini Report), increases the relevant universe of cases to 298 death-eligible offenses, 125 of which proceeded to the penalty-trial phase. *345Our consideration of these data in this case will not affect any argument proffered by Martini in his proportionality review.
-B-
METHOD OF CLASSIFYING CASES
Having determined the universe of cases, we next convert that universe into the data base for comparison purposes. In Marshall, we analyzed the cases in two ways. The first method followed an a priori, or clinical, approach, in which we analyzed the cases according to features that experience has shown probably influenced the life/death decision. Id. at 141—42, 144, 613 A.2d 1059. The second approach was an empirical one: we analyzed the cases according to characteristics that best explained the sentence actually imposed. Id. at 142-43, 144, 613 A.2d 1059. Following the Special Master’s methods, we “[took] advantage of the available data to sort out the cases on the basis of the characteristics that both prosecutors in the charging process and juries in the deliberative process deem most relevant.” Id. at 143, 613 A.2d 1059.
The coding of variables in the companion cases continues to be a source of contention between the Public Defender and the Attorney General. In Marshall, we recognized their differences and urged them to cooperate in developing a data base. Id. at 216-18, 613 A2d 1059. Since then, the AOC has conducted meetings to resolve issues concerning the standards for each coded characteristic. Many issues have been resolved, and the characteristics have been reduced to statistical codes. We recognize, however, that the codes inevitably incorporate subjective determinations. Implicit in the seemingly-objective review of statistics lies an unavoidably-subjective view of deathworthiness.
Remaining are some issues first raised in Marshall and raised again here. One issue questions the reliability of the coding of the thirty-four cases that remain coded as death-sentenced although reversed for various errors, such as improper jury instructions. *346Specifically, defendant identifies errors in all seven of the cases involving prior murder convictions (State v. Biegenwald, 106 N.J. 13, 53, 524 A.2d 130 (1987) (Biegenwald IA); State v. Biegenwald, 126 N.J. 1, 8, 594 A.2d 172 (1991) (Biegenwald IB); State v. Coyle, 119 N.J. 194, 218-20, 220-21, 229-32, 574 A.2d 951 (1990); State v. Erazo, 126 N.J. 112, 128, 594 A.2d 232 (1991); State v. Pennington, 119 N.J. 547, 565, 575 A.2d 816 (1990); State v. Purnell, 126 N.J. 518, 523, 601 A.2d 175 (1992); and State v. Ramseur, 106 N.J. 123, 312-13, 524 A.2d 188 (1987)). On remand, these cases either were not pursued to the penalty-trial phase or resulted in life sentences. Defendant argues that the initial death sentences in these cases, therefore, are not proper indicators of deathworthiness.
If we exclude these seven prior-murder-conviction cases from the pool of death-sentenced cases, Bey would remain as the only prior murderer whose death sentence we affirmed. That fact alone would not'compel a finding that Bey’s death sentence is disproportionate. In Marshall, we faced a similar situation. We stated that “simply because Marshall may be the first [contract-killer to receive an affirmed death sentence] does not mean that his death will be disproportionate under our statute.” 130 N.J. at 166, 613 A.2d 1059. After reviewing the frequency data in Marshall, we concluded that “ ‘[although lesser sentences than death are frequently imposed in domestic murder cases, it does not follow that the death penalty would not be authorized for the murder of one spouse by another under any circumstances.’ ” Id. at 174, 613 A.2d 1059 (quoting Tyler v. State, 247 Ga. 119, 274 S.E.2d 549, 555 (1981)). The circumstances in Marshall were that the defendant hired another to kill his wife so that he could collect life-insurance benefits on her life. “[T]he data show[ed] that among those for whom death is a fitting punishment, contract killers, whether principal or agent, are among the more frequent recipients of the death sentence.” Id. at 166-67, 613 A.2d 1059. Similarly, if Bey were to remain as the only defendant who had *347been previously convicted of murder and whose death sentence was affirmed, that sentence need not be disproportionate.
Furthermore, we decline to follow defendant’s suggestion to re-code as life-sentenced cases those death-sentenced cases in which the sentence was reversed. The AOC continues to code these cases as death-sentenced cases. In Marshall, we stated that “[w]e believe, ... as does the [Special] Master, that the original penalty trials, although reversed for various reasons, most often for the burden-of-proof and Gerald issues, have reflected juror values of deathworthiness in terms of deterrent effect.” Id. at 194 n. 10, 613 A.2d 1059. The phrase “Gerald issues” derives from our holding in State v. Gerald “that a defendant who is convicted of purposely or knowingly causing ‘serious bodily injury resulting in death’ ... as opposed to one who is convicted of purposely or knowingly causing death ... may not be subjected to the death penalty.” 113 N.J. 40, 69, 549 A.2d 792 (1988). Marshall recognized that such issues pertaining to procedural fairness, as distinguished from those that affect the substance of the crime, do not necessarily bear on the jury’s determination of deathworthiness. 130 N.J. at 169 n. 5, 194 n. 10, 613 A.2d 1059.
For example, in defendant’s first trial for the sexual assault and murder of Cheryl Alston, the jury returned a sentence of death. We reversed because of a statutory mandate, L. 1985, c. 478 (codified at N.J.S.A. 2C:ll-3g), that a defendant younger than eighteen-years old could not receive a capital sentence. Bey I, supra, 112 N.J. at 95-105, 548 A.2d 846. Consequently, Bey, who was only ten days short of his eighteenth birthday when he sexually assaulted and murdered Cheryl Alston, received a life sentence. Our reversal, however, does not detract from the initial jury’s view that defendant deserved the death penalty for the Alston murder. Defendant does not explain why some errors that have caused us to reverse a death sentence necessarily reflect on the jury’s ability to assess a defendant’s deathworthiness. In the absence of an acceptable explanation, we continue to believe that a *348death sentence, even when reversed, represents a societal consensus concerning the deathworthiness of a defendant.
Moreover, the reasons for the State’s failure to pursue capital sentencing a second time or for the imposition of a life sentence at a second penalty-phase trial are varied and indeterminable. We cannot conclude that in any given case a life sentence resulted from the view that the defendant was not initially deathworthy, rather than, for example, from the strength of the prosecutor’s case, including the availability of witnesses, or the adequacy of the State’s resources.
We therefore treat as death-sentenced a case that initially resulted in a death sentence but that was reversed. As we stated in Marshall, “[w]e have been candid to acknowledge that there is no scientific infallibility in the frequency data that we cite.” 130 N.J. at 169 n. 5, 613 A.2d 1059. Indeed, as stated above, all coding decisions necessarily rely on subjective determinations of deathworthiness that may not be completely accurate representations of death-sentencing decisions of jurors or prosecutors. Supra at 345, 645 A.2d at 691. We rely, as we did in Marshall, on “what we know,” 130 N.J. at 169 n. 5, 613 A.2d 1059; of the 117 death-eligible cases proceeding to the penalty phase, thirty-four cases received the death penalty.
Our dissenting colleague urges, post at 403, 645 A.2d at 719, as he did in Marshall, 130 N.J. at 249, 253-57, 613 A.2d 1059 (Handler, J., dissenting), that reversed cases are not valid indicators of deathworthiness. We continue to believe, however, as we did in Marshall, that cases in which prosecutors seek and juries impose the death penalty reflect the conscience of the community on the propriety of the imposition of that penalty. A reversal does not necessarily erase “the complex nature of the jury’s deliberation in the penalty-phase.” Post at 416, 645 A.2d at 725. We acknowledge that a reversed death penalty is a less persuasive indicator of deathworthiness than one that is affirmed, but we continue to believe that even reversed death sentences are suffi*349ciently valid indicators to remain for statistical purposes in the pool of death-sentenced cases.
We disagree also with our colleague’s suggestion, based on a recommendation of the Special Master, that we should adopt “a rebuttable presumption that reversed death sentences are invalid determinations of deathworthiness.” Post at 406, 645 A.2d at 720. The suggestion is reminiscent of his statement in his Marshall dissent, “I believe that a death sentence is disproportionate unless defendants with similar characteristics generally receive death sentences for committing factually similar offenses.” 130 N.J. at 248, 613 A.2d 1059. Implicit in both statements is the proposition that the State bears the burden of establishing the proportionality of a death sentence. We believe, however, that once this Court has sustained a death sentence on direct appeal, the defendant should bear the burden of proving disproportionality. Indeed, the language of the Act indicates that the Legislature intended that the defendant should bear that burden. N.J.S.A. 2C:ll-3e provides that “the Supreme Court shall determine whether the sentence is disproportionate to the penalty imposed in similar cases, considering both the crime and the defendant.” The use of the word “disproportionate,” rather than “proportionate,” signifies the legislative intention that we should search not for proof that a defendant’s death sentence is perfectly symmetrical with other death sentences, but for proof that the sentence is an outlier.
Consistent with the dissent’s underlying premise that no defendant should ever receive the death penalty, the dissent argues that no defendant should be the first to die. Hence, the dissent attacks as “hypothetical” our conclusion that even if we were to exclude reversed death-sentenced cases, leaving Bey as “the only prior murderer finally sentenced to death,” his death sentence would not be disproportionate. Post at 406, 645 A.2d at 721. As we said in Marshall, in which our dissenting colleague raised the same argument, 130 N.J. at 267-68, 613 A.2d 1059, simply because a defendant “may be the first does not mean that his death will be *350disproportionate under our statute,” id. at 166, 613 A.2d 1059. The grim fact is that some defendant must be the first to receive the death penalty.
-III-
COMPARISON OF CASES
After the Court has determined the universe of cases and the criteria for coding those cases, the third step of proportionality review is to group cases according to similarities relevant to the determination of deathworthiness. In Marshall, we selected measures of blameworthiness, or culpability, based on our consideration of both statutory aggravating and mitigating factors and nonstatutory factors based on “objectively-verified measures of blameworthiness.” Id. at 145, 613 A.2d 1059.
We then evaluated these factors in two ways:' the frequency analysis and the precedent-seeking analysis. The frequency analysis computes the frequency of death sentences within a pool of similar cases. It depends on a statistical analysis that measures the societal consensus that death is the appropriate penalty in the measured cases. See David C. Baldus, Death Penalty Proportionality Review Project, Final Report to the New Jersey Supreme Court 27 (Sept. 24, 1991) (Final Report). The precedent-seeking analysis is more intuitive, comparing a defendant’s deathworthiness with that of defendants in factually-similar cases. Id. at 30-31. Combining these two analyses helps to ensure the reliability of our evaluation of the proportionality of a defendant’s death sentence. The pool of cases remains small. As that pool expands, we can rely more heavily on the frequency analysis. For the time being, we are forced to rely more heavily on the precedent-seeking analysis.
-A-
THE FREQUENCY APPROACH
The frequency analysis consists of three different methods of assessing criminal culpability: the salient-factors test, the nu*351merical-preponderance-of-aggravating-and-mitigating-factors test, and the index-of-outcomes test. Marshall, supra, 130 N.J. at 154, 613 A.2d 1059. These tests are statistical analyses that assess the criminal culpability of a defendant when compared to other defendants. Because the frequency approach is a form of statistical analysis, our discussion is necessarily steeped in the underlying data.
