ON APPEAL FROM THE DENIAL OF SUCCESSIVE PETITION FOR POST-CONVICTION RELIEF
BOEHM, Justice.Zolo Agona Azania, formerly known as Rufus Averhart, was convicted of murder and sentenced to death. In this appeal from the denial of his second petition for post-conviction relief, Azania argues that his death sentence must be vacated because the jury that recommended imposition of the death penalty was the product of a system for jury pool selection that systematically and materially reduced participation of African-American jurors.
In an ordinary lawsuit we would not find the irregularities in the Allen County jury selection process sufficient to require a reversal. The disproportionate reduction of African-Americans in the jury pool was, as the Chief Justice’s dissent observes, the result of a “computer glitch,” more precisely, a flawed program, not a hardware defect. But computer failures can have serious consequences, and this is an example of that. Because of the heightened need for public confidence in the integrity of a death penalty, we conclude that although the conviction was proper, the jury pool selection process was fundamentally flawed, and reversal of the death penalty and a new penalty phase or resentencing is required.
Factual and Procedural Background
Azania was convicted of murder and sentenced to death for the 1981 slaying of Gary Police Lieutenant George Yaros in the course of a bank robbery.1 In 1984, this Court affirmed his conviction and sentence on direct appeal. Averhart v. State, 470 N.E.2d 666 (Ind.1984). Azania was denied post-conviction relief, and in a 1993 appeal from that ruling, this Court affirmed Azania’s conviction but reversed his sentence, citing ineffective assistance of counsel at the sentencing phase and the failure of the prosecution to provide gunshot residue test results to the defense. Averhart v. State, 614 N.E.2d 924, 930 (Ind.1993).
After remand for a new penalty phase, Azania unsuccessfully moved to strike the entire jury pool on the ground that it did not represent a reasonable cross section of the community. A new jury was impaneled and it also recommended death. After the trial court again sentenced Azania to death, this Court affirmed the sentence on direct appeal. Azania v. State, 730 N.E.2d 646 (Ind.2000). Azania was then granted leave to file a successive petition for post-conviction relief on two grounds: newly discovered evidence, and alleged abnormalities in the Allen County jury pool selection system. Azania v. State, 738 N.E.2d 248 (Ind.2000). The successive post-conviction court denied relief, and this appeal followed.
I. Jury Pool Selection
A. The Statutory Standard
The method by which jury pools are selected in Indiana is governed by statute. Indiana Code section 33-4-5-2(c) *1257allows jury commissioners to use a computerized jury selection system, but requires that the system employed “must be fair and may not violate the rights of persons with respect to the impartial and random selection of prospective jurors.” This Court long ago held that the purpose of the jury selection statute is to ensure that the method used to select a jury is not arbitrary and does not result in the systematic exclusion of any group. Shack v. State, 259 Ind. 450, 459-60, 288 N.E.2d 155, 162 (1972). Nevertheless, there is no requirement that any particular segment of the population be represented on every jury, Daniels v. State, 274 Ind. 29, 35, 408 N.E.2d 1244, 1247 (1980), and completely random selection of jurors is not required as long as the system used is impartial and not arbitrary. State ex rel. Burns v. Sharp, 271 Ind. 344, 348, 393 N.E.2d 127, 130 (1979). Minor irregularities will not constitute reversible error unless there is a showing of substantial prejudice to the accused’s rights as a result of the irregularities. Porter v. State, 271 Ind. 180, 201, 391 N.E.2d 801, 816 (1979), overruled on other grounds. Despite these somewhat flexible standards, an accused is entitled to a trial by a jury selected in substantial compliance with the statute, and if there is a lack of substantial compliance, the accused need not show actual prejudice. Cross v. State, 272 Ind. 223, 226, 397 N.E.2d 265, 268 (1979); Wireman v. State, 432 N.E.2d 1343, 1354 (Ind.1982) (Hunter, J., dissenting); Rogers v. State, 428 N.E.2d 70, 72 (Ind.Ct.App.1981); Bagnell v. State, 413 N.E.2d 1072, 1075 (Ind.Ct.App.1980).
B. Allen County’s System of Pool Selection
The computerized system used to select the jury pool for Azania’s 1996 sentencing recommendation hearing was designed in 1980. The successive post-conviction court found that the system had four flaws, the net effect of which was exclusion of a number of jury pool members who resided in Wayne Township from the possibility of being called to serve. Specifically, in 1996, when Azania’s penalty phase was retried, these problems excluded 4364 of 5013, or 87%, of Wayne Township voters from jury service. In that year, the countywide jury pool was 14,364.
