Vincent P. Shutt, a citizen of the state of Washington, brought suit against Sandoz Crop Protection Corporation (Sandoz), a New York corporation, and-Zoecon Corporation (Zoecon), a Delaware corporation, alleging that he had been discharged by the defendants in violation of the Age Discrimination in Employment Act (ADEA), 29 U.S.C. §§ 621-634 , and the Washington law against discrimination, Wash.Rev.Code Ann. § 49.60 et seq. After a bench trial judgment was given for Shutt. We reverse.

PACTS

Shutt was born on April 3, 1930. He went to work for the Crop Protection Division of Sandoz in January 1980 as a salesman of agricultural chemicals. His territory included Washington, Oregon, Idaho, and Montana. In 1984 Sandoz acquired Zoecon, and Shutt was then employed by the Crop Protection Division of Zoecon.

Throughout the period of employment by Sandoz and Zoecon Shutt’s performance appraisals indicated that his work was satisfactory. For the first six months of 1986 his sales of the various pesticides for which he was responsible were in excess of the budget assigned to him for his territory.

In the spring of 1986 Sandoz bought the agribusiness of Velsicol Chemical Corporation. Following the sale this business was known.as V.S. Crop Protection Corporation (V.S. Crop). In May 1986 Sandoz announced its intent to merge the Crop Protection Division of Zoecon with V.S. Crop, effective June 2, 1986. As of this date the Crop Protection Division of Zoecon became subject to the management of V.S. Crop.

V.S. Crop was a much larger operation than Zoecon. Its sales in 1985 totaled $100 million, while Zoecon’s were $21 million. V.S. Crop had 85 sales representatives, Zoecon 21. V.S. Crop’s sales were dominated by a few products such as Banvel, a dicamba-based herbicide. Zoecon’s products did not include Banvel or any other dicamba-based pesticide.

In the reorganization under the supervision of V.S. Crop, 81 sales territories were created. Because the combined sales forces of Zoecon and V.S. Crop totaled 106 salesmen, there were 25 more sales representatives than there were sales territories. It became a business necessity for a number of salesmen to be terminated. All of the persons making decisions for the merged company were former managers of V.S. Crop.

Of the Zoecon salesmen six were terminated, one was given a chance to transfer, and 14 were retained. Of the six who were terminated, five were 40 or older. Shutt was among those terminated.

The decision to terminate was based on subjective evaluation of the work of the salesmen. As the new managers were not familiar with Zoecon they relied in part on a ranking of salesmen for Zoecon made by its former national sales director. This ranking was not followed slavishly. For example, Mike Svoboda, age 63, who was ranked eighth in the Western Region, was offered a chance at a transfer, while Shutt, who was ranked seventh in the Western Region, was fired. In the Eastern Region the two men at the top of the list, who were over 40, were retained, but of the five men at the bottom of the list two were also retained who were under 40 and three were discharged who were respectively 66, 38 and 54. The trial court concluded that the list had not been relied on.

Shutt’s discharge was effective July 15, 1986. His territory was then covered by four former V.S. Crop salesmen, of whom three were under 40.

PROCEEDINGS

Shutt brought suit, charging age discrimination. The district court concluded as a matter of law that

The district court gave judgment for the plaintiff. The defendants appealed.

ANALYSIS

This case was decided as a “disparate impact” case where the discriminatory effect of an employment practice, neutral on its face, is shown by statistical evidence. See Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 986-87 , 108 S.Ct. 2777, 2784-85 , 101 L.Ed.2d 827 (1988). It is established that a subjective method of selection may constitute such an employment practice. Id. at 991 , 108 S.Ct. at 2787 . At the same time it has been cautioned that once the practice has been identified, the statistical disparities must be “sufficiently substantial that they raise ... an inference of causation.” Id. at 995 , 108 S.Ct. at 2789 . The statistical evidence may not be probative if the data are “small or incomplete.” Id. at 996-97 , 108 S.Ct. at 2789-90 .

The district court reached its conclusion as to discriminatory impact by looking at an exceedingly small class of employees, the 21 former salesmen of Zoecon. The district court offered no explanation why it chose this segment of the merged company as its sample. True, these persons had once worked for Zoecon as a separate entity, but the Zoecon salesmen had been merged with sales representatives from V.S. Crop and there was a single business with sales representatives that now numbered 106. To carve out the former Zoecon employees was arbitrary. No doubt the V.S. Crop managers favored their former employees: that may have been favoritism, it was not age discrimination. When a merger occurs, it is often the harsh reality that one company is dominant and the other company’s employees are more easily terminated. See Rose v. Wells Fargo & Co., 902 F.2d 1417 (9th Cir.1990) (merger of Wells Fargo and Crocker National banks displaced Crocker employees disproportionately).

There were discharges from the old sales representatives of the Velsicol/V.S. Crop group as well as from the old Zoecon group. The only way in which it could have been statistically determined whether the new management was discriminating on the basis of age was to look at the total statistics of the merged sales force. If this method were not adopted, no sufficiently substantial statistical pattern existed that could be identified as discriminatory. Shutt made no effort to show what the statistics of the entire sales force would have demonstrated.

In Sengupta v. Morrison-Knudsen Co., 804 F.2d 1072 (9th Cir.1986), we faced an analogous situation. In that case, Sengup-ta belonged to Department 222 within the Mining Group of the Morrison-Knudsen Co. Id. at 1073 . Layoffs precipitated by the shale oil industry collapse hit Sengup-ta’s department harder than the Mining Group as a whole. Id. at 1073-74 . Seng-upta claimed that the proper labor pool from which to draw statistical evidence of race discrimination was his department, consisting of 28 persons, rather than the Mining Group, consisting of 281 persons. Id. at 1074 . We rejected that argument, stating:

We went on to say, “[w]hen the relevant statistical pool, the entire Mining Group, is considered, there is no disparate impact.” Id. at 1077 . So here, there existed no necessity to employ a small group. The relevant pool was 106 sales persons. When that pool is considered, no disparate impact has been shown. Shutt’s case evaporates.

Shutt also challenged Sandoz’s termination decisions under the Washington law against age discrimination, Wash.Rev. Code Ann. § 49.60' et seq. In Washington, however, disparate impact analysis does not extend to subjective, discretionary employment practices. Shannon v. Pay n’ Save Corp., 104 Wash.2d 722 , 709 P.2d 799, 806-07 (1985); Oliver v. Pacific Northwest Bell Co., 106 Wash.2d 675 , 724 P.2d 1003, 1006 (1986). Shutt therefore failed to state a claim of disparate impact age discrimination under Washington law.

The judgment is REVERSED and the case REMANDED with instructions that judgment be entered for the defendants.