Generally speaking, statistical results become more reliable as the data sample increases and the correlation grows between two variables. In Marshall we stated that “[t]he higher the frequency of a death sentence among the comparison group of ‘similar cases,’ the more certain the determination that the sentence is proportionate. The lower the frequency, the more strictly the Court must scrutinize the case for the possible influence of impermissible factors.” Id. at 153, 613 A.2d 1059. As a general rule, “ ‘[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.’ ” Id. at 153-54, 613 A.2d 1059 (quoting Tichnell, supra, 468 A.2d at 17 n. 18). “Generally,” however, does not require a threshold rate over fifty percent. Id. at 152-54,167, 613 A.2d 1059. Even if the frequency were less than fifty percent, it could serve as evidence of reliability of the sentence, particularly if confirmed by the precedent-seeking analysis. Id. at 154, 167, 613 A.2d 1059.
As in Marshall, 130 N.J. at 265-67, 613 A.2d 1059, Justice Handler urges, post at 408, 645 A.2d at 722, that we set a more specific standard in the frequency analysis than that of general comparability with other death sentences. A general standard, although admittedly imprecise, is not necessarily arbitrary. Indeed, a standard that applies generally is the antithesis of one that applies arbitrarily. Hence, as in Marshall, 130 N.J. at 152-54, 613 A.2d 1059, we decline to define more specifically the standard for defining an acceptable frequency for the imposition of the death penalty.
*352Proportionality review seeks to determine only whether a particular death sentence is aberrational, not whether it compares perfectly with other sentences. Id. at 131, 613 A.2d 1059. Not every statistical disparity establishes disproportionality. After conducting all three tests, we conclude that defendant’s death sentence is neither random nor aberrational.
At first glance, defendant’s death sentence might seem disproportionate. In the Bey Report, forty-four percent (117/266) of the death-eligible cases in the universe proceeded to the penalty-trial phase, but only twenty-nine percent (34/117) of those cases resulted in a death sentence. When the Martini data are added, the ratios remain approximately the same: forty-two percent (125/298) of the death-eligible cases proceeded to the penalty phase and thirty percent (38/125) of those cases resulted in a death sentence.
The more significant basis of comparison is not all death-sentenced cases, but only those with similar characteristics relevant to the sentencing decision between life and death. Marshall, for example, was the only defendant sentenced to death whose death sentence was affirmed among all 227 death-eligible eases and 113 penalty-trial cases as of September 24, 1991. 130 N.J. at 166, 613 A.2d 1059. Yet, we found his death sentence not to be disproportionate. Id. at 174, 613 A.2d 1059. Marshall’s status as one who hired a contract-killer put him in a category of cases with defendants who more likely than not received a death sentence. Id. at 166-67, 613 A.2d 1059. So here, the data show that of all deathworthy defendants those, like defendant, with a prior murder conviction, more frequently receive the death penalty.
Preliminarily, defendant urges that we should not include him in the study because to do so would be to compare his case to himself. In Marshall, we recognized good reasons for both including and excluding a defendant’s case from review. Thus, we decided to review the statistics under both alternatives. Id. at 167-68, 613 A.2d 1059. Here, we use the same approach.
*3531. THE SALIENT-FACTORS TEST
The salient-factors test, which compares sentences in cases that are factually similar, is the most persuasive test. Id. at 168, 613 A2d 1059. Its methodology is simple: the test measures the frequency of death sentences in similar cases. Ibid. In Bey’s case, the most salient factor is that he had been convicted of a prior murder. We believe that prosecutors and juries would find a previously-convicted murderer to be more blameworthy than a first-time offender.
Among the seventeen cases in the Bey Report in which defendants had been convicted of a prior murder, seventy-five percent (9/12) of these defendants reaching the penalty-trial phase received the death penalty, and fifty-three percent (9/17) of all death-eligible defendants received the death penalty. The figures are:
Penalty Trial Death Eligible
Including Bey ' .75 (9/12) .53 (9/17)
Excluding Bey .73 (8/11) .50 (8/16)
The Martini Report, which compiles data for John Martini’s proportionality review, adds three death-eligible cases, all of which proceeded to the penalty phase. When these additional cases are considered, the ratios remain high: including Bey, sixty percent (9/15) of the defendants with a prior murder conviction who reached the penalty phase received the death penalty, and forty-five percent (9/20) of all such defendants who were death eligible received the death penalty.
The death-sentencing rate for defendants with prior murder convictions in both the Bey Report and the Martini Report exceeds the death-sentencing rate for contract-killer principals such as Marshall. Id. at 168, 613 A.2d 1059. In Marshall, we found significant a thirty-three-percent death-sentencing rate among penalty-trial cases and a twenty-five-percent death-sentencing rate among death-eligible cases. Id. at 169, 613 A.2d *3541059. By comparison, the death-sentencing rate for cases most similar to defendant’s case illustrates a higher correlation between a prior murder conviction and a death sentence.
These figures illustrate an even-higher correlation when the pool is narrowed to include cases more factually comparable to defendant’s case. A significant factor in Bey’s case is that in addition to being a two-time murderer, Bey committed his second murder during a sexual assault. Prosecutors and juries regard as highly blameworthy those defendants who have a prior murder conviction and whose current case involves either one additional aggravating circumstance or particular violence or terror (the violence/terror factor). In thirteen cases in the Bey Report involv: ing defendants who had been convicted of a prior murder, juries found one additional aggravating circumstance or the violenee/terror factor. Among those cases, one-hundred percent (8/8) reaching the penalty-trial phase resulted in the death penalty, and sixty-two percent (8/13) of all death-eligible cases resulted in the death penalty. The figures are:
Penalty Trial Death Eligible
Including Bey 1.0 (8/8) .62 (8/13)
Excluding Bey 1.0 (7/7) .58 (7/12)
Again, these ratios remain high when we consider the relevant data from the Martini Report: eighty-nine percent (8/9) of all cases reaching the penalty phase, and fifty-seven percent (8/14) of all death-eligible cases in this category resulted in a death sentence.
Defendant disputes the validity of these results on several grounds. First, he argues that the cases most similar to his are not reliable indicators of deathworthiness because the sentences in those cases are fraught with procedural and other errors. The argument proceeds that if these questionable cases were excluded from the pool of death-sentenced cases and instead were coded as life-sentenced cases, the death-sentencing rate would be much *355lower. For reasons set forth above, however, supra at 345-349, 645 A.2d at 690-692, we shall continue to include them. Consequently, we shall continue to treat as death-sentenced cases those cases in which we have reversed the death sentence. Therefore, the cases that initially resulted in a death sentence should remain in the pool of factually-comparable cases. Marshall, supra, 130 N.J. at 169 n. 5, 194 n. 10, 613 A.2d 1059.
Second, defendant argues that his death sentence suffers from several procedural errors that affected the verdict. Generally speaking, the errors concerned jury selection and the admission of evidence. These “distorting factors,” defendant alleges, “inflated the frequency leading to” his death sentence. In Bey III, however, this Court held that it was “extremely unlikely [that the errors] had the capacity materially to affect the jury’s deliberations or produce an unjust result.” 129 N.J. at 616, 610 A.2d 814. For this reason, we believe that these “errors” do not impugn defendant’s death sentence.
Third, defendant asserts that other categories of factually-comparable cases do not demonstrate a high rate of death sentences. In particular, defendant points to the results of the sexual-assault and robbery cases. The sexual-assault pool in the Bey Report consists of thirty-five cases, none of which involved a prior murder conviction. The death-sentencing rate for the eighteen cases reaching the penalty-trial phase is twenty-eight percent (5/18), and fourteen percent (5/35) for all thirty-five death-eligible cases in the pool. If the analysis were limited, as defendant contends, to cases with the violence/terror factor, the death-sentencing rate increases slightly to thirty-six percent (5/14) of the penalty-trial cases and nineteen percent (5/26) of the death-eligible eases resulting in the death penalty. The figures are:
*356Penalty Trial Death Eligible
Sexual assault including Bey .32 (6/19) .17 (6/36)
Sexual assault excluding Bey .28 (5/18) .14 (5/35)
With violence including Bey .40 (6/15) .22 (6/27)
With violence excluding Bey .36 (5/14) .19 (5/26)
When the Martini data are added, the ratios remain approximately the same: including Bey, sixteen percent (7/44) of all death-eligible sexual-assault cáses received a death sentence, and thirty-five percent (7/20) of these cases proceeding to the penalty phase received the death sentence. When we narrow our focus to sexual-assault eases exhibiting the violence/terror factor, twenty-one percent (7/34) of all death-eligible cases and forty-four percent (7/16) of penalty-trial cases, including Bey, received a death sentence.
The robbery pool includes ninety cases. Like the cases in the sexual-assault pool, none of these cases involved a prior murder conviction. Consequently, the robbery pool does not include Bey. Thirty of the robbery-pool cases proceeded to the penalty phase. Among those cases, twenty percent (6/30) resulted in the death penalty; only seven percent (6/90) of all death-eligible cases in this category received the death penalty. As with the sexual-assault pool, the sub-group of cases in this category exhibiting the violence/terror factor does not significantly increase the death-sentencing rates for robberies. Of the thirty-four cases in this smaller pool, thirty-one percent (4/13) of the penalty-trial cases and twelve percent (4/34) of all death-eligible cases resulted in a death sentence. The figures are:
Penalty Trial Death Eligible
Robbery including Bey .23 (7/31) .08 (7/91)
Robbery excluding Bey .20 (6/30) .07 (6/90)
With violence including Bey .36 (5/14) .14 (5/35)
With violence excluding Bey .31 (4/13) .12 (4/34)
*357The Martini data do not significantly change the death-sentencing rate. Excluding Bey — because of his prior murder conviction — twenty-one percent (7/33) of all robbery cases that proceeded to the penalty-trial phase and seven percent (7/100) of all such cases, including death-eligible defendants, resulted in the death penalty.
Although we agree with defendant that the death-sentencing rates in the sexual-assault pool and the robbery pool are lower than the rate in the prior-murder-conviction pool, the difference is meaningless. Both these categories as defined in the Bey and Martini Reports exclude cases with prior murder convictions. Because Bey was convicted of the prior murder of Cheryl Alston, his case is not even included in the categories of cases in which defendants have committed only a sexual assault or robbery. As demonstrated above, a prior murder conviction is one of the most significant indicia of blameworthiness. In both Bey and Martini, sixty-four percent (9/14) of all death-eligible cases having two aggravating factors, one of which is a prior murder conviction, resulted in a death sentence. Therefore, to compare defendant’s case to cases involving a sexual assault or robbery, but not involving a prior murder conviction, is to disregard one of the most influential elements in death sentencing — the prior murder conviction.
As outlined above, moreover, a smaller pool of cases accounts for defendants with prior murder convictions whose crimes exhibit one additional aggravating factor or the violence/terror factor, such as murder during the course of a sexual assault or robbery. Cases with both characteristics are most like Bey’s case. Neither the Bey Report nor the Martini Report indicates in which of these cases the additional aggravating factor was a sexual assault or robbery, or both. Prosecutors, however, frequently seek the death penalty when prosecuting murders involving sexual assaults. Final Report, supra, at 81. We believe, therefore, that a jury *358would deem as highly blameworthy convicted prior murderers who commit a sexual assault in conjunction with a subsequent murder.