1. Overview of the Problem
The problem in Allen County’s jury selection procedures may be readily stated in broad overview. The number of jurors needed for 1996 was first identified as 14,000. The program then selected 14,364 registered voters to be assigned a random number. Only persons assigned a number could be drawn for a panel. The assignment stopped after 10,000 voters had received numbers. Because the program worked through the voter list by township in alphabetical order, all of the excluded 4364 registered voters were Wayne Township residents. As a result, 87% of Wayne Township was excluded. This had a materially disproportionate effect on African-Americans because African-Americans comprised 8.5% of the total population of Allen County, and three fourths of that 8.5% resided in Wayne Township. The remainder of this Part I:B explains the details of how this occurred. Its legal implications are addressed in Part C.
2. Truncation
The first problem resulted from a truncation feature embedded in the program since 1980. The program would first read the registered voter list and determine the total number of registered voters in the county and in each township. The program would then determine the percentage of all Allen County registered voters who resided in each township. Before *1258each calendar year, the court administrator determined the desired number of jurors required for all Allen County courts for the entire year. Based upon the requested size of this “master pool,” the program then determined the number of jurors it needed to select from each township to ensure proportional representation of that township in the master pool. The total voter list for the township was then to be divided into that number of “selection groups” by dividing the total number of registered voters in the township by the number of jurors needed from the township. One juror was then to be chosen from each group. This division rarely produced an integer (e.g., 21). In almost all cases, it produced a real number (e.g., 21.2439). The program then truncated this real number by eliminating everything after the decimal point and converting the real number (21.2439) into an integer (21). The program then used the integer, rather than the real number, to select groups, identifying the first 21 as group 1, then 22 through 42 as group 2, etc. By using the truncated integer, which was a fraction smaller than the real number, rather than rounding to the nearest integer, the program produced roughly 5% more groups than the requested size of the master pool. A random number was then used to select one juror from each group, producing a response in the range of 10,500 names to a request for 10,000 jurors. Thus, from the outset of the program in 1980, this truncation caused more voters than were requested to be chosen for assignment of a random number.2
3. The Effect of Growth in the Requested Number of Jurors
Regardless of how many names were included on the master jury pool list, from the outset the program assigned random numbers — necessary for actual selection to serve — to only 10,000 voters. When the list exceeded 10,000 names, the effect of this was to cut the list off at 10,000. From 1980 to 1994, the court administrator requested annual master jury pools of 10,000 people. During that period, the approximately 500 excess jurors produced by the truncation feature were excluded from service, but only those 500 jurors were affected. In 1995, however, the requested number grew to 12,000 jurors, and the truncation feature added another 693, so 12,693 voters were selected. As a result of assigning a random number to only 10,000 jurors, 2693 of those jurors could not be called to serve. In 1996, the year of Azania’s resentencing, the requested jury pool was 14,000, and the truncation feature added 364 names. As a result of the limitation to 10,000, 4364 of those did not receive random numbers and could not serve.
4. The Accident of the Alphabet
Finally, and importantly, the computer organized the county jury pool by town*1259ships in alphabetical order. This placed all Wayne Township jurors at the end of the list of 14,364. Thus, in each year since 1980 all of the excluded jury pool members were Wayne Township residents. The effect of these problems was not unfocused or randomly distributed over the county or over population groups. According to the 1990 census, African-Americans comprised 18,552 or 8.5% of the total age 18 and over Allen County population of 217,332. In addition, 13,937 (75.1%) of these 18,552 African-Americans resided in Wayne Township. Accordingly, the program excluded 87% of the jury pool members from the township in which 75.1% of Allen County’s age 18 and over African-Americans resided.
