*308 OPINION AND ORDER
I. INTRODUCTION
Before me is defendant’s motion for summary judgment of plaintiffs’ disparate impact discrimination claims. Because plaintiffs have failed to establish that defendant’s selection process had a statistically significant adverse impact on members of plaintiff class who belong to various statutorily protected groups 1, defendant’s motion is granted.2
II. FACTUAL BACKGROUND
Plaintiffs’ class action against defendant, Pepsi-Cola Metropolitan Bottling Company, Inc. (“Pepsi” or the “Company”) arises out of the administration of the Company’s Hourly Selection Process (the “HSP”) in the summer of 1990.3 All members of plaintiff class were formerly employed at Pepsi’s Exeter Avenue facility (“Exeter”). The HSP was administered to all candidates for employment as Production Technicians (“Technicians”) at Pepsi’s newly created Detroit Distribution Center (“DPC”). Plaintiffs allege that the HSP discriminated against them on the basis of their race, gender and age in violation of Michigan’s Elliott-Larsen Civil Rights Act.4 Jurisdiction is premised on the existence of a federal question.5
The HSP was administered to 189 candidates, including Exeter employees and outside candidates (persons who either never worked for Pepsi or did not work at the Exeter facility), as a two-phase selection device. The first phase was comprised solely of a written or “cognitive” test.6 Those candidates who achieved an average percent correct score of 55% or higher passed the written test and were given the opportunity to proceed to the second phase, which consisted of both a group assessmeni/problem-solving exercise and a structured interview (“GAE” phase).7 All candidates were required to successfully complete both phases of the HSP in order to be considered for the nineteen (19) Technician positions.
Of the 189 individuals (112 minorities/70 *309non-minorities 8; 44 women/145 men; and 52 persons forty-or-older/137 persons under forty) who participated in the HSP, 131 (72 minorities/52 non-minorities; 31 women/100 men; and 22 persons forty-or-older/109 persons under forty) successfully completed the first phase and were permitted to participate in the second phase. Of the 1209 candidates (67 minority/46 non-minority; 29 women/91 men; 20 persons forty-or-older/100 persons under forty) who participated in phase two, 19 candidates (10 minority/9 non-minority; 3 women/16 men; 1 person forty-or-older/18 persons under forty) were offered employment as DPC Technicians.
III. DISPARATE IMPACT DISCRIMINATION
Prohibited race discrimination includes not only overt discrimination, but also practices or procedures which are neutral in form but discriminatory in operation, i.e., disparate impact discrimination. Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 645, 109 S.Ct. 2115, 2118-19, 104 L.Ed.2d 733 (1989). In disparate impact discrimination cases such as this, an employment practice may be deemed violative of Title VII [and Elliott-Larsen] where the employment practice has a statistically significant adverse impact on members of a protected group. Id. at 646, 109 S.Ct. at 2119. An adverse result is statistically significant where the disparity between the expected result and the actual result exceeds two to three standard deviations.10 Hazelwood School District v. U.S., 433 U.S. 299, 309, n. 14, 97 S.Ct. 2736, 2742, n. 14, 53 L.Ed.2d 768 (1977) (citing Castaneda v. Partida, 430 U.S. 482, 496-497, n. 17, 97 S.Ct. 1272, 1281, n. 17, 51 L.Ed.2d 498 (1977)).