Under the salient-factors measure, the data do not show that defendants similar to Bey generally receive a sentence other than death. To the contrary, the data demonstrate that defendants like Bey, who have killed before and who kill again during a sexual assault, are highly blameworthy. Indeed, defendants having a prior murder conviction and an additional aggravating factor receive the death penalty sixty-two percent of the time. The imposition of the death penalty in sixty-two percent of all comparable death-eligible eases is strong evidence of the reliability of defendant’s death sentence.
2. THE NUMERICAL-PREPONDERANCE-OF-AGGRAVATING-AND-MITIGATING-FACTORS TEST
The numerical-preponderance test compares the subject case with cases having the same number of aggravating and mitigating factors. In addition to this purely quantitative analysis, the test also attempts to account for the qualitative value that juries place on certain aggravating and mitigating factors.
In defendant’s case, the jury found two aggravating and two mitigating factors. Concerning the aggravating factors, the jury found that defendant had been convicted of a prior murder and that he had murdered Ms. Peniston during a sexual assault and robbery. For mitigating factors, two jurors found that defendant suffered from extreme emotional disturbance, and six found that he was entitled to the catch-all factor. Unlike the application of the numerical-preponderance test in Marshall, which reflected an infrequency of death sentences for cases with two mitigating factors and only one aggravating factor, the application of that test to cases such as this one, with two mitigating factors and two aggravating factors, demonstrates a high frequency of death sentencing. Among the twenty penalty-trial cases in which jurors identified the relevant aggravating and mitigating circumstances, weighed them, and then returned a sentence, fifty-five percent *359(11/20) resulted in a death sentence. The Martini Report, which adds three cases, shows a slight increase in the death-penalty rate to fifty-seven percent (13/23).
Defendant argues that the frequency of death-sentencing rates for cases having two aggravating and two mitigating factors is low when all death-eligible cases are considered. Among the forty-three death-eligible cases in this category, only twenty-six percent (11/43) received the death penalty. These figures are summarized:
Penalty Trial Death Eligible
Including Bey .55 (11/20) .26 (11/43)
Excluding Bey .53 (10/19) .24 (10/42)
In the Martini Report, which includes five additional death-eligible cases, the rate is twenty-seven percent (13/48).
We agree that the probability of a death sentence, on considering all death-eligible cases with two aggravating and two mitigating factors, is comparatively low. The death-sentencing rate for all such cases, however, is much higher than the rate for cases similar to Marshall. The death-penalty rate for all death-eligible cases with one aggravating and two mitigating factors, as was the case in Marshall, was seven percent (3/44) including Marshall and five percent (2/43) excluding him. Notwithstanding those frequencies, we found that Marshall’s death sentence was proportionate because the payment-for-murder aggravating factor, N.J.S.A 2C:ll-3e(4)(e) (the c(4)(e) factor), produced an above-average death-sentencing rate. Id. at 172, 613 A.2d 1059. Similarly, when one of the aggravating circumstances is the c(4)(a) factor, a prior murder conviction, the death-sentencing rate is much higher for death-eligible cases in the category of cases having two aggravating and two mitigating factors. In both the Bey and the Martini Reports, seventy-one percent (5/7) of such cases resulted in the imposition of the death penalty. The numerical-preponderance analysis, therefore, does not indicate that defendant’s death sentence is disproportionate.
*360Defendant further argues that his case should be compared to cases with three, not two, mitigating factors. The additional mitigating factor that he claims is his age, because he was eighteen years old at the time he murdered Carol Peniston. No member of the jury, however, found age to be a mitigating factor.
According to defendant, the jury’s rejection of his age as a mitigating factor indicates that the verdict is irrational. Defendant argues that age is the most significant of the mitigating factors and is the factor most often found by a jury to be relevant in sentencing decisions. See Final Report, supra, at 92 (stating “defendant’s age (5c) has the greatest mitigating effect”). His argument is that in only seven percent (6/83) of all death-eligible cases and fifteen percent (6/41) of all eases proceeding to the penalty phase in which age was found to be a mitigating factor did the jury return the death penalty. The Martini data increased the rate slightly with sixteen percent (7/43) of all penalty-trial cases and eight percent (7/91) of all death-eligible cases resulting in the death penalty. Furthermore, because proportionality review includes non-statutory factors, defendant invites us to reconsider factors that the jury rejected or found less persuasive than others. We decline the invitation.
We held in Bey III that the jury had not erred in failing to find age as a mitigating factor. 129 N.J. at 613, 610 A.2d 814. The trial court had properly instructed the jury to consider both chronological age and psychological maturity at the time of the crime. Although a jury may not ignore a defendant’s youth, it need not find that age is relevant to his or her culpability. Ibid. We do not believe that all twelve members of the jury were acting irrationally when each declined to find defendant’s age to be a mitigating factor. Furthermore, although the jury did not find Bey’s youth relevant to the age factor, it may have considered his youth in conjunction with the catch-all factor.
Other juries confronting young defendants also have rejected age as a mitigating factor. For example, when first sentencing Phillip Dixon, who was eighteen at the time he beat, sexually *361contacted, and murdered a thirteen-year-old girl, the jury failed to find age to be a mitigating factor. State v. Dixon, 125 N.J. 223, 231, 593 A.2d 266 (1991); Detailed Narrative of Summaries for Death Eligible Cases 44 (Detailed Narrative Summaries). Dixon’s death sentence was reversed on appeal. In that case, as in Bey III, however, the jury’s failure to find age to be a mitigating factor was not a reason to reverse the death sentence. See Dixon, supra, 125 N.J. at 228, 593 A.2d 266; Bey III, supra, 129 N.J. at 613, 610 A.2d 814.
In sum, we are unpersuaded by defendant’s arguments. Like the sentencing jury in Marshall, the sentencing jury in Bey considered a number of aggravating and mitigating factors, but found only some. The mere fact that defendant was eighteen when he murdered Ms. Peniston does not mean that the jury must find his youth to be a mitigating factor as a matter of law. Our system contemplates that juries will reject some factors, including age. Although juries may find age to be a mitigating factor in many cases, they need not so find it in every case.
Moreover, our consideration of non-statutory factors does not entitle us to overrule the jury findings. We may not reject the jury’s findings even if we might disagree with them. Our role, as previously stated, is to search for aberrations that might be the result of impermissible factors. In the frequency analysis, we will consider only those factors that the jury found relevant to the imposition of the death penalty. By comparison, in the precedent-seeking analysis, we will expand our review to include objective factors that are clearly present in the record even if the jury did not find them to be relevant. We will not include these additional factors in the frequency analysis because of the need to maintain the uniformity of the statistics. Otherwise, we would be obliged to reconsider and recalculate the ratios for each case in the universe of cases. As with judicial review generally, we must recognize our limits in proportionality review.
The dissent argues here, as it did on direct review of Bey’s conviction, see 129 N.J. at 632-48, 610 A.2d 814, that the *362trial court’s exclusion of the report of one of the State’s experts, Dr. Cooke, and its refusal to permit leading questions of Bey’s mother, could not have been harmless error. Post at 412, 645 A.2d at 724. Further, the dissent asserts that harmless-error analysis has no place in death-penalty eases. Post at 414-418, 645 A.2d at 725-727. We continue to believe, however, as we did on direct appeal, that the asserted errors were harmless. 129 N.J. at 586-94, 610 A.2d 814. Our role in proportionality review is not to second-guess rulings that we made on direct appeal but to determine if the imposition of the death sentence on the defendant, when compared to sentences imposed on other defendants, is irrational or aberrant. At some point, even a death-penalty case must end.
3. THE INDEX-OF-OUTCOMES TEST
The index-of-outcomes approach seeks “to identify the characteristics common to the cases in terms of their degree of blameworthiness as perceived by prosecutors and juries.” Marshall, supra, 130 N.J. at 172, 613 A.2d 1059. It organizes cases according to statistically-relevant measures of culpability, such as the infliction of severe physical pain or mental suffering on the victim, a contemporaneous sexual assault or robbery, and the commission of a prior murder. In the data compiled for Robert Marshall’s proportionality review (Marshall Report), the Special Master “estimated for each offender the probability of receiving a death sentence____ On the basis of those predictions, [h]e created five level culpability scales which cut the cases at each 20-percentage points of increasing probability of a death sentence, i.e., 0-19, 20-39, etc.” Technical Appendix 9 at 5.
Bey scores high in blameworthiness. Using indices that include both statutory and non-statutory factors, we find that the predicted probability of a death sentence in his case is seventy-six percent among all penalty-trial cases, with a lower limit of thirteen percent and an upper limit of ninety-nine percent. Defendant’s case falls within culpability level four (60-80% culpability), which *363contains six other cases, and has an overall death-sentencing rate of forty-three percent (3/7). When we consider the Martini data, Bey’s predicted probability of receiving a death sentence increases to eighty-one percent, with a lower limit of thirty-five percent and an upper level of ninety-seven percent. Accordingly, Bey moves to level five (80-100%), the highest culpability level, which has a death-sentencing rate of eighty-eight percent (23/26).
Among all death-eligible cases, the predicted probability of a death sentence in defendant’s case is fifty-one percent, with a lower limit of nine percent and an upper limit of ninety-two percent. The seven eases most comparable to defendant’s in terms of blameworthiness fall into culpability level three (40-60% culpability), which has an overall death-sentencing rate of fifty-seven percent (4/7). Defendant’s predicted probability of receiving a death sentence in Martini is forty-seven percent, with a lower limit of ten percent and an upper limit of eighty-eight percent. At culpability level three, the death-sentencing rate is fifty percent (5/10).
When we consider only statutory factors, the predicted probability of a death sentence for Bey among all penalty-trial cases, within a range extending from fourteen to ninety-six percent, is sixty-seven percent. Eleven cases similar to defendant’s case fall within culpability level four (60-80% culpability). The overall death-sentencing rate for these cases is eighty-three percent (10/12). In Martini, Bey’s predicted probability of receiving a death sentence is sixty-two percent, with a lower limit of sixteen percent and an upper limit of ninety-four percent. At culpability level four, defendant’s level, the death-sentencing rate is sixty-seven percent (10/15).
Among all death-eligible eases, the predicted probability of a death sentence in defendant’s case is twenty-five percent, with a lower limit of seven percent and an upper limit of sixty-one percent. The comparison includes eighteen cases similar to defendant’s case. In culpability level two, defendant’s level, the overall death-sentencing rate is fifty-eight percent (11/19). In Martini, *364Bey’s predicted probability of receiving a death sentence is thirty-three percent, with a lower limit of ten percent and an upper limit of sixty-eight percent. This would place him in culpability level two, which has a death-sentencing rate of fifty-two percent (12/23).
We are constrained, as we were in Marshall, by the small sample of cases with the same level of blameworthiness as defendant’s case. As in Marshall, “ “we have a much less solid basis for saying that cases like his either will or will not be associated with frequent death sentencing over the long run.’ ” 130 N.J. at 173, 613 A.2d 1059 (quoting Marshall Report, supra, at 41).
