Cotton appeals the district court’s order granting summary judgment in favor of the City of Alameda, its police department, and its chief of police (Alameda) on Cotton’s age discrimination claim brought pursuant to the Age Discrimination in Employment Act of 1967 (ADEA), 29 U.S.C. §§ 621-634 . The district court concluded that Alameda produced evidence that it had legitimate, nondiscriminatory reasons for hot hiring Cotton, and that Cotton had failed to produce evidence that these reasons were a pretext for discrimination. We have jurisdiction pursuant to 28 U.S.C. § 1291 , and we affirm.

Cotton is a 47 year old police officer with the Bay Area Rapid Transit (BART) police department. In the fall of 1981, he applied for a position with the Alameda police department and passed the structured oral examination with a score of 66.71%. Nevertheless, in January 1982, he signed a waiver removing his name from the eligibility list. Cotton applied again in March 1983, but failed to achieve a passing score *1247 of 60%. He applied for a third time in July 1983, this time passing with a 65% score. Alameda conducted a background investigation on Cotton. Although Alameda did not hire Cotton, it did hire seven other applicants with higher test scores. In April 1984, Cotton applied a fourth time, scored 62.50%, and was placed on the eligibility list. Again, Alameda hired four applicants with higher scores, but did not hire Cotton.

On November 1, 1984, Cotton filed a complaint in federal district court alleging age discrimination in violation of ADEA and state law. Alameda moved for summary judgment on the ground that it had stated legitimate reasons for not hiring Cotton and Cotton had not produced any evidence that the reasons were a pretext for discrimination. The district court granted this motion for summary judgment and then dismissed the pendent state law claims. We review de novo the decision to grant summary judgment. Darring v. Kincheloe, 783 F.2d 874, 876 (9th Cir.1986).

Under the ADEA, it is “unlawful for an employer ... to fail or refuse to hire ... any individual [who is at least 40 years of age but less than 70] ... because of such individual’s age.” 29 U.S.C. §§ 623 (a), 631(a). A litigant can establish a violation of ADEA using either of two related theories: he can show disparate treatment — that he suffered intentional discrimination in employment because of his age, or he can show an adverse or disparate impact — that facially neutral employment practices in fact treated more harshly employees in the protected age group. See Sakellar v. Lockheed Missiles and Space Co., 765 F.2d 1453, 1455-56 (9th Cir.1985) (Sakellar), cert. denied, — U.S. —, 106 S.Ct. 856 , 88 L.Ed.2d 896 (1986). The district court liberally construed Cotton’s arguments as stating disparate treatment and disparate impact theories. We consider each theory in turn.

To establish a prima facie ADEA case under the disparate impact theory, “a plaintiff must show that a neutral practice had a significant discriminatory impact upon a protected class,” Sakellar, 765 F.2d at 1456 , of which the plaintiff is a member. Since our focus is on the consequences of employer action, statistical evidence usually provides the only rational means of determining whether a facially neutral employment practice had a discriminatory impact.

Cotton challenges Alameda’s practice of bifurcating the application process. Alameda processed applicants who were employed for a year or more by another police department as “lateral” applicants and all others as “entry level” applicants. Lateral applicants were treated in all respects like entry level applicants, except that lateral applicants were not required to pass the written test and the writing skills test, and were given a different version of the structured oral examination. Following the administration of the oral examination, Alameda prepared separate lists ranking the lateral and entry level applicants according to their test scores. The lists were effectively combined, however. Alameda first evaluated for employment those applicants with the highest test scores. A lateral applicant with a given test score was considered only after all lateral and entry level applicants with higher scores had been evaluated, but before any applicant with lower scores was considered.

Cotton contends that lateral applicants as a group must be older than entry level applicants and that Alameda did not hire lateral applicants as frequently as entry level applicants. Therefore, he argues, a genuine issue of material fact exists as to whether the bifurcated process had a disparate impact on older applicants. We conclude, however, that the evidence does not establish a prima facie case of disparate impact. First, Cotton has produced no statistical evidence with respect to applicants over age 40. The record permits us to conclude only that Alameda had at least one lateral applicant over age 40, namely Cotton. Cotton has provided no evidence *1248 with respect to the ages of other lateral and entry level applicants. We cannot draw any conclusions from a statistical base of one. Second, even if we accept Cotton’s unsupported assertion that lateral applicants are older than entry level applicants, the only statistical evidence presented suggests that the bifurcated application process did not adversely impact the older group. During the relevant period, Alameda hired 8% of the lateral applicants, but only 2% of the entry level applicants.

