OPINION

B. FLETCHER, Circuit Judge:

Plaintiffs William Schechner and John Lobertini were television news reporters at KPIX-TV, one of the two San Francisco affiliates of CBS Broadcasting, Inc.1 They were laid off after CBS issued a directive requiring each of its affiliates to reduce its annual budget by ten percent. Schechner and Lobertini were sixty-six and forty-seven years old, respectively, when they lost their jobs. They brought suit alleging that KPIX discriminated against them on the basis of age and gender, in violation of California law. The district court granted KPIX's motion for summary judgment, dismissing all of Plaintiffs' claims. We affirm. We write to clarify that a plaintiff can make out a prima facie case of disparate-treatment age discrimination using statistical evidence, even where that evidence does not account for the defendant's legitimate non-discriminatory reason for the discharge.

I. BACKGROUND

This case demonstrates that reduced advertising revenues, whether caused by competition from online news outlets or our nation's economic downturn, have taken a significant toll on local television news stations. Schechner and Lobertini were performing their jobs well when KPIX laid them off. Both are experienced reporters with distinguished careers that include numerous awards. KPIX does not allege that performance issues played any role in the decision to lay off either Schechner or Lobertini.

In March 2008, KPIX was faced with the difficult task of reducing its annual budget by ten percent before the end of its first fiscal quarter. Although the budget reduction mandate came from CBS's headquarters in New York, the senior management at local stations decided how to implement the required cuts. Ronald Longinotti, the President and General Manager of KPIX-TV and Dan Rosenheim, the Vice-President and New Director at KPIX-TV were responsible for implementing the required cuts at KPIX-TV.

Longinotti and Rosenheim were responsible for hiring and firing decisions in the KPIX news department in March 2008. They had signed Schechner to a new two-year contract in 2004, when Schechner was sixty-two years old, and to new one-year contracts in 2007 and 2008, when he was sixty-five and sixty-six years old, respectively. Similarly, Longinotti and Rosenheim had signed Lobertini to a new two-year contract in October 2006, when Lobertini was forty-six years old. It is undisputed that KPIX was under no obligation to sign Schechner or Lobertini to these contracts when it did so. Longinotti and Rosenheim also made the decision to lay off Schechner, Lobertini and three other members of the KPIX "on-air" news team2 as part of a reduction in force necessitated by the CBS-mandated budget cut. The other three on-air employees that KPIX laid off were fifty-seven-year-old Tony Russomano, fifty-six-year-old Manny Ramos, and fifty-one-year-old Rick Quan.3

KPIX offered legitimate non-discriminatory reasons for its layoff decisions. Longinotti and Rosenheim testified that they first decided that news anchors would not be subject to layoff because they are the "face" of KPIX and Longinotti and Rosenheim wanted the reductions to be as invisible as possible to the viewing public. Then, they decided that they would lay off general assignment reporters based on next date of contract expiration. Rosenheim testified that they excluded "specialty reporters," meaning those focusing on a specific beat, because "they were the people that we were promoting and pushing the brand of the station." Schechner and Lobertini dispute whether KPIX followed the decision-making model that Longinotti and Rosenheim described.

Schechner and Lobertini submitted reports by expert statistician William Lepowsky. Lepowsky's reports compared the on-air talent who were laid off with the entire pool of on-air talent in the KPIX-TV news department. He concluded that "those individuals laid off, as a group, are older than the group of those not laid off, and the disparity between the two groups is statistically significant." Lepowsky found statistically significant age disparities using three different statistical methods and using a number of different groups of KPIX-TV's on-air talent. Based on his statistical analyses, Lepowsky opined that the age of KPIX's on-air talent "correlates closely" with those selected for layoff. He acknowledged that his analyses assumed that all on-air talent had an equal probability of being laid off (i.e., he did not exclude anchors from the pool of talent subject to lay off) and did not account for contract expiration date. KPIX's statistical expert, Dr. Bernard Siskin, opined that Lepowsky's report failed to account for obvious, valid and important factors because it failed to account for the decisionmaking process that KPIX said it followed.

The district court granted KPIX's motion for summary judgment on Plaintiffs' disparate treatment claim. The district court found that Schechner and Lobertini failed to make out a prima facie case of age discrimination. Specifically, the district court concluded that where a plaintiff's statistical analysis fails to preemptively account for a defendant's legitimate non-discriminatory reason for discharge, the statistical results cannot show a stark pattern of discrimination. We disagree and write to clarify that a plaintiff who relies on statistical evidence to establish a prima facie case of disparate treatment bears a relatively low burden of proof. Nonetheless, we affirm because Plaintiffs have not carried their burden at step three of the McDonnell Douglas analysis.