To compensate for the dearth of cases in his culpability range, defendant, following a suggestion of the Special Master, has modified the culpability ranges. Instead of using five standard ranges of culpability of twenty percent each, defendant has altered the ranges so that more eases fall within the middle-range levels two through four. Defendant’s modified culpability ranges for all penalty-trial cases are:
Culpability Level Culpability Range Death Sentencing Rate
1 0-.0019 0% ( 0/33)
2 .0019-.012 0% ( 0/19)
3 .012-.145 5% ( 1/21)
4 .145-.89 43% (10/23)
5 .89-1 96% (28/29)
The inescapable problem with defendant’s modified culpability ranges is that they consist of dissimilar cases. In level four, defendant’s culpability level, the range of cases is vast: a case having an overall blameworthiness index of .145 is in the same comparison group as a case having a blameworthiness index of .89. Thus, level four includes cases in which defendants have significantly different levels of culpability. For example, defendant’s extended version of culpability level four includes Joseph Hicks, who has a culpability ratio of .15. Hicks shot his victim once in *365the head during a struggle over a sale of marijuana. Detailed Narrative Summaries, supra, at 126. Level four also includes Nicholas Muscio, who has a culpability ratio of .16. Muscio repeatedly stabbed a woman during the course of robbing her apartment. Id. at 208. Joseph Guagenti, another defendant included in this culpability level, has a culpability ratio of .18. He shot his ex-girlfriend at a bar where she was dancing and was spending time with a new boyfriend. Id. at 113. Finally, Raymond Kise has a culpability ratio of .20. He beat and then drowned a neighbor of his co-defendants because the victim called Kise’s girlfriend a “slut” while the group of men were drinking in the apartment of one of the co-defendants. Id. at 82.
Unlike Bey, none of these defendants had a prior murder conviction. Nor did they sexually assault their victims. Guagenti had been committed to a forensic psychiatric hospital for fifteen months for depression and had attempted suicide following the break-up of his relationship with his victim. Kise was intoxicated. These defendants are not comparable to Bey, who has a culpability level of .76 for sexually assaulting, beating, strangling, and stomping his victim, Carol Peniston, and who had a prior conviction for beating, sexually assaulting, and murdering Cheryl Alston. Because these cases are dissimilar from Bey’s case, they should be excluded from the pool of comparable cases. Although the Special Master noted that the pool should be expanded to include a sufficient number of cases for comparison purposes, he never said that the pool should include dissimilar cases.
The overall result of the three types of analyses constituting the frequency approach demonstrates that the capital-sentencing rate for prior murderers such as defendant is not random or aberrational. Defendant has failed to offer reliable evidence showing that for cases similar to his, a sentence other than death generally is imposed.
*366-B-
THE PRECEDENT-SEEKING APPROACH
The second part of proportionality review involves the precedent-seeking analysis. This analysis, which supplements the frequency approach, is a less mechanical and more traditional case-by-case comparison of similar death-eligible cases. As with the frequency analysis, our study incorporates not only cases decided by March 24, 1993, the date of the Bey Report, but also cases through June 25, 1993, the date of the Martini Report. The precedent-seeking analysis also persuades us that defendant’s death sentence is not disproportionate.
1. RELEVANT FACTORS
In Marshall, the Court extended the factors involved in the precedent-seeking approach beyond the statutory factors to include other “objective criteria rooted in traditional sentencing guidelines.” 130 N.J. at 159, 613 A.2d 1059 (citing N.J.S.A. 2c:44-1). The Court identified three elements of criminal culpability as examples of the types of such additional factors. Id. at 155-59, 613 A.2d 1059. The first is the defendant’s moral blameworthiness, which includes motive, premeditation, provocation, mental disease, knowledge of helplessness of the victim, knowledge of the effects on surviving victims, the defendant’s age and maturity, and his or her involvement in planning the crime. Second is the degree of victimization, which includes the extent of mutilation of the victim and injury to surviving victims. The third factor is the character of the defendant, which includes the defendant’s prior record and other acts of violence, cooperation with authorities, remorse, and capacity for rehabilitation.
In Marshall, with a few exceptions, id. at 178-79, 188, 613 A.2d 1059, we confined the precedent-seeking analysis to the same universe of cases that we considered in the frequency analysis. Id. at 178-79, 613 A.2d 1059. Here, we will maintain the same universe for both analyses. Proportionality review provides two *367different methods of analyzing the same data. If the case universe varies, the two analyses cannot confirm each other.
Defendant questions the identification of comparable cases. He argues that the comparison group should include those cases in which the juries found certain mitigating factors, such as age, a history of child abuse, mental or emotional disturbance, or remorse, or in which the prosecutor offered the defendant a plea bargain. The argument misperceives the method of selecting comparable cases. Initially, from the universe of all death-eligible cases, we select a class of eases according to their salient factors. Id. at 155, 613 A.2d 1059. Those factors, which differ from mitigating factors, are simply a means of classifying the comparable cases. Ibid. Defendant’s error when selecting those cases is in substituting mitigating factors for salient factors. We use mitigating factors to make a more detailed consideration of the comparable cases to the case at hand, not to make the initial selection of those cases.
The salient factors for determining the proportionality of Bey’s death penalty are not his age (although he was only eighteen when he murdered Carol Peniston), his alleged mental infirmity, history of child abuse, expression of remorse, or the fact that he was offered a plea bargain. Rather, Bey’s essential attribute is that before he was convicted of sexually assaulting and murdering Carol Peniston, he had been convicted of sexually assaulting and murdering Cheryl Alston. As we stated earlier, a two-time murderer is among the most blameworthy of defendants. Supra at 352-353, 645 A.2d at 694. Only eleven of the fifty-two cases that defendant offers as “similar” involve a prior murder conviction. As in Marshall, we decline to consider cases that do not exhibit the salient factors of the case under review.
Having determined how to identify the group of comparable cases, we now turn to how to compare those cases. Defendant proposes that we consider not only the factors that the sentencing jury found, but also those that the jury rejected, such as age, or *368factors that defendant never asserted, such as intoxication or the offer of a plea bargain.
As discussed above, in Marshall we distinguished similar cases based on a broader range of factors than the statutory aggravating and mitigating factors. The reason for expanding the range of factors was to reflect more accurately the factors juries consider when determining whether to impose the death penalty. 130 N.J. at 157, 613 A.2d 1059. If, however, evidence of a factor is not objective or was not submitted to a jury, we will not consider it. Thus, although the statute does not expressly include factors such as child abuse, if evidence of such abuse is clearly present on the record, a jury is likely to consider it, as will we. Here, the sentencing jury did not hear evidence that the State had offered defendant a plea bargain, that defendant had been intoxicated, or that defendant had not served prison time for the Alston murder. Hence, we will not consider evidence of those factors.
We will, nonetheless, consider defendant’s age, child abuse, and remorse. These factors are objective, rooted in traditional sentencing guidelines, were clearly presented to the sentencing jury, and are likely to influence a jury’s sentencing decision. As we have indicated, although the jury rejected age as a separate mitigating factor, it may have considered defendant’s youth in finding the catch-all factor. Supra at 360-361, 645 A.2d at 697-698.
Evidence of defendant’s abusive childhood could have influenced the jury’s findings of either the catch-all factor or the extreme-mental-or-emotional-disturbanee factor. Also, Bey’s apology to Carol Peniston’s family could have influenced the jury’s findings on the catch-all factor.
In sum, we will analyze the twenty-one cases in the Bey and Martini Reports in which defendants had a prior murder conviction, compare those cases, and determine whether Bey’s case is more like those of defendants who received a capital sentence or those who received a non-capital sentence. In conducting our analysis, we will consider objective factors clearly present on the *369record that reflect on blameworthiness, victimization, and character.
2. COMPARISON OF MARKO BEY’S CASE TO SIMILAR CASES
In the salient-factors approach, the AOC grouped Bey’s case with other cases in which the defendants had a prior murder conviction. Twenty-one cases involving thirteen different defendants exhibited a separate murder conviction that was or could have been offered as a prior murder conviction. Excluding Bey, eight of those cases resulted in a death sentence, and the remaining twelve resulted in life sentences.
Our task is to determine whether, by comparison to jury sentences of defendants in comparable cases, Bey’s sentencing jury acted aberrantly by sentencing him to death. A defendant’s sentence is not disproportionate simply because other defendants who have committed similar crimes have not received sentences other than death. No two murders are identical. The comparable cases, although similar in many respects, involve different defendants, different facts, different legal issues, and different juries. We therefore anticipate some inconsistency between the results of the comparable cases and the case before us. Of necessity, the persuasiveness of the comparison of Bey’s case to others will depend on the similarity of the facts presented to the sentencing jury in those cases. We glean those facts from the published opinions or, if the opinions are unpublished, from the AOC’s Detailed Narrative Summaries. We conclude that juries in comparable cases generally sentence defendants like Bey to death and that Bey’s jury did not act aberrantly by sentencing him to death.
a. THE CASES
RICHARD BIEGENWALD I and II
These cases involve the murder of Anna Olesiewicz (Biegenwald IA, IB, & IC) and that of William Ward (Biegenwald II). On August 27, 1982, eighteen-year-old Anna Olesiewicz and a friend, Denise Hunter, drove from Camden to Neptune City to spend the *370evening at the Asbury Park boardwalk and then stay at the home of Hunter’s uncle. While at the boardwalk, Olesiewicz sat on a bench and Hunter went to the women’s room. When Hunter returned, she could not find Olesiewicz. Hunter returned to her uncle’s home, and the next morning filed a missing persons report.
On January 14, 1983, Olesiewicz’s remains were discovered in a vacant lot behind a fast-food restaurant. Biegenwald had encouraged Theresa Smith, whom he considered a “protege,” to become “tough” by killing someone. When Smith reneged on a plan to kill one of her co-workers, Biegenwald decided to kill Olesiewicz. He lured the victim to his house by promising her marijuana. Then he shot her in the head four times. Biegenwald removed a gold ring from the victim’s finger and gave it to Smith.
The State alleged two aggravating factors: the prior-murder-conviction factor, c(4)(a), and the depraved-mind factor, e(4)(c). Biegenwald asserted three mitigating factors: extreme emotional disturbance, c(5)(a); mental disease or defect, c(5)(d); and the catch-all factor, c(5)(h). In support of the mental-disease and catch-all factors, Biegenwald presented the videotaped testimony of a forensic psychiatrist who claimed that Biegenwald had been abused as a child and had been institutionalized at the age of eight. During his institutionalization, Biegenwald had been diagnosed as schizophrenic and subjected on twenty occasions to electro-shock treatment. The psychiatrist diagnosed Biegenwald as suffering from an anti-social personality disorder with paranoid traits, a condition that prevented him from appreciating the wrongfulness of his conduct. The jury found both aggravating factors, but rejected extreme emotional disturbance as a mitigating factor. Three jurors found mental disease or defect, and four jurors found the catch-all factor. After weighing the two aggravating factors against the two mitigating factors, the jury sentenced Biegenwald to death. Biegenwald 1A, supra, 106 N.J. at 18-25, 524 A.2d 130.
We affirmed the conviction, but remanded for a new sentencing proceeding because the jury had not been instructed to find that *371the aggravating factors outweighed the mitigating factors beyond a reasonable doubt. Id. at 67, 524 A.2d 130. A second jury sentenced Biegenwald to death, but this Court reversed that sentence because of a defective voir dire. Biegenwald IB, supra, 126 N.J. at 43, 594 A.2d 172. After this Court reversed the two death sentences imposed on Biegenwald for the Olesiewicz murder, a third jury sentenced Biegenwald to a life sentence.