. We conclude that Cotton has not raised a genuine issue with respect to whether Alameda’s bifurcated application scheme adversely impacted older applicants. Therefore, the district court properly granted summary judgment on the disparate impact theory.

To prevail on a disparate treatment theory, a plaintiff must prove intentional discrimination. A plaintiff makes out a prima facie case of intentional discrimination under the ADEA if he demonstrates that he was within the protected class of individuals between forty and seventy years of age, that he applied for a position for which he was qualified, and that a younger person with similar qualifications received the position. Sakellar, 765 F.2d at 1455 . Establishing a prima facie case raises an inference of discrimination which the employer can rebut by demonstrating that it had a legitimate, nondiscriminatory reason for its decision. Id. The plaintiff then must come forward with evidence that the employer’s reason is a mere pretext to conceal its discriminatory motive. Id.

Alameda assumed for purposes of its summary judgment motion that Cotton could make out a prima facie case. Alameda, however, presented three reasons for not hiring Cotton: (1) Cotton had an unstable employment history, (2) candidates more promising than Cotton were available, and (3) Alameda did not receive encouraging reports from those who had worked with Cotton. No one disputes that these are legitimate, nondiscriminatory reasons. All argument has therefore focused on

whether these reasons are credible or instead merely pretexts.. A plaintiff can show pretext in two ways, “either [1] directly by persuading the court that a discriminatory reason more likely motivated the employer or [2] indirectly by showing that the employer’s proffered explanation is unworthy of credence.” Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 256 , 101 S.Ct. 1089, 1095 , 67 L.Ed.2d 207 (1981) (Burdine). Of course, a motion for summary judgment does not require a plaintiff to prove that the employer’s reasons are pretextual. Rather, he need only “tender a genuine issue of material fact as to pretext in order to avoid summary judgment.” Steckl v. Motorola, Inc., 703 F.2d 392, 393 (9th Cir.1983). To do so, he must “produce ‘specific facts showing that there remains a genuine factual issue for trial’ and evidence ‘significantly probative as to any [material] fact claimed to be disputed.’ ” Id., quoting Ruffin v. County of Los Angeles, 607 F.2d 1276, 1280 (9th Cir.1979), cert. denied, 445 U.S. 951 , 100 S.Ct. 1600 , 63 L.Ed.2d 786 (1980).

We begin by considering the direct evidence avenue for showing pretext. Cotton’s only direct evidence of discrimination relates to Alameda’s past employment policies. Prior to 1979, Alameda considered age to be a job related requirement for police officers and consequently refused to hire anyone over the age of 35. Since that practice was terminated before Cotton’s applications, this fact sheds little or no light on Alameda’s employment policies at times relevant to this case. Standing alone, it does not reasonably permit the inference that Alameda was more likely motivated by Cotton’s age than by the reasons it articulated for declining to hire Cotton.

We are therefore left to determining whether Cotton can proceed down the second avenue for showing pretext by producing sufficient indirect evidence. We must consider whether a genuine issue exists with respect to the credibility of each of the employer’s proffered explanations. Cotton argues that the first reason given — job instability — was not credible because he only *1249 had 4 jobs in 24 years. Taken out of context, this argument has plausibility: in some fields, an average of one job change every six years might be considered a stable rather than an unstable employment history. Nevertheless, Alameda produced the affidavit and deposition testimony of its Chief of Police who was familiar with employment patterns in the law enforcement field. He testified that Cotton had moved from the San Francisco Police Department to the Department of Fish and Game, to a job in real estate, and then to BART. The police chief thought that this pattern “indicated a lack of commitment to law enforcement and instability because the moves were not in a particular career direction.” He also stated that “leaving a municipal police department and going to a quasi law enforcement position with the fish and game ... represents a negative.”

Cotton, by contrast, has offered no evidence that his employment history was in fact stable and directed. He did not, for example, produce for comparison the prior employment histories of the applicants hired by Alameda, nor did he offer an empirical study of police officer employment patterns, nor did he offer the testimony of someone familiar with employment patterns in law enforcement who would support his contention that he had a relatively stable employment history. In short, Cotton has failed to point to any evidence from which we might infer that Alameda’s first reason “is unworthy of credence.” Burdine, 450 U.S. at 256 , 101 S.Ct. at 1095 .