Azania argues that the result of these problems was that in the quarterly draw from which his jury pool was taken, African-Americans — who in a truly representative system would have comprised 8.5% of the pool — in fact comprised only 4.4% of the pool. The post-conviction court rejected Azania’s calculation as unreliable. The court ruled that using 1990 census data “as a proxy for the racial composition of the 1996 voter registration list” — as well as using a mathematical formula to estimate the number of African-Americans in the quarterly draw from which Azania’s jury was comprised — was akin to “asking the court to make an inference from an inference, something the court is not allowed to do.” The. post-conviction court may be correct that African-American citizens do not necessarily register to vote in proportion to their population, but Allen County did not maintain racial information about the voter list and we have nothing to go by except the census. Both the United States Supreme Court and the lower federal courts have repeatedly upheld the use of census figures in constitutional assaults on jury selection procedures. See Duren v. Missouri, 439 U.S. 357, 365, 99 S.Ct. 664, 58 L.Ed.2d 579 (1979) (upholding the use of six-year-old census data in fair cross-section challenge); Alexander v. Louisiana, 405 U.S. 625, 627, 92 S.Ct. 1221, 31 L.Ed.2d 536 (1972) (upholding the use of six-year-old census data in equal protection challenge); Davis v. Warden, 867 F.2d 1003, 1014 (7th Cir.1989); United States v. Osorio, 801 F.Supp. 966, 977-78 (D.Conn.1992). We agree with the courts that have concluded that under these circumstances a “defendant should not be expected to carry a prohibitive burden in proving underrepresentation.” Davis, 867 F.2d at 1014. Similarly, because no statistical data was available regarding the number of African-Americans in the quarterly draw from which Azania’s jury was comprised, it was appropriate for Azania’s expert witness to use a mathematical formula derived directly from the operation of Allen County’s computerized system to estimate that number.
C. The Effect of the Elimination of 87% of Wayne Township from Jury Service
The United States Supreme Court has long held that “the selection of a petit jury from a representative cross section of the community is an essential component of the Sixth Amendment right to a jury trial.” Taylor v. Louisiana, 419 U.S. 522, 528, 95 S.Ct. 692, 42 L.Ed.2d 690 (1975). We think our state statute, in requiring an “impartial and random selection” demands no less. Although we reach our holding today under Indiana Code section 33-4-5-2(c) and not under the Sixth Amendment to the Federal Constitution, we think that the Indiana statute ultimately turns on an issue very similar to Sixth Amendment analysis: whether the flaws in a jury selection system are so minor as to be inconsequential or are material enough that a *1260segment of the population has been materially excluded.
The federal courts have developed two competing tests under the Sixth Amendment to determine if a jury pool adequately represents the community. Under the absolute disparity test, the “disparity” is the difference between the percentage of the distinctive group eligible for jury duty and the percentage represented in the pool. In this case, where the percentage of African-Americans eligible for jury duty in Alen county is 8.5% and the percentage represented in the pool is 4.4%, this amounts to an absolute disparity of 4.1%. Under the comparative disparity test, the “disparity” is calculated by dividing the absolute disparity by the percentage of the group eligible for jury duty. Here, that results in the division of 4.1% by 8.5%, for a comparative disparity of 48.2%. Put differently, as the result of flaws in Alen County’s system, Arican-Americans as a group had roughly half the chance of being included on a jury panel than a truly random system would have produced. Nevertheless, the post-conviction court concluded that in Azania’s case the computerized system “impartially and randomly selected] citizens to be jurors, and thus substantially complie[d] with [section 33A-5-2(c) ].” We agree this may be true for non-death penalty cases, but we do not agree that the Alen County system in place in 1996 was sufficiently impartial or random to support a jury recommendation of the death penalty.
As the Supreme Court of the United States held in Powers v. Ohio, 499 U.S. 400, 413, 111 S.Ct. 1364, 113 L.Ed.2d 411 (1991):
The purpose of the jury system is to impress upon the criminal defendant and the community as a whole that a verdict of conviction or acquittal is given in accordance with the law by persons who are fair. The verdict will not be accepted or understood in these terms if the jury is chosen by unlawful means at the outset.
The Indiana jury selection statute is designed to ensure that the method used to select a jury is not arbitrary and does not result in the systematic exclusion of any group. The United States Supreme Court has long emphasized that “the qualitative difference of death from all other punishments requires a correspondingly greater degree of scrutiny of the capital sentencing determination.” California v. Ramos, 463 U.S. 992, 998-99, 103 S.Ct. 3446, 77 L.Ed.2d 1171 (1983). The Supreme Court has also held, in a death penalty case, that a jury’s being chosen from a fair cross section of the community is “critical to public confidence in the fairness of the criminal justice system,” and that the systematic exclusion of “identifiable segments playing major roles in the community cannot be squared with the constitutional concept of jury trial.” Taylor, 419 U.S. at 530, 95 S.Ct. 692. Widespread concern over the fairness and reliability of death sentences demands that the courts and the public have no significant doubts as to the integrity and fairness of the process. These same considerations require heightened sensitivity in a death penalty case in determining whether a jury selection system is “random” and “impartial” as required by Indiana law.