In the Ward murder, Biegenwald and the State’s principal witness, hit-man Dherren Fitzgerald, met with William Ward to arrange the terms of a “hit” that Fitzgerald wanted to perform for $25,000. Fitzgerald joined Ward in Ward’s ear and the two drove to Fitzgerald’s home. Biegenwald followed them in Fitzgerald’s car. Before Biegenwald arrived at Fitzgerald’s apartment, Fitzgerald and Ward discussed the terms of the “hit.” Fitzgerald, who wanted no witnesses, refused to permit Ward to watch the “hit.” Ward responded by displaying his revolver. The men wrestled over the gun. Fitzgerald claims the gun went off, shooting him in either his shoulder or neck. Fitzgerald then reached for a .22 caliber pistol with a silencer. Because he could not cock the gun with one hand, Fitzgerald hit Ward on the head with the barrel, rendering the gun inoperable.
The struggle ended with Fitzgerald on top of Ward, who was on his back, still clutching the gun. Fitzgerald stated that Biegenwald then appeared and shot Ward in the head five times. Biegenwald and Fitzgerald then stuffed Ward into the car, returned home, and stored the body in the garage until they buried it.
The prosecution served notice of only one aggravating factor, the prior-murder-conviction factor, c(4)(a). Biegenwald presented two mitigating factors: mental disease or defect, c(5)(d), and the catch-all factor, c(5)(h). The jury found the aggravating factor and both mitigating factors, but was unable to reach a verdict. Therefore, the court sentenced Biegenwald to life imprisonment with a thirty-year parole disqualifier. The Appellate Division affirmed in an unreported opinion.
*372 JAMES KOEDATICH
On November 23, 1982, at approximately 9:30 p.m., eighteen-year-old Amie Hoffman left her part-time job in a shopping center. Two days later her body was found floating face down in a water-retention tank located in a secluded area. Koedatich had abducted her in the mall parking lot. When discovered, she was wearing the same clothing as on the day of her abduction.
An autopsy revealed a long gash on the left side of her head, a wound to her right shoulder, and injuries at the base of her neck. Her left ear had been severed, leaving a deep wound that extended to the spinal chord. She also had sustained two severe chest wounds, one penetrating four-and-one-half inches and the other seven inches, through her lungs and to her back. The medical examiner theorized that the knife had been inserted once, causing the shallower wound, and then thrust in deeply, causing the seven-inch wound. The victim’s hand revealed defensive wounds consistent with grabbing for the knife, and abrasions and bruises on her left thigh and lower arm, consistent with having been dragged over the retention-tank wall. Vaginal and rectal swabs revealed sperm, and the medical examiner estimated that intercourse had occurred within twenty-four hours of the victim’s death.
The State alleged the existence of four aggravating factors: a prior murder conviction, the c(4)(a) factor; depraved mind, the c(4)(c) factor; the murder was committed for the purpose of escaping detection for another crime, N.J.S.A. 2C:11 — 3e(4)(f) (the c(4)(f) factor); and the murder was committed in the course of either a kidnapping or aggravated sexual assault, the c(4)(g) factor. Koedatich refused to allow his counsel to present any mitigating evidence concerning his childhood trauma. The trial court, nonetheless, submitted the catch-all factor, c(5)(h), and charged the jury that the decision on this factor must be unanimous. The jury found that Koedatich had committed a prior murder, that of Deirdre O’Brien, for which he had received a life sentence. It also found that he had a depraved mind, but it did not unanimously find the catch-all mitigating factor. The jury *373sentenced Koedatich to death. State v. Koedatich, 112 N.J. 225, 231-49, 548 A.2d 939 (1988), cert. denied, 488 U.S. 1017, 109 S.Ct. 813, 102 L.Ed.2d 803 (1989).
This Court affirmed the conviction, but reversed Koedatich’s death sentence because of improper instructions regarding the catch-all factor. Id. at 325, 548 A.2d 939. In the re-trial of the penalty phase, the jury found all four aggravating factors: prior murder conviction, depraved mind, murder to escape detection, and contemporaneous murder and felony. The jury also found the catch-all factor. Because the jury could not unanimously agree on the weighing of the aggravating and mitigating factors, Koedatich received a life sentence.
THOMAS RAMSEUR
On August 25, 1982, Ramseur stabbed to death his former girlfriend, fifty-four-year-old Asaline Stokes. Stokes lived with her grandchildren across the street from Ramseur’s aunt’s house. Ramseur had frequently threatened to kill Stokes and had physically attacked her. On one occasion, when he had severely beaten Stokes, the police were called. Three or four months before the murder, Ramseur threatened to kill her and her grandchildren. The day before the murder Ramseur and Stokes had an argument, during which she stated she was tired of his drinking and threats. Ramseur told her, “you’ll be sorry,” stole a knife from her kitchen, and left.
On the day of the murder, Stokes was speaking to a mechanic near her house when Ramseur left his aunt’s house and walked over to the victim and the mechanic. He patted her on the shoulder, stabbed her, and continued to stab her as she fell to the ground. As she lay dying, Ramseur taunted her by saying, “if I see your kids again I’m going to kill them too.” Stokes finally succumbed at the hospital. She had major stab wounds in the face and chest, and two wounds in the chest penetrated eight-and-one-half-inches deep, piercing her lung.
The State alleged two aggravating factors: c(4)(a), the prior-murder-eonvietion factor, and c(4)(c), the depraved-mind factor. *374Ramseur presented testimony that his behavior had changed after he had been mugged in 1982. A neurologist testified that Ramseur had progressive shrinking of the brain in the frontal and temporal lobes. A psychiatrist testified that Ramseur suffered from psycho-motor seizures, a form of epilepsy that may cause loss of control during a seizure. The psychiatrist, who diagnosed Ramseur as paranoid, testified that the stabbing had occurred during a psycho-motor seizure. Ramseur alleged that four mitigating factors were present: e(5)(a), extreme emotional disturbance; e(5)(c), age; c(5)(d), mental disease or defect; and e(5)(h), the catch-all factor. The jury found both aggravating factors and found two mitigating factors — extreme emotional disturbance and mental disease or defect. It then sentenced him to death. Ramseur, supra, 106 N.J. at 160-66, 524 A.2d 188.
This Court affirmed Ramseur’s conviction, but vacated the death sentence because the supplemental instructions on jury deadlock had coerced the death sentence. Id. at 314, 524 A.2d 188.
SAMUEL ERAZO -
On July 20, 1986, Erazo stabbed his wife, Lucy, to death. The two had been married in 1982 while Erazo was in prison serving a sentence for the murder of a young girl. After Erazo was released, he moved into Lucy’s apartment. The relationship between the two worsened. Erazo hit Lucy on many occasions. One of Lucy’s daughters spoke to parole authorities about Erazo’s living in an apartment with children, a violation of his parole. As a result, Erazo was imprisoned for several months. On the night of the murder, during a party in their apartment, Erazo and Lucy drank heavily. Tension increased throughout the evening, and Erazo became infuriated when Lucy danced the merenge with another man.
When the party ended at 11:30 p.m., Erazo left to accompany his guests home. When he returned, Lucy was leaving the apartment. She returned after midnight. Shortly thereafter, neighbors heard glass breaking and Lucy screaming that Erazo *375was killing her. Erazo changed his clothes, left the apartment house, and told a friend to call an ambulance. Medical personnel found Lucy lying on the floor next to a bloodstained knife. She had sustained four knife wounds to her hands, arms, and chest, three slashes to the neck, and a single stab wound to the back that had killed her instantly. The State asserted that Erazo had killed her because after she purposely cut her hand, she had threatened to call the police, a call that could have led to the revocation of Erazo’s parole.
At the penalty phase, the State relied on two aggravating factors: c(4)(a), prior murder conviction; and c(4)(c), depraved mind. In mitigation, Erazo presented testimony from his brother and sister urging the jury to spare his life for them and for his mother, who was ill. Corrections officers at Rahway State Prison testified that during Erazo’s imprisonment, he had been a model prisoner. Erazo also gave a statement of allocution. He offered this evidence in support of six mitigating factors: the c(5)(a) factor, extreme emotional disturbance; N.J.S.A. 3C:ll-3c(5)(b) (the c(5)(b) factor), victim participation in conduct that led to her death; the c(5)(d) factor, intoxication; N.J.S.A. ll:3e(5)(e) (the c(5)(e) factor), unusual or substantial duress; the c(5)(c) factor, age; and the e(5)(h) catch-all factor. The jury found both aggravating factors and four of the mitigating factors: extreme emotional disturbance, victim participation, intoxication, and extreme duress. It rejected the age and catch-all factors, and sentenced Erazo to death. Erazo, supra, 126 N.J. at 127-31, 594 A.2d 232.
This Court reversed the conviction and remanded the case to the Law Division for a re-trial of the guilt phase because of an error in the jury charge on passion-provocation manslaughter and because of a Gerald error. Id. at 122, 594 A.2d 172. At the time of the Martini Report, Erazo’s guilt had not yet been determined.
FRANK PENNINGTON
Pennington arrived at a bar in East Rutherford, New Jersey at about 11:30 p.m, on September 2, 1986. Thirty minutes later, the victim, Arlene Connor, arrived to help her daughter close the bar. *376Connor announced closing time at about 1:00 a.m. Pennington asked for a beer, his fourth, and went to the men’s room. The other customers had left by the time Pennington returned to the bar. Pennington shot Connor once in the heart, killing her. In a statement to the police, Pennington conceded that he had pulled out a gun, told the victim he did not want to hurt anyone, and that he just wanted money. When Connor threw a glass that hit him in the chest, he ducked, straightened up, and pulled the trigger.
The State alleged two aggravating factors: c(4)(a), prior murder conviction; and c(4)(g), murder while engaged in a contemporaneous felony. Pennington alleged three mitigating factors: c(5)(a), extreme emotional disturbance; e(5)(d), mental disease or defect; and c(5)(h), the catch-all factor.'
He presented the testimony of various family members, who asserted that his mother was immature, promiscuous, bad tempered, and had not properly raised him. For example, she had taught him to steal cigarettes for her. His mother testified that Pennington’s father was an alcoholic who had beaten her and Pennington. Furthermore, in 1968 Pennington enlisted in the Marine Corps and served in Vietnam. Medical testimony indicated that Pennington suffered from multiple-personality disorder, and that after he had returned home from Vietnam he had suffered from post-traumatic stress syndrome. He also was an alcoholic and had suffered a brain injury.
Finding both aggravating factors and the mental-disease-or-defect mitigating factor, the jury sentenced Pennington to death. Pennington, supra, 119 N.J. at 557-60, 575 A.2d 816. This Court reversed the sentence because the trial court had failed to require the jury to determine whether Pennington had intended to cause death rather than just serious bodily injury, a Gerald error. Id. at 561, 575 A.2d 816. Pennington received a life sentence in the re-sentencing trial.
BRAYNARD PURNELL
On August 28, 1988, at about 6:00 p.m., Purnell, a thirty-six-year-old cocaine user, asked a friend, Jeffrey Davis, to purchase *377$20 of cocaine from Lawrence Talley. Davis went to a local playground but was unable to make a buy. Talley later sent an associate to deal with Purnell, but Purnell wanted more cocaine than Talley would agree to sell for $20. Talley and Purnell walked to Purnell’s house, where in the backyard Purnell killed Talley and hid the body. The body was found several days later. Talley had been killed by fifteen stab wounds to the neck, chest, and abdomen. Scratches on his back were consistent with the body having been dragged. According to Davis, on the night of the murder, Talley was carrying cocaine, but the police did not find drugs or cash on Talley’s body.