To establish a prima facie case of disparate impact discrimination, plaintiffs must (1) identify a specific employment practice and (2) offer reliable statistical evidence of deficiencies sufficiently substantial to show that the practice has caused the exclusion of applicants because of their membership in a protected group. Wards Cove, 490 U.S. at 657, 109 S.Ct. at 2124-25. It is not enough for plaintiffs to identify a statistical disparity; they must link the disparity to the specific practices or procedures claimed to have caused the disparity. Id. After plaintiffs make such a showing, the burden of production shifts to defendant to show that each challenged practice “serves, in a significant way, [its] legitimate employment goals.” Id. at 659, 109 S.Ct. at 2126. If defendant meets this burden of production, plaintiffs can only rebut such evidence by coming forward with an equally effective alternative selection process that would result in a less disparate impact or proving that specific HSP procedures are not job-related. Id. at 660-61, 109 S.Ct. at 2126-27. The burden of persuasion remains with the plaintiff at all times.11
Because plaintiffs have not put forth reliable statistical proof that either HSP procedure had a statistically significant adverse impact on minorities or women, plaintiffs, as a matter of law, have failed to establish a prima facie case of disparate impact discrimination on the basis of race or gender. Defendant is entitled to summary judgment as a matter of law. Although the statistical analysis reveals that the written test had a statistically significant adverse impact on persons forty or older, the HSP was properly validated and job-related. Plaintiffs failed to show that the HSP was not job-related or that a viable alternative exists which would not have an adverse impact on the plaintiff class. Because plaintiffs failed to carry their burden of production, as relates to the disparate impact age discrimination claim, summary judgment will be entered in favor of Pepsi.
*310A. Identification of the Speciñc Employment Practice or Procedure.
To establish a prima facie case of disparate impact discrimination, plaintiffs must first identify the specific facially neutral employment practice which they allege had a statistically significant adverse effect on them as members of various protected groups. Id. at 645-46, 109 S.Ct. at 2118-19. Because each phase of the HSP acts as a pass/fail barrier, it must be analyzed separately for adverse impact. See also Taylor v. James River Corp, 51 Fair Empl.Prac.Cas. (BNA) 893, 898, 1989 WL 165953 (S.D.Ala. 1989) (written test and structured interview analyzed separately for disparate impact).
Plaintiffs attack the HSP and its individual components. Plaintiffs challenge: (1) the written test as having a disparate impact as to age; (2) the GAE as having a disparate impact as to age, race and gender under the 80% Rule on Exeter and outside candidates combined; and (3) the GAE as having a disparate impact as to age, race, and gender on Exeter candidates only, also applying the 80% Rule.12 Plaintiffs have met the first requirement of their prima facie case by identifying the specific employment practice responsible for the alleged adverse impact.
B. Statistically Siynifícant Adverse Impact.
To meet the next requirement of their prima facie case, plaintiffs must demonstrate a statistical disparity and link the disparity to the specific practice(s) claimed to have led to the disparity. For statistical evidence to be probative, the statistical pool or sample used must logically be related to the employment decision at issue and the statistical method applied to the pool or sample must be meaningful and suitable under the facts and circumstances of the case. Wards Cove, 490 U.S. at 650, 109 S.Ct. at 2121; Hazelwood School Dist. v. U.S., 433 U.S. 299, 308, 97 S.Ct. 2736, 2741-42, 53 L.Ed.2d 768 (1977).
1. Appropriate labor poollstatistical sample.
In cases where the employment practice is a test, the actual examinees constitute the most logical statistical pool. Richardson v. Lamar County Bd. of Educ., 729 F.Supp. 806, 815 (M.D.Ala.1989). There is no requirement that a statistical showing must always be based on analysis of actual applicants. The pool must be one which is best able to reflect that the questioned practice has caused the exclusion of applicants for jobs because of their membership in a protected group. Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 977, 108 S.Ct. 2777, 2779, 101 L.Ed.2d 827 (1988); Dothard v. Rawlinson, 433 U.S. 321, 330, 97 S.Ct. 2720, 2727, 53 L.Ed.2d 786 (1977). When a test, which is part of a selection process, operates as a pass-fail barrier, the proper approach is to compare the pool of those taking the test with those passing it. Connecticut v. Teal, 457 U.S. 440, 452, 102 S.Ct. 2525, 2533, 73 L.Ed.2d 130 (1982).
The appropriate statistical pool in this case is all participants at each phase of the HSP in July and August of 1990. The statistical pool for analysis of the results of the written test is all 189 candidates. The appropriate pool for analysis of the results of the GAE phase is the 12013 candidates who passed the written test and actually participated in the second phase.