In this case, Azania properly preserved his right to contest the impartiality of the computerized system by moving to strike the entire jury pool. As the court below noted, the system’s programming error excluded 4364 people — roughly one-third of the jury pool — from possible service, and reduced by nearly one-half the odds that an African-American would appear on the jury panel. Every one of the excluded jury pool members was from *1261Wayne Township, the township in which three-fourths of Allen County’s African-Americans over age 18 resided. The net result was that the flaws inherent in the selection system materially reduced the probability that African-Americans would serve on Azania’s penalty phase jury. Accordingly, the system did not substantially comply with section 33-4-5-2(c), and a new penalty phase is required.
Finally, as the dissent observes, in 1982 Azania requested a transfer of this case from Lake County, where Officer Ya-ros was slain. Unlike the dissent, we do not consider that to be relevant here. Azania exercised his right under generally applicable procedures to seek a transfer to another county. In so electing, he did not forfeit his right to a properly selected jury in the new county, whatever its demographic composition.
II. Change of Judge
Prior to his second post-conviction hearing, Azania twice ■ unsuccessfully moved for a change of judge pursuant to Indiana Post-Conviction Rule l(4)(b). Azania alleged bias on the part of the trial judge, who as a member of the Allen Superior Court Board of Judges had some oversight responsibility for the computerized jury selection system. Rule l(4)(b) mandates a change of judge when the historical facts recited in the affidavit filed in support of the motion, if taken as true, support a rational inference of bias or prejudice. This Court will presume that a judge is not biased against a party. Lambert v. State, 743 N.E.2d 719, 728 (Ind.2001). This Court recently held that denial of a motion for change of judge under Criminal Rule 12 is reviewed under a clearly erroneous standard. Sturgeon v. State, 719 N.E.2d 1173, 1182 (Ind.1999). We think the same standard applies to post-conviction court motions under Rule l(4)(b). Both rules call for a change of judge if the affidavits support “a rational inference of bias or prejudice.”
The pertinent historical facts recited in Azania’s affidavits in support of his motions for a change of judge were that (1) the trial judge was personally involved in the investigation of the computer problems and, if not serving as judge, might be called as a witness in the case, (2) because of the destruction of some of the jury selection evidence, the court would be called upon to assess the credibility of some Allen County court employees and judicial officers, and (3) after the judge became aware of the problems, he did not notify Azania of any problems with the computer system, even though the judge had recently presided over Azania’s sentencing hearing with a jury selected by the same system. Our holding in Part I of this opinion renders moot Azania’s first two historical facts. As to the remaining one, we do not believe the trial judge’s failure to notify Azania of problems with the computer system raises a rational inference of bias against Azania. Azania points to no authority requiring — or even suggesting' — such a notification by a trial judge to a defendant in a closed matter. Nor are we aware of any. The trial court’s denial of Azania’s motions for a change of judge was not clearly erroneous.
III. Allegedly False Testimony
At Azania’s 1982 trial, James McGrew identified Azania as the man McGrew saw place a pistol and jacket in some bushes not far from the scene of the robbery. McGrew also testified that when a police officer pursuing Azania approached McGrew, McGrew told the officer, “I believe that the guy you’re looking for is over there,” and pointed in the direction Azania had gone. McGrew testified that when the officer returned “about a minute later,” with Azania now face down in the back of a *1262patrol car, McGrew positively identified Azania as the man who placed the objects in the bushes.
In a 1995 deposition in preparation for Azania’s penalty phase retrial, McGrew recanted his earlier testimony and identification. In a 2001 videotaped deposition prepared for Azania’s successive post-conviction proceeding, McGrew claimed that he had never been able to identify Azania as the man he saw place the objects in the bushes, and that he told this to police and prosecutors, but that they pressured him to make the identification at trial anyway. McGrew claimed that when he was interviewed at the Gary police station, he saw a photograph of Azania and Azania’s name on a bulletin board behind the officer who interviewed him, and that the officer pointed to the picture and told McGrew that Azania had killed a police officer. McGrew also testified that in 1982, while waiting in a room adjacent to the courtroom and preparing to testify, an armed man McGrew assumed to be a bailiff pointed Azania out to McGrew through the room’s doorway. McGrew claimed he felt threatened by the armed man’s action and the trial atmosphere, and was afraid that if he did not identify Azania his own life would be in jeopardy. McGrew testified that he could not otherwise have identified Azania, since he never saw the face of the man who placed the objects in the bushes.