The State alleged two aggravating factors: c(4)(a), prior murder conviction; and e(4)(g), murder committed during the course of a robbery. In mitigation, Purnell asserted c(5)(h), the catch-all factor; and c(5)(b), victim participation in the conduct that precipitated his death. Defense witnesses testified about Purnell’s redeeming character and personality, his good works for others, and his non-use of drugs. The jury found both aggravating factors. Three jurors found that the victim had participated in his own death, the e(5)(b) factor, and two jurors found the catch-all factor, c(5)(h). Purnell was sentenced to death. Purnell, supra, 126 N.J. at 524-30, 601 A.2d 175.
This Court reversed the conviction and the sentence because the trial court had not charged the jury on the lesser crime of felony murder. Id. at 534, 601 A.2d 175. At the second sentencing hearing, Purnell received a life sentence.
BRYAN COYLE
In 1983, shortly after his release from prison after serving a term for murder, Coyle moved to Old Bridge. He soon became sexually involved with Rhonda Lemberg, his married next-door neighbor. Lemberg told Coyle that she was unhappy with her husband, who had beaten her and her children. She also disclosed her fear that her husband would one day use against her the gun that was in their house.
*378On July 28, 1983, Lemberg and her husband had an argument, following which she went to Coyle’s house. Coyle had taken mescaline, a narcotic, before her arrival. Shortly thereafter, the husband arrived and demanded that his wife return home. When no one answered, the husband broke the window, cutting his hand. Before opening the door, Coyle retrieved his nine-millimeter handgun, loaded it, and put it in his back pocket. The husband entered and, ignoring Coyle’s efforts to placate him, moved towards his wife. When Coyle fired a warning shot, the husband fled and returned home to call the police. The husband, seeing his wife and Coyle enter Coyle’s car, ran into the street and used a discarded garage door to block the car. Lemberg believed that her husband had a gun. She told Coyle that her husband would kill her. She fled from the car, but her husband caught her and they engaged in a heated verbal exchange. Her husband walked back to his home and Coyle walked down the block with Lemberg. Shortly thereafter, the husband stormed out of his house and hurried after Lemberg. Coyle chased the husband and fired his handgun. The first two shots missed. Coyle shot again, this time hitting the husband in the leg. The husband crawled across a lawn and hid behind a tree. Coyle followed him and fired three more shots, two of which hit the husband, one in the back of the head, killing him. At trial, both Lemberg and Coyle testified that the husband had been acting irrationally on the night of the murder. Coyle also claimed that he fired at the husband to save Lemberg, that he had intended not to kill her husband, but only to stop him from attacking her.
The State alleged two aggravating factors: c(4)(a), prior murder conviction; and c(4)(c), depraved mind. Coyle asserted four mitigating factors: c(5)(a), extreme emotional disturbance; c(5)(d), intoxication; c(5)(b), victim participation; and e(5)(h), the catch-all factor. The jury found both aggravating factors and only the victim-participation factor, c(5)(b), as a mitigating factor. It sentenced Coyle to death. Coyle, supra, 119 N.J. at 201-08, 574 A.2d 951.
*379This Court reversed the death sentence because of various errors, including the absence of an instruction on the intent to cause death as opposed to serious bodily injury, and an improper charge on passion-provocation. Id. at 221, 574 A.2d 951. On re-sentencing, Coyle received a life sentence.
CARLOS VASQUEZ
On June 3, 1988, at about 8:00 a.m., Vasquez, who was forty-three-years-old, abducted a thirteen-year-old girl and then sexually assaulted and killed her. The victim, whose hands and feet were tied together behind her back with electrical cord and clothesline, was found in a box that had been put out to be collected with the trash. Vasquez said that he had made sexual advances toward the girl. When she resisted and became hysterical, he grabbed her neck to prevent anyone from hearing her. The cause of death was asphyxia caused by gagging, ligature strangulation, and fracture of the cervical spine.
At the request of the victim’s parents to spare them the stress of a trial, defendant was allowed to plead guilty to felony murder, despite the fact that he had a prior murder conviction. Vasquez received an aggregate sentence of life imprisonment plus twenty years, with a forty-year parole disqualifier. He denied any physical, mental-health or substance-abuse problems. Detailed Narrative Summaries, swpra, at 285-86.
JIHAD MUHAMMED
On August 3, 1984, Muhammed approached Dawn Andrew and Clarence Maxwell on the street and offered to sell them “speed.” When Andrew declined, Muhammed left but returned twenty minutes later with his co-defendant, Forrest Boyer. Muhammed pulled out a handgun, pointed it at the couple, and then fired it into the ground. Boyer then took Andrew’s purse, rummaged through it, and stole marijuana. Maxwell told Boyer to give back the purse. Muhammed took two steps, pulled out a sawed-off shotgun, and shot him. When Andrew’s father came out of a nearby house and asked why Muhammed had shot the victim, Muhammed replied: “I didn’t like his attitude.”
*380Muhammed pleaded guilty to murder and various weapons charges, and was sentenced to life imprisonment with an aggregate parole disqualifier of thirty-six years. Id. at 189-91.
ALBERTO NIEVES
On March 25,1987, Nieves was leaving a grocery store when he heard Hector Rentas sound his automobile horn at Nieves’s wife. In the exchange that followed, Nieves took a gun from his car and pointed it at Rentas’s head, telling him that if he wanted his girl, he should take her. Nieves then lowered the gun and returned to his car.
Three days later, on March 28, Rentas was parked outside a store with his six-year-old son when Nieves walked up to him and told him to “stop messing with my girl.” When Rentas responded that he was not messing with Nieves’s girl, Nieves raised a gun and shot Rentas once in the head. The bullet passed through Rentas’s head and lodged in the seat between him and his son. The medical examiner later testified that at the time of the shooting, the gun had been within six inches of the victim’s head.
A jury convicted Nieves of purposeful or knowing murder. At the penalty phase, the prosecution offered two aggravating factors: c(4)(a), the prior-murder-conviction factor; and c(4)(b), the grave-risk-of-death factor, for endangering the victim’s six-year-old son. The defendant offered four mitigating factors: c(5)(a), extreme emotional disturbance; e(5)(b), victim participation; N.J.S.A. 2C:ll-3c(5)(g) (the c(5)(g) factor), substantial assistance to the State; and c(5)(h), the catch-all factor. The catch-all factor was supported by evidence that Nieves was one of eighteen children and had grown up in extreme poverty. One of his siblings had been murdered and another had been imprisoned for avenging that murder.
The jury found both aggravating factors and two of the mitigating factors: c(5)(b), victim participation; and c(5)(h), the catch-all factor. One juror refused to deliberate, and the jury could not reach a unanimous verdict. The court sentenced the defendant to *381an aggregate term of life imprisonment, with a thirty-two-and-one-half-year parole disqualifier. Id. at 222-25.
GEORGE BOOKER
On January 6, 1972, George Booker was convicted of murder and was sentenced to twenty-seven to twenty-nine years in the State Prison. He was paroled on November 15, 1983. On September 11, 1985, after being asked to leave the home of friends with whom he had been staying, Booker went to the home of a thirty-one-year-old female friend, pulled out a knife, sexually assaulted her, and stole her car. As he drove away, Booker ran down a pedestrian and stole his wallet.
Booker then went to the home of two women who were living together. Booker raped and sodomized one woman, bashed in her mouth and forehead, and then strangled her with an electrical cord. When the other woman returned home, he forced her to undress and lie in the bed next to her dead roommate. Then he stabbed her to death. Booker, knife in hand, was arrested on September 13 while inside the home of an elderly female.
Booker was convicted of capital murder of both victims. The sentencing jury found aggravating factors for a prior murder conviction, c(4)(a); depraved mind, c(4)(c); and murder to escape detection, c(4)(f). Concerning the murder of the first victim, the jury also found e(4)(g), the contemporaneous-felony factor, as an aggravating factor. In mitigation of both murders, the jury found c(5)(a), the extreme-mental-or-emotional-disturbance factor. Also, the jury found c(5)(h), the catch-all factor. Because the jury could reach a unanimous decision concerning either murder, the court sentenced Booker to an aggregate sentence of life imprisonment, with a sixty-year parole disqualifier. Id. at 29-32.
HECTOR SANABRIA
Sanabria’s first murder occurred on September 25, 1984, when Sanabria, a drug dealer, shot and killed another drug dealer, Omar. The shooting was an attempt by Sanabria to obtain a monopoly over the sale of drugs in Paterson. The pre-sentence *382report stated that Sanabria and his brother, Junior, were concerned that Omar was attempting to move into their territory, and that he had robbed one of their workers. During a meeting between Omar and the Sanabria brothers, Junior struck Omar with a gun. When Omar reached for his gun, both Sanabria brothers fired their weapons. Sanabria also grabbed Omar’s gun and shot him with his own weapon. Substantial proof suggested that Sanabria had provoked the encounter. Omar’s body was found with seven bullets in it, five in the heart.
The second murder occurred on December 3, 1984, when, during an argument about the ownership of drugs, Sanabria shot and killed Edwin and Nelson Aponte on a street in Paterson. As in the killing of Omar, evidence supporting Sanabria’s claim of self defense was weak. The cause of death for both victims was multiple gunshot wounds in the chest.
A jury convicted Sanabria of Omar’s murder, and on April 25, 1986, the court sentenced Sanabria to life imprisonment, with a thirty-year parole disqualifier. Despite this murder conviction, the prosecutor did not file a notice of the c(4)(a), prior-murder-conviction factor in the case involving the murder of the Aponte brothers. The jury convicted Sanabria of the knowing and purposeful murder of both brothers. The court sentenced Sanabria to two thirty-year terms, each with a thirty-year parole disqualifier and each consecutive to the other. Detailed Narrative Summaries, supra, at 227-28.
b. THE COMPARISON
The totality of the evidence, which includes Bey’s prior murder conviction and the sexual assault of both of his victims, leads us to conclude that Bey’s death sentence is not disproportionate.
Bey argues that when measured by victimization, moral blameworthiness, and character, he is not as deathworthy as other defendants who received either life sentences or death sentences. We disagree.
*383First, Bey compares the victimization in his murder and sexual assault of Carol Peniston with that in the murders of other defendants in the comparison group. Bey points to the fact that Biegenwald has been convicted of killing a total of five people, and that Booker has been convicted of killing a total of three people, but that he has killed only two people. Although the number of victims is a factor that bears on the degree of victimization, the prosecutor presented the sentencing juries in the Biegenwald cases with evidence of only two other murders, not four. One of those murders was a 1959 conviction for a murder that had occurred during the course of a robbery. As for Booker, two of the murders occurred during the same crime spree. Unlike Bey’s two murders, they were not separate and independent incidents.
Some defendants, such as Biegenwald, Ramseur, Purnell, Coyle, and Sanabria, shot or stabbed their victims several times. Others, like Ramseur, Erazo, and Nieves, threatened their victims over a period of time, with Nieves killing his victim in the presence of a young child. Unlike Bey, however, none of these defendants sexually assaulted his victim. By comparison, Bey sexually assaulted, beat, strangled, and stomped on his victim, a woman whom he did not know and who had done nothing to provoke his rage. Koedatich, Vasquez, and Booker, who sexually assaulted their victims, are closer to Bey in terms of victimization. As we subsequently discuss, infra at 386-387, 645 A.2d at 710-711, these cases are distinguishable from Bey’s case for other reasons.
