OPINION

{¶ 1} Plaintiffs-appellants, Yolanda Albaugh et al., appeal from a judgment of the Franklin County Court of Common Pleas granting judgment in favor of defendant-appellee, the City of Columbus, Division of Police, on appellants' sexual discrimination claims. For the following reasons, we affirm that judgment.

{¶ 2} In 1997, appellants, Albaugh and 96 other persons, filed a complaint against appellee in the Franklin County Court of Common Pleas. Appellants were civilian employees working as communications technicians for appellee. The complaint alleged that appellee's policy of denying appellants' second day-off overtime constituted disparate impact discrimination on the basis of their sex in violation of R.C. 4112.02. Second day-off overtime allows an employee to be compensated at double his or her normal rate of pay if the employee works on what would normally be that employee's second day off. The complaint alleged that sworn employees (police officers), who are predominately male, were permitted to work second day-off overtime while non-sworn civilian employees, who are predominately female, were not allowed to work second day-off overtime. The suit sought, among other things, back pay, lost fringe benefits, compensatory and punitive damages, as well as damages for emotional distress.

{¶ 3} On March 30, 1998, appellee filed a motion for summary judgment based upon appellants' alleged failure to establish a prima facie case of sexual discrimination under the disparate impact theory of discrimination. The trial court granted appellee's motion, holding that, pursuant to Wards Cove Packing Co., Inc. v. Atonio (1989), 490 U.S. 642, appellants' statistical evidence was not relevant to their disparate impact claims because the statistics compared very different job classifications (civilian vs. sworn employees). Therefore, appellants could not establish a disparate impact discrimination claim. The trial court also noted that appellants had not overcome appellee's stated business justification for their policy: lower payroll costs.

{¶ 4} On appeal to this court, we reversed the trial court's decision and found that appellants could bring a disparate impact sexual discrimination claim even though the job classifications at issue were different. Albaugh v. Columbus Division of Police (1999),132 Ohio App.3d 545. We held that the reasoning in Wards Cove Packing was not applicable under the facts presented. Id. at 554. Accordingly, we found that appellants' statistical evidence, despite the differing job classifications, was relevant to their claim that the facially neutral policy of allowing only sworn employees the opportunity to work second day-off overtime had an adverse impact on females. Id. at 554. We further noted that a genuine issue of material fact existed as to the legitimacy of appellee's stated business justification for the policy. The case was remanded to the trial court for further proceedings.

{¶ 5} On remand, appellants' claim was adjudicated by bench trial for the period between January 1991 through March 1996. At the close of appellants' case, appellee moved for a directed verdict for the years 1991 through 1994. The trial court granted appellee's motion based upon appellant's failure to present sufficient evidence showing a disparate impact on civilian employees during that period of time. Following the trial, the trial court issued findings of fact and conclusions of law and awarded judgment in favor of appellee for the remaining time period (1995 through March 1996). The trial court concluded that appellants failed to show that appellee's facially neutral employment policy caused a disparate impact on civilian employees. The trial court also held that appellants failed to show a valid statistical disparity, noting that appellants' putative expert witness was not an expert in the field of statistics, and that his conclusions were flawed because they were based on invalid data. Lastly, the trial court concluded that, even if there were a disparity, appellee demonstrated a legitimate business purpose.

{¶ 6} Appellants appeal, assigning the following errors:

{¶ 7} "1. Whether, on the evidence adduced at trial, the court below erred by finding that plaintiffs failed to prove, through lay and admissible expert testimony, documents and mathematical calculation, an adverse impact from defendants' overtime policy in 1995.

{¶ 8} "2. Whether, on the evidence adduced at trial, a reasonable fact-finder could find an adverse impact from defendants' overtime policy for years other than 1995 and a directed verdict on those years was therefore improper.

{¶ 9} "3. Whether, as required by law of the case in the Court of Appeals opinion, defendants met their burden of proof to demonstrate a legitimate business justification in light of the disparate impact from their overtime policy.

{¶ 10} "4. Whether the Court's Findings of Fact and Conclusions of Law are clearly erroneous as they pertain to the identified employment practice required under disparate impact discrimination and the statistical evidence."