II. JURISDICTION AND STANDARD OF REVIEW

The district court had jurisdiction under 28 U.S.C. §§ 1332 and 1441(a). We have jurisdiction under 28 U.S.C. § 1291. Plaintiffs did not appeal the district court's grant of summary judgment on their disparate impact age discrimination claim or their gender discrimination claim. We therefore do not address these claims.

We review de novo the district court's grant of summary judgment. Earl v. Nielsen Media Research, Inc., 658 F.3d 1108, 1112 (9th Cir.2011). The court determines, viewing the evidence in the light most favorable to the non-moving party, whether there are any genuine issues of material fact, and whether the district court correctly applied the relevant substantive law. Id. "As a general matter, the plaintiff in an employment discrimination action need produce very little evidence in order to overcome an employer's motion for summary judgment." Diaz v. Eagle Produce Ltd. P'ship., 521 F.3d 1201, 1207 (9th Cir.2008) (quoting Chuang v. Univ. of Cal. Davis, Bd. of Trs., 225 F.3d 1115, 1124 (9th Cir.2000)). "The requisite degree of proof necessary to establish a prima facie case ... on summary judgment is minimal and does not even need to rise to the level of a preponderance of the evidence." Wallis v. J.R. Simplot Co., 26 F.3d 885, 889 (9th Cir.1994).

Because the district court concluded that Plaintiffs failed to establish one of the elements of a prima facie case, it did not complete the remainder of the McDonnell Douglas analysis with respect to their disparate treatment claims. See Diaz, 521 F.3d at 1208. We may affirm the district court on any grounds supported by the record. See Townsend v. Univ. of Alaska, 543 F.3d 478, 485 (9th Cir.2008).

III. DISCUSSION

Schechner and Lobertini allege that KPIX discriminated against them on the basis of age, in violation of the California Fair Employment and Housing Act (FEHA), when the station laid them off as part of its reduction in force. See Cal. Gov't Code § 12940(a). California applies the McDonnell Douglas4 burden-shifting framework and other federal employment law principles when interpreting the FEHA. See Guz v. Bechtel Nat'l, Inc., 24 Cal.4th 317, 100 Cal.Rptr.2d 352, 8 P.3d 1089, 1113 (2000); Earl, 658 F.3d at 1112.

The district court held that Schechner and Lobertini failed to make a prima facie case of age discrimination under the familiar McDonnell Douglas burden shifting framework.5 Specifically, the district court concluded that under Ninth Circuit law, when a plaintiff relies on statistical evidence as the primary support for his prima facie case of age discrimination "the three-step McDonnell Douglas analysis collapses into a single step." The district court added that "where a plaintiff's statistical analysis fails to preemptively account for a defendant's legitimate, non-discriminatory reason for discharge, the statistical results cannot show a stark pattern of discrimination unexplainable on grounds other than age." We clarify that a plaintiff's statistical evidence need not necessarily account for an employer's proffered non-discriminatory reason for the adverse employment action to make a prima facie case of discrimination.

The employee in an age discrimination case makes a prima facie case of disparate treatment "by demonstrating that he was (1) at least forty years old, (2) performing his job satisfactorily, (3) discharged, and (4) either replaced by substantially younger employees with equal or inferior qualifications or discharged under circumstances otherwise `giving rise to an inference of discrimination.'" Diaz, 521 F.3d at 1207 (quoting Coleman v. Quaker Oats Co., 232 F.3d 1271, 1281 (9th Cir. 2000)). An inference of discrimination can be established by "showing that others not in [plaintiff's] protected class were treated more favorably." Id. A plaintiff laid off during a reduction in force will generally have to rely on evidence giving rise to an inference of discrimination — often statistical evidence — because the plaintiff is unlikely to have been replaced. See id. at 1207 n. 2; see also Coleman, 232 F.3d at 1281.

There is no dispute that Schechner and Lobertini satisfied the first three elements of their prima facie case. The key issue before us is whether Lepowsky's statistical analyses necessarily fail to satisfy the fourth element because they do not account for variables related to KPIX's proffered non-discriminatory reasons for discharging Schechner and Lobertini. As the district court recognized, the rule it applied would "permit[] savvy employers to eliminate older employees from the workplace during a reduction-in-force without any consequences, the exact harm the FEHA was enacted to prevent."

A.

We begin by reviewing a number of our precedents in age discrimination cases. We have explained that when a plaintiff seeks to establish a prima facie case of disparate treatment based solely on statistics, these statistics must show a "stark pattern of discrimination unexplainable on grounds other than age." Palmer v. United States, 794 F.2d 534, 539 (9th Cir.1986) (internal quotation marks omitted) (quoting Gay v. Waiters' & Dairy Lunchmen's Union, 694 F.2d 531, 552 (9th Cir.1982)). In Rose v. Wells Fargo & Co., we found that two high-level managers who were laid off after a merger failed to satisfy this standard. 902 F.2d 1417, 1423 (9th Cir. 1990). The Rose plaintiffs submitted statistical evidence showing that employees over the age of fifty were far more likely to lose their jobs in the layoffs than younger employees. Id. We found that these disparities could be explained by the fact that older employees tended to occupy the key management positions, which were eliminated after the merger in order to avoid duplication. Id. Palmer and Rose instruct that in many cases, statistical evidence will need to address variables other than age in order to give rise to an inference of discrimination sufficient to make a plaintiff's prima facie case.