Second, defendant also contends that he is less culpable than several other defendants because he is not as morally blameworthy. Specifically, he contends he was much younger at the time of his second murder than they were when they murdered their victims. Although age often mitigates a defendant’s culpability, as we have stated above, supra at 360-361, 645 A.2d at 697-698, in light of the totality of evidence, defendant’s age by itself does not compel the return of a sentence other than death.
Bey also offered evidence of chronic child abuse and possible organic brain syndrome to support his argument that he is less *384culpable than other defendants. Like many of the comparable defendants, Bey suffered an abusive childhood. His violent childhood, however, does not differ materially from that of Biegenwald, Koedatich, Ramseur, Pennington, and Nieves, who also had suffered from child abuse or other violence. But these defendants, unlike Bey, offered additional, uncontroverted evidence to demonstrate the impact of the abuse and violence they had suffered.
For example, Biegenwald had been institutionalized as a youth from the age of eight. He also had experienced twenty electroshock treatments, and had been diagnosed as suffering from schizophrenia, anti-social personality disorder, and paranoia. Ramseur and Pennington offered uncontradicted physical evidence to support their defense of mental disease or defect. Ramseur exhibited a shrinking of his brain in the frontal and temporal lobes and suffered from psycho-motor seizures, which he had experienced during the murder of his victim. Pennington was a Vietnam veteran who suffered from post-traumatic stress syndrome, multiple personality disorder, and a brain injury. In Bey’s case, however, the State disputed the evidence of organic brain damage and introduced evidence showing that Bey suffered merely from an anti-social personality disorder that did not prevent him from understanding his actions or acting purposely. The uncontradicted evidence of physical brain damage to the comparable defendants could explain why their juries did not deem them to be deathworthy.
The dissent attacks our explanation of the difference between Bey’s sentence and that of other defendants by pointing to evidence of the abuse that Bey endured during his childhood that was excluded at trial. Specifically, the dissent points to the exclusion of a report of a state psychologist, Dr. Cooke, and the preclusion of leading questions to Bey’s mother. Post at 411-412, 645 A.2d at 723-724. Implicit in the dissent is the notion that the excluded evidence would establish that Bey is like other defendants who have received a sentence other than death. On Bey’s direct appeal, we found that the evidence was cumulative and therefore *385that the errors were harmless. 129 N.J. at 590, 594, 610 A.2d 814. We continue to believe that other evidence adequately established the abuse that Bey suffered as a child and that the excluded evidence would not have made a difference. Thus, we find that the excluded evidence does not account for the difference between Bey’s death sentence and the sentences imposed on other defendants.
Moreover, Bey failed to offer the sentencing jury any evidence of intoxication. In contrast, Booker, Erazo, and Pennington presented evidence that they had been intoxicated when they committed their crimes. Furthermore, many defendants, e.g., Biegenwald (in the murder of William Ward), Erazo, Purnell, Coyle, Pennington, and Sanabria offered evidence that they had been provoked by the victim or otherwise had been motivated by passion or duress. Bey, in contrast, did not know his victim and was not provoked by her. We conclude that the differences between Bey’s case and the comparable cases suggest that Bey is more blameworthy than these defendants.
Third, defendant attempts to distinguish his character from that" of other defendants. He states that unlike the other defendants, with the exception of Sanabria, he had not served any prison time for a prior murder. Defendant, however, does not explain how this fact reflects on his character. From this evidence, we cannot conclude that Bey’s character is any better than that of the other defendants.
Last, defendant also offers his remorse as evidence that his character is less culpable than that of other defendants. In support, he points to his apology to the victim’s family. Bey, however, expressed remorse only at the sentencing phase when facing the death penalty. Under these circumstances, his belated apology does not demonstrate that his character is any better than that of the other defendants.
In sum, the results of our analysis of the degree of victimization, moral blameworthiness, and character of defendants in comparable eases do not support the conclusion that Bey’s sentence is *386disproportionate. Bey may seem no worse in some respects than defendants who have received a sentence other than death. Yet, as the preceding discussion illustrates, the cases of those other defendants differ significantly from Bey’s ease. So viewed, Bey’s death sentence is not aberrant.
Each of the defendants in the comparison group had a prior murder conviction. Except for Koedatich, Vasquez, and Booker, however, their cases share a common characteristic that distinguishes them from Bey’s case: the absence of sexual assault of the victims. As culpable as defendants with prior murder convictions may be, a jury could find that a defendant such as Bey, with two convictions for murder and aggravated sexual assault, is particularly deathworthy.
Koedatich, Vasquez, and Booker exhibit similar levels of culpability because each of these defendants sexually assaulted his victim and each had been convicted of a prior murder. But these three defendants, unlike Bey, ultimately received life sentences for their crimes. From this, defendant argues that he also should have received a life sentence. We disagree.
First, we expect that juries may decide similar cases differently. Disparity alone does not demonstrate disproportionality. Marshall, supra, 130 N.J. at 181, 613 A.2d 1059. As we stated in Marshall:
The ultimate question concerns whether the fact that a jury spared [other defendants in comparable cases] requires the invalidation of Robert Marshall’s death sentence. We do not believe that statutory disproportionality ever contemplated that two New Jersey juries must reach identical verdicts even in closely-similar circumstances. Our search should be for some impermissible or invidious factor or pattern that has been broken. That the [other defendants] were spared their lives does not establish a pattern of life-sentencing for such killings. We do not sense that some invidious factor tainted Marshall’s sentencing process.
[Ibid.]
Here, as in Marshall, we do not find a pattern of life sentencing or the taint of an invidious factor that would require us to reverse Bey’s death sentence. Unusual circumstances arising in Koedatich and Vasquez preclude a finding of a pattern of life sentencing for sexual-assault murders. In Koedatich, after a jury unanimous*387ly decided to sentence Koedatich to death, one juror in the second penalty-trial phase, in light of Koedatich’s abusive childhood, refused to consider the death penalty. The mere fact that one juror in one case prevented the imposition of the death penalty need not prevent other juries from imposing that penalty on another defendant in another case. Similarly, in Vasquez, the victim’s parents insisted that Vasquez be allowed to plead to a non-capital offense to spare them the trauma of a trial.
As previously noted, Booker is distinguishable on the facts. It involved a defendant who went on a crime spree apparently caused by substance abuse. Booker committed his sexual assaults and murders during this spree. The jury heard evidence that Booker had used marijuana and anti-depressants, and had drunk a quart of beer before commencing his rampage. Uncontroverted expert testimony explained that Booker had an unusually exaggerated reaction to these drugs and had become uncontrollable while under their influence. By comparison, the second Bey jury did not hear any evidence of intoxication. Bey, moreover, committed his sexual assaults and murders in separate, independent criminal episodes.
Finally, in neither Vasquez nor Booker was the prior murder joined with a sexual assault. Only Bey’s and Koedatich’s murders joined rape and murder. As we stated above, the Koedatich case was unusual because one juror, in light of mitigating evidence, refused to consider imposing the death penalty. Despite their basic similarities, the Koedatich, Vasquez, and Booker cases differ sufficiently to support our conclusion that Bey’s death sentence is not disproportionate.
3. OTHER CASES
In the category of cases of defendants with prior murder convictions, defendant includes the cases of Leroy Taylor, Orlando Montalvo, and Marcus Rogers. We decline to consider these cases because none of the defendants had been convicted previously of murder. The prosecutor could not have asserted the prior murder conviction in any of them. In Taylor, the defendant was *388adjudicated a delinquent in his first murder case and therefore was not “convicted” of murder. Detailed Narrative Summaries, supra, at 257-60. In Montalvo, the defendant pled guilty to manslaughter in his first murder case before the second murder trial began. Id. at 179-83. Finally, in Rogers, the defendant’s prior murder conviction was reversed on appeal, and he then pled guilty to manslaughter before he committed the second murder. Id. at 215-19. In none of these cases could the State have asserted the prior-murder-conviction aggravating factor. Hence, they are distinguishable from Bey.
-TV-
RACE AS AN IMPERMISSIBLE FACTOR
Finally, defendant contends that prosecutors and juries impermissibly consider the race of defendants and of victims when imposing the death sentence. His point is that if he were not an African-American, the prosecutor would not have sought and the jury would not have imposed the death penalty. The statistics do not support his contention. Our abiding problem with analyzing the effect of race is that the case universe still contains too few cases to prove that the race of a defendant improperly influences death sentencing.
That fundamental point distinguishes our opinion from the dissent. The inescapable fact is that we lack enough cases to conclude with any degree of statistical reliability whether race is working impermissibly in death sentencing. For the dissent, however, the “under-sized data pools and consequently large margins for error,” post at 425, 645 A.2d at 730, merely mean that the Court has not met its burden to ensure that the imposition of the death penalty is proportionate. As we have explained above, however, we believe that the burden remains that of the defendant to prove disproportionality. Supra at 348-349, 645 A.2d at 692-693.
*389In Marshall, we reaffirmed our commitment to equality in the administration of justice, stating that
were we to believe that the race of the victim and race of the defendant played a significant part in capital-sentencing decisions in New Jersey, we would seek corrective measures, and if that failed we could not, consistent with our State’s policy, tolerate discrimination that threatened the foundation of our system of law.
[130 N.J. at 209, 613 A.2d 1059.]
We remain committed to that belief. Consequently, we will continue to monitor any correlation between race and the imposition of the death penalty.
Also in Marshall, we indicated that we would find the race-based disparities described in McCleskey v. Kemp, 481 U.S. 279, 326-27, 107 S.Ct. 1756, 1785-86, 95 L.Ed.2d 262 (1987), to be constitutionally significant. Id. at 210, 613 A.2d 1059. In McCleskey, the United States Supreme Court sustained the imposition of the death penalty notwithstanding certain disparities in death sentencing according to the race of the defendant and of the victim. 481 U.S. at 291, 107 S.Ct. at 1766, 95 L.Ed.2d at 277. Those disparities were that white-victim cases received capital sentences at a rate eleven times that of cases involving black victims; black defendants who killed white victims were sentenced to death at nearly twenty-two times the rate of black defendants who killed black victims and seven times the rate that pertained to white defendants who killed black victims. Id. at 326-27, 107 S.Ct. at 1785, 95 L.Ed.2d at 300-01. The McCleskey data further indicated that prosecutors sought the death penalty for seventy percent of black defendants who killed white victims, but for only fifteen percent of black defendants who killed black victims and nineteen percent for white defendants who killed black victims. Id. at 327, 107 S.Ct. at 1785, 95 L.Ed.2d at 301. Although the United States Supreme Court found these data not to be significant under the Federal Constitution, we believe that these disparities could be significant under the New Jersey Constitution. Marshall, supra, 130 N.J. at 210, 613 A.2d 1059.
Unlike the data in McCleskey, the Marshall data did not demonstrate that race played a constitutionally-significant role in death sentencing. Ibid. In Marshall, the Special Master presented two tables: Table 18, which treats the race of defendants; *390and Table 18A, which treats the race of victims. Table 18 illustrated that at culpability level four, black defendants are sentenced to death sixty-four percent more often than non-black defendants. Ibid. According to Table 18A, white-victim cases are 1.4 times more likely to advance to the penalty trial than cases involving other victims. Id. at 211, 613 A.2d 1059.