The second reason Alameda gave for not hiring Cotton was that more promising candidates were available. Cotton declares that he had 24 years of experience in law enforcement and contends that most people hired by Alameda had no experience. He argues that this evidence creates a genuine dispute as to whether the other candidates were more promising. This argument, however, misses the mark. The question is not whether Cotton in the abstract had better qualifications than the other candidates. The question is whether the other candidates are more qualified with respect to the criteria that Alameda actually employs. The primary criterion used by Alameda was the applicant’s score on the structured oral examination. It is irrelevant that some other criterion, such as experience, might more accurately indicate someone’s law enforcement abilities. The ADEA does not make it unlawful for an employer to do a poor job of selecting employees. It merely makes it unlawful to discriminate on the basis of age. See Douglas v. Anderson, 656 F.2d 528, 534 (9th Cir.1981). We will not second guess the selection criterion used by an employer unless that criterion itself has a disparate impact on the protected class. In this case, Cotton agrees that the testing itself was fair. He has failed to show a disparate impact from the selection criteria.

The first time Cotton applied, he passed the examination but voluntarily removed his name from the eligibility list. On his second attempt, Cotton failed the examination. On his third attempt, he scored 65% and was placed on the eligibility list dated August 18, 1983. On October 19, 1983, Alameda hired four applicants from this list whose test scores ranged between 93.38% and 81.71%. Then in November, Alameda hired two more applicants who had scored 80.04% and a third who had scored 70.04%. On Cotton’s fourth attempt, he scored 62.50% on the test,, and was placed on the June 7, 1984, eligibility list. Alameda hired four people from this list who scored above 70%. Furthermore, for this entire period, all lateral candidates hired by Alameda scored above 80% — more than 15 percentage points above Cotton’s scores. The test results speak for themselves. There is no genuine dispute that better applicants were available according to the criterion Alameda actually employed.

The third reason Alameda gave for not hiring Cotton was that it did not receive encouraging recommendations from those who had worked with Cotton. Alameda’s background investigator, Parsons, had an informal conversation about Cotton with two BART employees, Tamisiea and Nunes. Although Parsons could recall only the substance of the conversation, he testified at his deposition that Tamisiea and *1250 Nunes described Cotton “as being a borderline employee, hinging on submarginal; would work to an output only that he felt was necessary to get by, and was a constant complainer____ They did harp on the point that he only did as much work as he felt was necessary to get by his supervisors.” Parsons also spoke with Villa, a former BART employee now employed by Alameda. Parsons testified that Villa described Cotton as being “less, than a striver ... as doing just what was necessary to get by.”

Cotton makes two attacks with respect to Alameda’s third reason for not hiring him. First, he argues that if Alameda had conducted a more thorough investigation, it would have discovered that he had a distinguished record at BART. It is irrelevant, however, that Alameda’s background investigation did not discover all there is to know about Cotton. The only issue is whether Alameda received and relied upon less than favorable reports about Cotton, as it stated. In any case, Parsons could not have readily conducted a more thorough investigation because Cotton refused to give Parsons the necessary permission to contact Cotton’s employer directly-

Second, Cotton argues that it is disputed whether Alameda received discouraging reports because Tamisiea and Nunes have denied that they made the statements attributed to them by Parsons. The record, however, does not support this argument. Tamisiea testified that he could not remember the specifics of his conversation with Parsons but that in substance he had said that Cotton “was no ball of fire.” He also testified that he did not believe he used the phrase “submarginal” because he considered Cotton “mediocre” but not “terrible.” Nunes testified that he could not remember anything about the conversation with Parsons. Finally, Villa testified that he told Parsons that Cotton was lackadaisical, not assertive, was less than a striver, aiid was a mediocre policeman. This testimony does not reasonably place Parsons’s version of the facts in doubt. If anything, Tamisiea and Villa support Parsons’s general recollection that he was told that Cotton was a mediocre police officer. We conclude that Cotton has failed to produce evidence that would permit a reasonable inference that Alameda had not received unfavorable recommendations as a result of its.background investigation.

Ill

Cotton has not produced evidence that Alameda’s bifurcated application process adversely impacts those, over 40. Neither has Cotton produced direct evidence that Alameda’s reasons for not hiring him were mere pretexts for discrimination. Finally, the evidence does not reveal a genuine dispute as to the credibility of any of Alameda’s proffered reasons. Summary judgment was appropriate.

AFFIRMED.