{¶ 11} Before addressing appellants' assignments of error, we must first discuss the nature of a disparate impact discrimination claim. Disparate impact discrimination involves employment practices that are facially neutral in their treatment of different groups but fall more harshly on one group. See Hazen Paper Co. v. Biggins (1993), 507 U.S. 604,609. Proof of discriminatory motive is not required under this theory of discrimination. Id.; Watson v. Fort Worth Bank and Trust (1988),487 U.S. 977, 986-987, citing Griggs v. Duke Power Co. (1971),401 U.S. 424. In a disparate impact case, a plaintiff must begin by identifying the specific employment practice that is challenged and that is allegedly responsible for any observed statistical disparity. Wards Cove Packing Co., at 656, quoting Watson, at 994. "Once the employment practice at issue has been identified, causation must be proved; that is, the plaintiff must offer statistical evidence of a kind and degree sufficient to show that the practice in question has caused" the alleged discrimination. Watson, at 994; Dunnigan v. Lorain, Lorain App. No. 02CA008010, 2002-Ohio-5548, at ¶ 18. If the plaintiff sets forth this prima facie case, the defendant-employer then has the burden to show a business justification for the practice. See Wards Cove Packing, at 659. If the defendant succeeds in demonstrating business necessity, the plaintiff has the opportunity to show that another policy or practice, without a similarly undesirable effect, would also equally serve the employer's legitimate business interests. Id. at 660-661.

{¶ 12} We now turn to appellants' first assignment of error. Appellants contend that they submitted sufficient statistical evidence at trial to show appellee's second day-off overtime policy caused an adverse impact on appellants from January 1995 through March 1996. This statistical evidence was presented through the testimony of appellants' expert witness, David F. Sharma. Sharma is a chartered life underwriter and a chartered financial consultant with experience in insurance risk analysis. He based his statistical analysis on a summary of W-2 payroll records compiled by Donna Jarrell, a former communications technician and a named plaintiff. Jarrell analyzed the payroll records of appellee's employees to determine the amount of second day-off overtime worked by civilian and sworn employees for the period in question. Her summary counted the number of employees in each classification, the gender of those employees and the number of second day-off overtime hours worked by each classification. She did not attempt to determine the reasons why these employees worked second day-off overtime hours.

{¶ 13} Using Jarrell's figures, Sharma applied a standard deviation analysis. The United States Supreme Court has approved a standard deviation analysis as a method of proving disparate impact discrimination. Castaneda v. Partida (1977), 430 U.S. 482. Standard deviation is a measure of the extent to which an observed result is likely to vary from an expected result. The larger the number of standard deviations an observed result is from an expected result, the lower the probability that the observed result is random. More specifically, in the context of disparate impact discrimination claims, the larger the number of standard deviations, the more likely the observed result is the product of discrimination rather than chance. See Jefferson v. Morgan (C.A.6, 1992), 962 F.2d 1185, 1189. A disparity of two or three standard deviations has been held to be sufficient to establish a prima facie case of disparate treatment. See Castaneda, at 496 n. 17; Hazelwood School Dist. v. United States (1977), 433 U.S. 299, 308 n. 14, 311 n. 17.

{¶ 14} Sharma testified that, for the year 1995, the standard deviation of the number of civilian employees working second day-off overtime versus the sworn employees working second day-off overtime was a minus 13. The standard deviation of the total hours of second day-off overtime worked between the two jobs was a minus 60. Sharma testified that, given these deviations from the optimum, a female employee's chance of working second day-off overtime was infinitely less than that of a male employee. Sharma performed this analysis only for the year 1995. Appellants contend that this testimony was sufficient to establish a prima facie case of disparate impact sexual discrimination for that year. Although Sharma's testimony, standing alone, could be sufficient to establish a statistical disparity, the trial court disregarded Sharma's testimony and gave it no weight.

{¶ 15} Courts are not obligated to assume that statistics are reliable. Watson, at 996. It is for the trier of fact, in this case the trial court, to afford testimony the weight it deserves and determine the credibility of the witness. State v. Nemeth (1998), 82 Ohio St.3d 202,210; Clark v. Clark (Sept. 1, 1998), Franklin App. No. 97APF10-1360; Charles v. Anthony (Sept. 15, 1992), Franklin App. No. 92AP-51. A trier of fact may accept or reject any or all of the testimony of any witness, including testimony of an expert witness. Croft v. State Farm Mut. Auto. Ins. Co. (Jan. 8, 2002), Allen App. No. 1-01-72. This decision is within the exclusive province of the trier of fact and will not be disturbed absent an abuse of discretion. Clark, supra.

{¶ 16} There is no basis to conclude that the trial court abused its discretion in disregarding Sharma's testimony. It was undisputed that the numbers Sharma relied on were incomplete, as Jarrell testified that she was missing documents and could not account for some employees and any second day-off overtime hours they may have worked. Sharma admitted that, if the numbers he relied on were incorrect, his analysis would be incorrect.