Although these tables demonstrate a degree of disparity that troubled us in our analysis of McCleskey, we ultimately found in Marshall no substantial discrimination in the application of the Act. One reason was that the tables did not provide an extensive set of relationships between the statistical variables. Id. at 210, 613 A.2d 1059. Table 18 and Table 18A showed the rate of death sentencing for defendants and victims individually, but they did not show, as the McCleskey data showed, the death-sentencing rate for race-of-victim and race-of-defendant combinations.
Another reason we rejected Marshall’s arguments was that the number of cases involving defendants, black or non-black, with comparable culpability factors was too few to support any rehable conclusion. Id. at 211, 613 A.2d 1059. Finally, and most importantly, the data showed no race-of-victim effects in penalty-trial decisions. Id. at 212-13, 613 A.2d 1059. For these reasons, we concluded that the data showed no “ ‘substantial discriminatory effect in the application of the [Act].’ ” Id. at 211, 613 A.2d 1059 (quoting Final Report, supra, at 103).
Here, amicus curiae Association of Criminal Defense Lawyers of New Jersey and New Jersey State Conference of NAACP Branches have attempted to correct the deficiencies we identified in Marshall. Amici developed a more extensive set of relationships by evaluating the interaction between race-of-victim and race-of-defendant combinations and by assessing the influence of statutory and non-statutory factors such as socio-economic status and the gender of the defendant. Also, amici updated the Marshall universe by adding forty additional cases. It recently supplemented this larger universe with the data from the Martini Report. The Martini data add eight more penalty-trial eases to *391Table 18 and twenty-eight more penalty-trial cases to Table 18A. Using these amended tables, Bey argues that an overall racial disparity exists at a statistically-significant level, particularly at the middle ranges of culpability where the choice between life and death is less certain.
As in Marshall, Bey’s Table 18 displays race-of-defendant disparities in death-penalty-sentencing decisions among penalty-trial cases after adjusting the standard-culpability levels. As stated above, only seven cases fall within culpability level four, Bey’s level. The Martini data add only one case. Bey, to include enough cases for a statistically-reliable comparison, redefined level four from .60-.80 to .145-89. His new level four now contains twenty-three cases, ten of which resulted in the death penalty. Thus, forty-three percent (10/23) of the cases at culpability level four resulted in the death penalty. The Martini data include twenty-seven cases, thirteen of which resulted in the death penalty, thereby increasing the death-sentencing rate at level four to forty-eight percent (13/27).
Bey argues that a disproportionate number of these death sentences were imposed on black defendants. Of the ten cases included in the Bey data that resulted in death sentences, eight involved black defendants, but only one defendant was white and one was Hispanic. Therefore, the death sentence was imposed on black defendants in culpability level four, Bey’s level, at a rate of eighty percent (8/10); for non-black defendants, the rate was only fifteen percent (2/13). The results are produced below:
Culpability Level Black Defendant Non-Black Defendant % Disparity
1 0 ( 0/21) 0 ( 0/12) 0
2 0(0/8) 0 ( 0/11) 0
3 .17 ( 1/6 ) 0 ( 0/15) 17
4 .80 ( 8/10) .15 ( 2/13) 65
5 1.0 (12/12) .94 (16/17) 6
*392The Martini data reflect a similar, although slightly decreasing disparity between black and non-black defendants. Of the thirteen cases that received a death sentence, ten involved black defendants and only three involved white or Hispanic defendants. Therefore, Bey’s data show that at culpability level four, the death sentence was imposed on black defendants at a rate of eighty-three percent (10/12), and on non-black defendants at a rate of twenty percent (3/15). The results are:
Culpability Level Black Defendant Non-Black Defendant % Disparity
1 0 ( 0/16) 0 ( 0/10) 0
2 0 ( 0/12) 0 ( 0/15) 0
3 .2 ( 2/10) .13 ( 2/16) 7
4 .83 (10/12) .20 ( 3/15) 63
5 1.0 (11/11) .94 (15/16) 6
Despite amici’s best efforts, defendant’s analysis remains flawed. Defendant’s redefinition of the culpability levels distorts culpability level four, the level that evidences the highest percentage of disparity and that includes Bey. The basic problem is that level four includes too much. To create middle ranges that contain a sufficient number of cases, defendant extended culpability level four from a range of .20 to .75. This extended range fails to achieve the underlying purpose of creating culpability levels consisting of similar cases. Supra at 364-365, 645 A.2d at 700. In Marshall, the Special Master chose the original twenty-percent ranges so that each range would contain sufficiently similar cases in terms of blameworthiness. Admittedly, he also stated that culpability level four would need to be expanded to include a sufficient number of cases for a valid statistical analysis. Marshall Report, supra, Technical Appendix 9 at 5. The Special Master, however, never stated that any range so expanded would be statistically reliable. As expanded, culpability level four includes cases that are dissimilar. Thus, the level is unreliable.
*393The dissent’s attempt to develop a reliable statistical base fails for the same reason. Although the dissent’s culpability level four is smaller than defendant’s, it still includes cases that are dissimilar. The dissent’s level four includes cases -with a predicted probability of a death sentence ranging from .19 to .85, a sixty-six percentage-point differential. A culpability range that spans sixty-six percentage points, although narrower than defendant’s range, is still too broad to ensure the inclusion only of comparable cases. Furthermore, the dissent’s perceived “true mid-range cases,” post at 424, 645 A.2d at 730, with a predicted probability of .30 to .70, includes only fifteen cases at level four. So meager a number of cases is too small to support the dissent’s conclusion that “an obvious disparity between races is visible.” Post at 424, 645 A.2d at 730. We note, moreover, that the dissent’s compilation of cases, unlike defendant’s Table 18, is not limited to penalty-trial cases and includes cases that are not even death eligible. Post at 423, 645 A.2d at 729 n. 4.
Implicit in the extensions of level four as proposed by defendant and by the dissent is the admission that without extending the range to include additional cases, level four would contain too few cases to support a reliable statistical conclusion. The lack of sufficient cases becomes clear if we confine our analysis to the standard twenty-percent levels contained in the Bey and Martini Reports. Of the cases included in level four in the Bey Report, only seven proceeded to the penalty phase, three of which resulted in the imposition of the death penalty. The comparable data in the Martini Report show only eight cases, four of which resulted in the death sentence. Neither table contains a sufficient number of cases to determine whether a significant statistical disparity exists between death-sentencing black and non-black defendants. In the Martini Report, moreover, Bey’s predicted probability of receiving a death sentence increases to .81, which places him in culpability level five, the highest culpability level.
Without a sufficient number of similar cases, we cannot hold that race impermissibly influences the imposition of the death *394penalty. As vexing as waiting for more data may be, we have no alternative but to wait. To force the analysis by adding dissimilar cases, as defendant and the dissent propose, would disserve the ends of justice. We do not foreclose all attempts to modify the culpability ranges to produce a sufficient number of cases for a valid statistical analysis. Any such modification, however, must consist of ranges containing similar cases.
Defendant’s Table 18 also addresses other impermissible factors, such as socio-economic status. Defendant argues that socioeconomic status aggravates racial disparity at Bey’s culpability level. The flaw in defendant’s analysis is that he subjectively defines socio-economic status. The problem is not that we should never consider socio-economic status. In Marshall, we stated that such data might be relevant. 130 N.J. at 135, 203, 214, 613 A.2d 1059. The data, however, must be objective and rooted in traditional sentencing guidelines. Supra at 366, 645 A.2d at 700-701. In Marshall, we also accepted the defendant’s argument that we should not undertake subjective, moralistic judgments when considering non-statutory factors. Id. at 155, 613 A.2d 1059. Socioeconomic status, as defined by Bey, invites precisely that kind of subjective judgment.
Defendant appears to have defined socio-economic status according to general job descriptions without considering other relevant facts about the defendants’ or the victims’ lifestyles. For example, defendant identifies high socio-economic status as including victims or defendants who are employed as secretaries, government workers, and store managers. Consequently, he identifies Carol Peniston as having a high socio-economic status simply because she was a secretary. From the record, however, we cannot glean sufficient information to justify that conclusion. Bey contends that defendants of low socio-economic status include those who have never worked, have worked sporadically, or are engaged in organized crime. According to this classification, defendant deemed William Todd Lewis to be of low socio-economic status, although Lewis had worked as a truck driver consistently *395since 1971, had earned $400 per week, and was married to a woman who owned her own house and car. Similarly, defendant classifies Samuel Mincey as being of low socio-economic status, although Mincey had owned his landscaping business for five years, worked in construction, and had been a maintenance worker.
Defendant’s Table 18A illustrates the race-of-victim disparities in penalty-trial death-sentencing decisions after adjusting the culpability levels to the same extent that we find unacceptable in Table 18. Unlike in Marshall, this table demonstrates a more extensive set of relationships that are similar to the McCleskey comparisons; defendant evaluated the interaction between race-of-victim and race-of-defendant combinations, and assessed the influence of statutory and non-statutory factors such as socio-economic status and gender of defendant. Defendant argues that the results definitively show that at the penalty-trial phase, defendants who kill white victims are more likely than defendants who kill non-white victims to receive the death sentence. He presses the point although both his victims were African-American women. We produce the results below.
Culpability Level White Victims Non-white Victims % Disparity
1 .19 ( 6/32) .06 ( 3/50) 13
2 .35 ( 8/23) .24 ( 9/37) 11
3 .65 (15/23) .29 ( 4/14) 36
4 .72 (13/18) .71 (12/17) 1
5 .90 (27/30) .91 (20/22) -1
As with the race-of-defendant data in defendant’s Table 18, the data in his Table 18A continues to reflect a disparity that generally decreases with additional cases. The Martini data add twenty-eight cases, four of which fall within culpability level four. The results are:
*396Culpability Level White Victim Non-white Victim % Disparity
1 .18 ( 7/38) .03 ( 2/59) 15
2 .42 (10/24) .29 (14/48) 13
3 .57 (13/23) .27 ( 3/11) 30
4 .74 (17/23) .75 (12/16) -1
5 .93 (27/29) .87 (20/23) 6
On examination, however, the tables do not show any disparity at Bey’s level of culpability in the imposition of the death penalty because of the race of the victim. Moreover, Table 18A suffers from the same flaws as Table 18: the modified culpability ranges include cases that are dissimilar and that are based on inadequate measures of socio-economic status. In sum, we do not find from the data presented that the race of either the defendant or the victim plays an impermissible role in death sentencing. Likewise, we do not find that the socio-economic status of the defendant or the victim plays any such role.
Defendant also argues that his sentence violates the United States Constitution because juries do not generally impose death sentences and because geographic disparities impermissibly affect death sentencing. We rejected those arguments in Marshall, supra, 130 N.J. at 188-206, 613 A.2d 1059, and continue to find them unpersuasive.
-V-
CONCLUSION
As in Marshall, we face a universe of cases that is too small to support reliable comparisons in some areas of the frequency approach and in our evaluation of racial disparities in sentencing. We also recognize that our method of comparison, such as the inclusion of cases in which the death sentence has been reversed, is not perfect. Overall, however, the statistical analyses and our own more traditional review of the cases support the conclusion *397that defendant’s death sentence is not disproportionate. Furthermore, defendant has failed to show that impermissible factors, such as race, have played a constitutionally-significant role in the imposition of the death penalty.
The imposition of the death penalty on defendant is not disproportionate.