Plaintiffs’ statistical pool consisted of the group of Exeter employees who participated in the GAE and who were not offered jobs at the DPC. Plaintiffs argue that the former Exeter employees constituted the appropriate statistical labor pool to gauge disparate impact because, unlike outside candidates, Exeter employees were assessed and interviewed by supervisors familiar with their work records and likely biased by the Exeter plant manager’s feeling about the group *311work ethic.14 Thus, they argue the GAE, as administered to former Exeter employees, was a different selection process than that given to outside candidates.15 Plaintiffs’ method of determining the appropriate statistical pool is incorrect and was rejected in Williams v. Mead Coated Bd., Inc., 836 F.Supp. 1552, 1565-66 (M.D.Ala.1993), and Lopez v. Laborers Int’l Union Local # 18, 987 F.2d 1210, 1216 (5th Cir.1993). Because plaintiffs use the wrong statistical pool in analyzing the results of phase one of the HSP, the results of their analyses are not probative of disparate impact on either of the protected groups at issue. Williams, 836 F.Supp. at 1565-66; Lopez, 987 F.2d at 1216.
2. The appropriate statistical method.
Defendant’s expert applied the Fisher’s Exact test (“Fisher’s Exact”) to the total candidates at each phase of the HSP. The Fisher’s Exact is the most appropriate form of statistical analysis to use in a ease, such as this, where the sample size is relatively small. See e.g. Bridgeport Guardians, Inc. v. City of Bridgeport, 735 F.Supp. 1126, 1131-32 (D.Conn.1990).16
Plaintiffs argue that the correct statistical method is the jéths or 80% Rule which provides that:
A selection rate for any race, sex, or ethnic group which is less than four-fifths ® (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact.
29 C.F.R. § 1607.4(D). Although the léths Rule is not binding upon the courts, it is an objective standard which the Equal Employment Opportunity Commission (“EEOC”) and other agencies employ to trigger investigations and to provide evidence of disparate impact. Thus, many courts give deference to the jéths Rule and its results. The guidelines, however, specifically recognize limitations in the use of the rule in certain cases:
Smaller differences in selection rate may nevertheless constitute adverse impact, where they are significant in both statistical and practical terms ... Greater differences in selection rate may not constitute adverse impact where the differences are based on small numbers and are not statistically significant ...
29 C.F.R. § 1607.4(D) (Emphasis added). Thus the %ths Rule is recognized as a rule of thumb to be used in conjunction with standard deviation or other statistical evidence.
Although the Sths Rule has been used in this circuit, with sample sizes smaller than 189,17 plaintiffs’ analysis fails because it does not combine the jtths test [80% Rule] with standard deviation or other statistical evidence. Plaintiffs’ results are not expressed in statistical terms and are not probative evidence of disparate impact discrimination on the basis of race or gender.18 Plaintiffs’ evidence does not establish a prima facie case of race or gender disparate impact discrimination. Thus, defendant’s suggestion, which does yield statistical results, or some other statistical method known to be sensitive to *312smaller sample sizes,19 is appropriate in this case.
3. Statistical significance of the disparities.
Plaintiffs’ expert, Dr. Thomson, performed two different analyses, one using the former Exeter employees as the labor pool and the other using all persons who participated in the GAE phase only. Because the former Exeter employees are an improper pool,20 and because plaintiffs expert did not perform a statistical analysis,21 the results of that analysis are irrelevant. Dr. Thomson’s tables on all persons who participated in the GAE phase do not give standard deviations or other information which is indicative of statistical significance, and therefore probative evidence of disparate impact discrimination. Additionally, Dr. Thomson specifically concludes, in his affidavit, that the shortfalls observed (based on percentages) are not statistically significant:
[Although there was a shortfall in terms of the share of the offers made to minority employees,22 the chances of observing such shortfalls were not “statistically significant”. This means that the shortfalls were not so extreme as to meet the standards frequently used for determining statistical significance.
