dissenting.
I find this ease particularly disturbing. The Equal Employment Opportunity Commission (“EEOC”) has a duty to intercede on behalf of vulnerable victims of unwarranted discrimination. But here the EEOC seems to have gone way over the line, especially in the age discrimination claim and the monetary remedy. For the reasons that follow I must respectfully dissent.
I. Disparate Treatment Claim
Ted Gryezkiewicz migrated to America from Communist Poland. In 1966 or 1967 he began his pursuit of the American dream by founding 0 & G Spring and Wire Forms Specialty Company (“0 & G”), a company which manufactures wire spring products. His entrepreneurial efforts have been reasonably successful. 0 & G now employs roughly fifty people at its Chicago plant, and the company is modestly profitable. For the last ten years, 0 & G has been defending itself from complaints by the EEOC. His defense, in attorney’s fees alone, has already cost him more than $400,000. Today, this court holds that O & G must pay some 451 persons “who might have applied to O & G” more than $378,000 in back pay. The court found discrimination because his thirty-five secondary department employees included Hispanic and Polish workers, but no black workers. The court reached this result notwithstanding the EEOC’s failure to produce even one black witness who applied to fill a vacancy at O & G. There was only a faulty statistical model presented by the EEOC for the purposes of litigation.
A. Sufficiency of the Evidence
This court concludes that there was sufficient evidence for the district court to find against O & G on the EEOC’s disparate treatment claim. “To succeed in a claim alleging disparate treatment, the EEOC ultimately had the burden of proving by a preponderance of the evidence that [O & G] engaged in a ‘pattern or practice’ of discrimination. ...” E.E.O.C. v. Sears, Roebuck & Co., 839 F.2d 302, 308 (7th Cir.1988). “This includes proof of the employer’s discriminatory intent....” Id. The EEOC must also prove a “pattern or practice” — in other words ‘“that racial discrimination was the company’s standard operating procedure— the regular rather than the unusual practice.’ ” Mozee v. American Commercial Marine Serv. Co., 940 F.2d 1036, 1050-51 (7th Cir.1991) (quoting International Bhd. of Teamsters v. United States, 431 U.S. 324, 336, 97 S.Ct. 1843, 1855, 52 L.Ed.2d 396 (1977)). The district court found intentional discrimination on the basis of statistical evidence and so-called “anecdotal” evidence. This court concludes that this evidence was sufficient to support the district court’s finding of intentional discrimination. I disagree.
First, a look at the so-called anecdotal evidence: This evidence consisted of testimony by four black witnesses who applied to O & G and were not hired. A court may rely on evidence of individual acts of intentional discrimination to prove disparate treatment. See Sears, 839 F.2d at 308. But in this case the testimony of the four black witnesses was insufficient to demonstrate that any individual discrimination had occurred because the EEOC did not prove that O & G had openings in its secondary department when the four applied. E.E.O.C. v. Chicago Miniature Lamp Works, 947 F.2d 292, 304 (7th Cir.1991) (evidence that the black applicants were not hired in itself was unpersuasive to show intentional discrimination). See also Pierce v. F.R. Tripler & Co., 955 F.2d 820, 824 (2d Cir.1992) (in order to be liable for discrimination, a position must be available). “[I]f there is no vacancy to be filled, ... that would normally be the end of a Title VII claim.” Rush v. McDonald’s Corp., 966 F.2d 1104, 1118 (7th Cir.1992). The district court did not find that any of the black witnesses had applied for vacant positions at O & G. O & G asserts that no vacancies existed and the *886EEOC does not dispute this fact in its brief. Instead, the EEOC asserts that the percentage of blacks applying when 0 & G was hiring would be the same as the percentage of whites. This fact, however, has no bearing on whether vacancies existed when the people applied. Without a finding that a “vacancy” existed, the testimony of these witnesses cannot be considered anecdotal evidence of individual acts of discrimination by 0 & G. See, e.g., Sears, 839 F.2d at 310, n. 8; Chicago Miniature, 947 F.2d 292; Pierce, 955 F.2d 820.1
The only remaining evidence supporting the EEOC’s position consists of a statistical model presented by Dr. deVise. While a district court has broad discretion in determining the probativeness of statistics, Sears, 839 F.2d at 310, if the statistical model fails to account for an important factor, reliance on that model constitutes clear error. Chicago Miniature, 947 F.2d at 301. Such is the case here:, the statistical model on which the district court relied was fatally flawed because the relevant labor market used by the EEOC failed to identify qualified and interested applicants.