In our more recent decisions we have questioned the strength of statistical evidence offered by plaintiffs to make a prima facie case, but have nonetheless resolved the cases at step three of the McDonnell Douglas framework. See Diaz, 521 F.3d at 1209, 1214; Coleman, 232 F.3d at 1282-83.

Diaz involved claims by farm workers who were at least fifty-five years old when they lost their jobs. 521 F.3d at 1205-06. The Diaz plaintiffs submitted statistical evidence showing that the average age of the employees laid off, 48.4 years, was higher than the average age of employees retained, 38.75 years. Id. at 1208-09. We raised concerns about the statistics because (1) many of the farm's employees, including plaintiffs, were already over age 50 when they were hired; (2) the disparity in the average ages of employees laid off compared with employees retained was not "so stark as to suggest bias rather than pure chance"; and (3) the two data sets (16 workers in the laid-off group and sixteen in the retained group) were "too small to form a reliable basis for analysis." Id. at 1209. There was also evidence, however, that the supervisor who laid off the plaintiffs (and who took over after plaintiffs were hired) exhibited a preference for younger workers, in some cases even when those workers were less experienced than employees laid off. Id. at 1209-11. The disparities in the ages of employees laid off compared with those retained increased after he took over. Id. at 1210.

We suggested that the Diaz plaintiffs' statistics standing alone would be insufficient to make a prima facie case. Id. at 1208-09. But we concluded that the plaintiffs had established their prima facie case based on the statistics along with plaintiffs' other circumstantial evidence of discrimination. Id. at 1209-11. We did not say that statistical evidence must account for variables associated with the employer's neutral reason for laying off a plaintiff. Id. at 1208-10.

We similarly questioned the strength of the plaintiffs' statistical evidence in Coleman, but resolved the case at step three of the McDonnell Douglas framework. 232 F.3d at 1281-82. The plaintiffs in Coleman were three employees aged fifty-five, fifty-three, and forty-nine, who were laid off along with hundreds of other Quaker Oats employees nationwide. Id. at 1277, 1279-80. Two of the Coleman plaintiffs submitted statistical evidence showing that older employees lost their jobs at twice the rate of younger workers. Id. at 1281. We characterized the statistical evidence as "problematic" because it did not account for any variables other than age. Id. at 1281. We explained that when variables such as education and job category were considered, the statistical disparities were far less dramatic or not statistically significant. Id. Nonetheless, "despite the weakness in the evidence offered by [plaintiffs] to establish their prima facie cases, given the low threshold required, we assume[d], without deciding, that ... [plaintiffs] have established such a case." Id. at 1282. We affirmed the district court's summary adjudication based on the plaintiffs' failure to present evidence that Quaker Oats' non-discriminatory explanation for the layoffs was pretextual. Id. at 1283.

Our resolution of the Diaz and Coleman cases at step three of the McDonnell Douglas framework is consistent with our admonition that "[t]he requisite degree of proof necessary to establish a prima facie case ... on summary judgment is minimal and does not even rise to the level of a preponderance of the evidence." Wallis, 26 F.3d at 889. Consistent with our precedents, we conclude that a plaintiff who submits statistical evidence that shows a stark pattern of age discrimination establishes a prima facie at step one of the McDonnell Douglas framework. We hold that statistical evidence does not necessarily fail to establish a prima facie case because it does not address the employer's proffered non-discriminatory reasons for the discharge. We do not hold that any statistical evidence of disparate treatment, regardless of its strength, will be sufficient to establish a prima facie case.

Here, Schechner and Lobertini submitted analyses showing stark age disparities between the on-air talent who were retained and those who were laid off. This evidence was sufficient to carry their minimal burden at step one of the McDonnell Douglas framework.

B.

KPIX met its burden at step two by offering a legitimate, non-discriminatory reason for its layoffs: that it laid off general assignment reporters based on date of contract expiration. We now turn to whether Schechner and Lobertini have shown that this was mere pretext for age discrimination. We conclude that they have not.

Schechner and Lobertini argue that KPIX did not actually follow the process Longinotti and Rosenheim described to make layoff decisions. They make the following claims: (1) Joe Vazquez and Simon Perez, two general assignment reporters who were retained, had contract expiration dates earlier than the contract expiration dates of the employees laid off; (2) anchors were not in fact exempt from the layoffs because sports anchor Rick Quan was laid off and another anchor was part of the reduction in force; (3) two of the reporters who were laid off were "specialty reporters," not general assignment reporters. We address these arguments in turn.