(Pls.’ Exh. A at 9.) Dr. Thomson concludes that the results do meet the %ths Rule which supports plaintiffs’ claims of disparate impact “although not at a level which is ‘statistically significant.’ ” Id. at 10. Because in his analysis of phase two, Dr. Thomson did not give his results in statistical terms, and because plaintiffs’ expert admitted that the results were not statistically significant, plaintiffs cannot establish the second element of their prima facie case. Therefore, defendant is entitled to summary judgment.
Using the Fisher’s Exact test, defendant’s expert, Dr. Siskin, analyzed the HSP to determine the extent to which each component of that selection device had a disparate impact on the basis of the candidates’ race, gender, or age. The results of his analysis are expressed in units of standard deviations:
Protected Group Phase One Phase Two
Minorities 1.31 0.51
Women 0.00 0.74
Persons 40 or over 40 4.73 1.31
The analysis of statistical evidence reveals that the HSP did not have a statistically significant adverse impact upon minorities or women. Defendant is entitled to summary judgment, as a matter of law, of plaintiffs’ disparate impact discrimination claims on the basis of race and sex.
However, defendant’s statistical evidence does reveal a statistically significant adverse impact to persons forty or older.23 Thus plaintiffs have established a prima facie case of disparate impact age discrimination.
C. The Phase One Written Test Serves A Legitimate Employment Goal.
Although the statistical evidence reveals disparate impact age discrimination, defendant has met its burden of production by “justifying” the procedure. Defendant has established that the written test is job-related and serves a legitimate employment goal. In Black Law Enforcement Officers Ass’n v. City of Akron, 824 F.2d 475, 480 (6th Cir.1987), the U.S. Court of Appeals for the Sixth Circuit stated:
The defendant may meet this burden by establishing that the procedure used measures important skills, abilities, and knowledge that are necessary for the successful performance of the job. In other words, the defendant must show, “by professionally accepted methods,” [that the test is] predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job *313or jobs for which candidates are being evaluated. (Citations omitted)
The written test phase of the HSP has been shown to be correlated with important and relevant elements of work behavior for production jobs in a Pepsi-Cola bottling plant.24 PDI developed and implemented a criterion-related validity study in accordance with technical standards set forth in the EEOC’s Uniform Guidelines on Employee Selection Procedures. 29 C.F.R. § 1607.14.25
D. Plaintiffs Have Failed To Prove That The Written Test Is Not Job-Related And Have Failed To Come Up With A Viable Alternative.
As discussed above, plaintiffs do not attack the business relatedness of phase one of the HSP (the written test); the phase which had a statistically significant adverse impact on persons forty or older. However, plaintiffs may, in the alternative, rebut defendant’s defense by providing an alternative to the written test which would not have a statistically significant adverse impact on the protected class. Wards Cove, 490 U.S. at 660, 109 S.Ct. at 2126-27. Plaintiffs’ suggested alternative is that defendant should have simply retrained Exeter employees to operate the newer equipment at the DPC. This, however, is not a selection device and is not a viable alternative. Because plaintiffs failed to give an explanation as to how defendant was to choose the most qualified 19 candidates for retraining out of the 70 Exeter employees, or to legally justify why nonExeter candidates should not be considered for the Technician positions, defendant is entitled to summary judgment because plaintiffs are unable to rebut defendant’s defense. See Brunet v. City of Columbus, 1 F.3d 390, 411 (6th Cir.1993) (alternative procedure which does not enable employer to select most qualified candidates is not “equally effective”); Bernard v. Gulf Oil Corp., 890 F.2d 735, 744 (5th Cir.1989) (plaintiffs’ proposed device, which tests merely for competence and not best hires, is not viable alternative).
III. CONCLUSION
Plaintiffs have not established that either phase of the HSP had a statistically significant adverse impact upon minorities or women. Although the written portion of the HSP had a statistically significant adverse impact on persons forty or older, defendant has established that the written test serves legitimate employment goals. Plaintiffs have failed to assert a viable alternative to the written test or to prove that the test is not job-related. Thus, plaintiffs have failed to carry their burden of production.26 For the *314reasons discussed above, Defendant’s Motion for Summary Judgment is Granted.
IT IS SO ORDERED.