Former Lake County Deputy Prosecutor James McNew, who assisted in the 1982 prosecution of Azania, testified that he was never aware that McGrew allegedly could not identify Azania. McNew denied directing anyone to coerce McGrew into identifying Azania, and testified that he did not consider Azania’s identity a problem at trial in light of the other evidence of Azania’s guilt, including security camera photographs from the bank and clothing evidence. McNew testified that to present a complete story to the jury, the State would have asked McGrew on direct examination if he could identify Azania even if the State knew McGrew could not do so.
Captain Michael Nardini interviewed McGrew the day after the murder in an interview room at the Gary police station. Nardini testified that McGrew told him McGrew could identify the man who placed the objects in the bushes. Nardini also testified that there was no bulletin board in the interview room, that he did not remember a photograph of Azania being posted anywhere in the station at the time of the interview, that at the time of the interview he did not know that Azania was a suspect in the case, and that he did not tell McGrew that Azania killed a police officer.
Allen County Deputy Sheriff Jerry Fru-chey served as a bailiff during Azania’s trial and closely meets McGrew’s description of the “armed man” who allegedly pointed Azania out to McGrew. Fruchey testified that he did not remember ever speaking with McGrew, did not tell him to identify Azania, did not point out Azania, and did not in any way threaten McGrew.
The successive post-conviction court considered all this evidence and held that McGrew’s deposition testimony was not credible and accordingly did not satisfy the nine-prong test for newly discovered evidence mandating a new trial. See Carter v. State, 738 N.E.2d 665, 671 (Ind. 2000). Substantial evidence contradicts McGrew’s recantation of his trial identification. First, as is true of all recanted testimony, McGrew’s 1982 trial testimony directly contradicts his current claims. Second, his current claims contradict his statements to police at the time of Azania’s apprehension and to Nardini the next day. Third, his claim that he was intimidated by an armed man is contradicted by Fruchey. *1263This issue turns on credibility of witnesses. The successive post-conviction court viewed McGrew and the other post-conviction witnesses and found that his recantation was not credible. That finding is not clearly erroneous, and is accordingly affirmed.
Conclusion
The successive post-conviction court erred when it concluded that Allen County’s computerized jury selection system substantially complied with Indiana Code section 33-4-5-2(c) on the facts of this case. The court did not err when it denied Azania’s motions for a change of judge and his claim for a new trial based on the prosecution’s use of allegedly false testimony. Accordingly, we vacate Azania’s death sentence and remand to the trial court for new penalty phase hearings, or, if the prosecution elects not to pursue the death penalty, for sentencing.
SULLIVAN and RUCKER, JJ., concur. SHEPARD, C.J., dissents with separate opinion, in which DICKSON, J., concurs. DICKSON, J., dissents, believing that this accidental, inadvertent, and impartial exclusion of jurors did not undermine Allen County’s essential substantial compliance with the statutory method for selection of jury pools, and that any error was harmless in light of the overwhelming evidence of aggravators which have previously led two separate penalty phase juries to unanimously recommend the death sentence.. For a detailed account of the robbery and killing, see Averhart v. State, 470 N.E.2d 666, 673-75 (Ind.1984).
. For example, in a hypothetical county comprised of 1100 registered voters evenly distributed across the county's 10 townships, the program would first determine that 10% of the registered voters, or 110 voters, resided in each township. If the requested size of the master jury pool was 200 jurors, the program would next determine that 20 jurors (10% of 200) were needed from each township to ensure proportional representation of that township in the master pool. As to hypothetical Township A, the program would divide the total number of registered voters in Township A(110) by the number of jurors needed from Township A(20), to determine that 5.5 voters should be placed into each of Township A's 20 "selection groups.” Next, the program would truncate 5.5 to the integer 5, and then take the first 5 voters on the list and select one, then take the next five voters on the list and select one, and so on. The result would be that 22 voters from Township A would be included on the list, rather than the 20 required for proportional representation.