{¶ 17} Moreover, the numbers Jarrell compiled did not make a distinction between voluntary second day-off overtime and mandatory second day-off overtime required by a court order. Although disparate impact analysis may be applied to different job classifications under some circumstances, the trial court could consider differences in duties and responsibilities between job classifications in determining whether discrimination occurred, or whether there was some non-discriminatory justification for the statistical disparity. Swiggum v. Ameritech Corp. (Sept. 30, 1999), Franklin App. No. 98AP-1031, quoting Adams v. Indiana Bell Telephone Co. (S.D.Ind. 1998), 2 F. Supp.2d 1077, 1098 (finding statistical evidence flawed where the evidence failed to take into account variables other than age).

{¶ 18} Here, the trial court noted the difference in job responsibilities between appellants and police officers. Police officers are often needed to testify in court proceedings. Therefore, they are routinely subpoenaed. Consequently, a police officer may be called to testify on his or her second day off, resulting in second day-off overtime hours. In fact, Jarrell did not contest the fact that sworn employees worked more than 20,000 of the 23,000 hours of mandatory second day-off overtime calculated for 1995 due to subpoenas. Appellee has no control over overtime hours worked because of the need for trial testimony. In contrast, appellants are rarely required to testify in court. Because Sharma's statistical analysis failed to take this significant variable into account, the trial court did not abuse its discretion in disregarding his conclusions. See, also, Goad v. Sterling Commerce, Inc. (June 13, 2000), Franklin App. No. 99AP-321 (disregarding statistical evidence as flawed that did not take into account factors other than sex in finding discrimination); Penk v. Oregon State Bd. of Higher Edn. (C.A.9, 1987), 816 F.2d 458 (affirming lower court's discounting of plaintiff's statistical evidence due to flaws in data); Johnson v. United States Dept. of Health Human Serv. (N.D.Ohio. 1992), 1992 WL 675943 (disregarding plaintiff's flawed statistics). Therefore, the trial court did not err in concluding that appellants failed to demonstrate a statistical disparity for the year 1995 sufficient to support a finding of disparate impact sexual discrimination.

{¶ 19} Furthermore, appellants' reliance on a guideline promulgated by the Equal Employment Opportunity Commission is misplaced. See 29 C.F.R. § 1607.40(D). The guideline, known as the four-fifths rule, has been looked to by courts for guidance in gauging the significance of statistical evidence of discrimination. Equal Employment Opportunity Comm. v. Joint Apprenticeship Committee of the Joint Industry Bd. of the Electrical Industry (C.A.2, 1999), 186 F.3d 110, 118; Waisome v. Port Auth. of New York and New Jersey (C.A.2, 1991), 948 F.2d 1370,1374-1376. It is not binding on courts, but is a "rule of thumb." Id. Here, however, because the trial court did not abuse its discretion in completely disregarding appellants' statistical evidence, the four-fifths rule does not apply.

{¶ 20} Appellants' first assignment of error is overruled.

{¶ 21} By their second assignment of error, appellants contend that the trial court erred in granting appellee's motion for directed verdict on appellants' discrimination claims for the years 1991 through 1994. In reviewing the trial court's decision, we note that a motion for a directed verdict will be granted only after construing the evidence most strongly in favor of the party against whom the motion is directed and finding that, upon any determinative issue, reasonable minds could only reach a conclusion adverse to such party. Swiggum, supra. We review the trial court's grant of a directed verdict de novo. Titanium Industries v. S.E.A., Inc. (1997), 118 Ohio App.3d 39, 47-48.

{¶ 22} Although appellants presented statistical data for the years 1991 through 1994, they offered no standard deviation analysis of this data or any other analysis relevant to a disparate impact claim. Without proper analysis, these statistics are meaningless. Wingfield v. United Technologies Corp. (D.Conn., 1988), 678 F. Supp. 973, 981; Foster v. Tandy Corp. (C.A.4, 1987), 828 F.2d 1052, 1057 (noting that "raw statistics devoid of any context which relates those statistics to the alleged discriminatory practice are of minimal probative value"). Moreover, these statistics suffered from the same flaw as the 1995 data analyzed by Sharma. The 1991 through 1994 data did not account for differences in duties and responsibilities of the two job classifications. Therefore, the trial court did not err in directing a verdict in favor of appellee for the years 1991 through 1994. Appellants' second assignment of error is overruled.

{¶ 23} Because appellants' statistical evidence was flawed and was properly disregarded by the trial court, appellants failed to demonstrate a statistical disparity caused by appellee's policy, and, therefore, failed to present a prima facie case of disparate impact discrimination for the liability period in question. The disposition of appellants' first and second assignments of error render appellants' third and fourth assignments of error moot. App.R. 12(A).

Judgment affirmed.

BRYANT and LAZARUS, JJ., concur.