The record does not support Plaintiffs' assertion that Vazquez and Perez had earlier contract expiration dates than the terminated employees. Perez and Vazquez did not have signed written contracts when the CBS mandate came down, but Rosenheim testified that they did have enforceable oral agreements in place. Internal CBS documents support Rosenheim's testimony and the documents Plaintiffs point to do not refute his assertion. There is no genuine issue of material fact or inference of pretext to be drawn from KPIX's retention of Vazquez and Perez.

There is no genuine issue of material fact as to whether a sixty-one-year-old anchor took voluntary termination. Longinotti acknowledges that he asked the anchor to consider this option, but that acknowledgment does not make the anchor's termination non-voluntary. Plaintiffs point to a KPIX budget document, which includes the anchor's name in a line item for "five terminated talent." The anchor's inclusion on this list does not mean that the anchor's termination was not voluntary. There is no declaration or testimony from the anchor, or other evidence in the record, regarding why the anchor took voluntary termination. We therefore conclude that the anchor's voluntary departure from KPIX does not support an inference of discrimination.

Rick Quan's termination presents a closer question, but we conclude that it does not support the conclusion that KPIX's proffered explanation for the layoffs was pretext. Rosenheim asserted that Quan was terminated "for very specific reasons having to do with the cost of the sports department and the role of sports in our news." Plaintiffs did not develop Rosenheim's testimony with regard to those "specific reasons." Quan's declaration contains no information tending either to support or refute Rosenheim's testimony. While it is true that Quan was an anchor and that his inclusion in the reduction in force tends to undermine KPIX's claim that anchors were not eligible for layoff, KPIX has consistently asserted that it made a separate business decision regarding the role of sports in its broadcasts and Plaintiffs point to no evidence to refute that assertion.

We conclude that there is a factual dispute as to whether Russomano and Lobertini were specialty reporters who should have been exempted from layoff under the system described by Rosenheim and Longinotti. Rosenheim testified that KPIX's specialty "political reporter" is Hank Plant and its "environmental reporter" is Jeffrey Schaub. But Plaintiffs presented evidence that Lobertini was a political reporter and that Tony Russomano was an environmental reporter.

We conclude, however, that even viewing the disputed facts in the light most favorable to Schechner and Lobertini, they do not support a finding of pretext. This is true largely because KPIX is entitled to a favorable "same-actor inference."

"[W]here the same actor is responsible for both the hiring and the firing of a discrimination plaintiff, and both actions occur within a short period of time, a strong inference arises that there was no discriminatory motive." Bradley v. Harcourt, Brace & Co., 104 F.3d 267, 270-71 (9th Cir.1996). The same-actor inference is "a `strong inference' that a court must take into account on a summary judgment motion." Coghlan v. Am. Seafoods Co., 413 F.3d 1090, 1098 (9th Cir.2005) (quoting Bradley, 104 F.3d at 271). The inference applies to favorable employment actions other than hiring, such as promotion. Id. at 1097. It also may arise when the favorable action and termination are as much as a few years apart. Id.

Schechner acknowledged in his deposition that Rosenheim accommodated Schechner's request to change from full-time to part-time employment in 2006, despite Rosenheim's expressed preference that Schechner remain full time. Schechner was then sixty-four years old. Schechner also acknowledged that Rosenheim and Longinotti signed him to a two-year contract in 2004, when he was sixty-two years old, and one-year contracts in 2007 and 2008, when he was sixty-five and sixty-six respectively. Rosenheim and Longinotti signed Schechner to his last contract just months before he was laid off. Similarly, Rosenheim and Longinotti signed Lobertini to new contracts when he was over forty years old. This includes his final two-year contract, which Longinotti signed when Lobertini was forty-six, less than two years before Lobertini was laid off. Plaintiffs acknowledge that the station was under no obligation to offer them new contracts when it did so.

KPIX is entitled to a favorable same-actor inference because Longinotti and Rosenheim signed Schechner and Lobertini to new contracts not long before they laid off Schechner and Lobertini. In light of the same-actor inference, we conclude that Schechner and Lobertini failed to present sufficient evidence of pretext to survive summary judgment at step three of the McDonnell Douglas framework.

IV. CONCLUSION

The McDonnell Douglas three-step burden-shifting framework does not collapse into a single step when plaintiffs rely on statistical evidence to make a prima facie case of disparate treatment age discrimination. We hold that a plaintiff's statistical evidence need not account for the employer's non-discriminatory reason for the discharge in order to show a stark pattern of discrimination. Although Schechner and Lobertini established a prima facie case of discrimination, they did not present sufficient evidence of pretext to survive summary judgment.

AFFIRMED.