The statistical model on which the district court relied was flawed first of all because 0 & G had a preference for hiring skilled or experienced workers since even in the secondary department employees were required to read blueprints to operate the manual punch and stamp presses. Also 0 & G desired to promote to more skilled jobs from within. 705 F.Supp. at 406. In making its hiring decision, Ted Gryezkiewicz testified, the preference for experienced workers played a role in approximately 90% of O & G’s hiring decisions. The district court found that O & G in fact had a preference for skilled workers and that this preference was reasonable. 705 F.Supp. at 402, 406. Notwithstanding this finding, however, the district court credited as evidence Dr. deVise’s statistical model which failed to account for this skill and experience factor. This variable was too important to be ignored because “where special qualifications are necessary, the relevant statistical pool for purposes of demonstrating discriminatory exclusion must be the number of minorities qualified to undertake the particular task.” City of Richmond v. J.A. Croson Co., 488 U.S. 469, 501-02, 109 S.Ct. 706, 725-26, 102 L.Ed.2d 854 (1989). Reliance on a statistical model which fails to account for such an important variable is clearly erroneous. See, e.g., Chicago Miniature, 947 F.2d at 301-03.
The EEOC argues that since O & G did not require all of its new hires to have experience or training, the failure to account for this preference is irrelevant. But as we noted in Holder v. Old Ben Coal Co., 618 F.2d 1198, 1201-02 (7th Cir.1980), the Supreme Court has said a “Title VII plaintiff must show that rejection for a job did not result from ‘the two most common legitimate reasons on which an employer might rely to reject a job applicant: an absolute or relative lack of qualification or the absence of a vacancy in the job sought.’ ” (quoting Teamsters, 431 U.S. at 358 n. 44, 97 S.Ct. at 1866 n. 44) (emphasis added). Thus Holder logically concluded: “[a] desire to hire the more experienced or better qualified applicant is a non-discriminatory, legitimate, and common reason on which to base a hiring decision.” Holder, 618 F.2d at 1202. Accord Heerdink v. Amoco Oil Co., 919 F.2d 1256, 1260 (7th Cir.1990). Or as now Chief Judge Posner put it in Mason v. Continental III. Nat. Bank, 704 F.2d 361, 364 (7th Cir.1983): “No rational enterprise that has several qualified candidates for a position selects among them by lot; it picks the best qualified.”
The statistical model was also fatally flawed because it failed to identify the interested portion of the labor market. “The identification of a relevant labor market — the key issue in a class-based Title VII case— means not only identifying qualified potential applicants for the job at issue but also identifying interested potential applicants.” Chicago Miniature, 947 F.2d at 302. The statistical model that the district court relied on in this case failed to identify the interested applicants for jobs at O & G because it did not *887consider that 0 & G does not require its workers to speak or read English. The lack of an English fluency requirement increases the interest of non-English speaking Polish and Hispanic job-seekers to work at 0 & G. 705 F.Supp. at 406. The statistical model also failed to account for the fact that Polish and Spanish were routinely spoken at 0 & G. This would also reduce the interest of the English speaking black (and white) labor market to work at 0 & G., Cf. 705 F.Supp. at 406.2
We held in Chicago Miniature, 947 F.2d at 301, that failing to account iii a statistical model for the fact that an employer did not require its workers to speak English rendered the statistical evidence defective. We also held that crediting such statistical' evidence as proof of disparate treatment constituted clear error. I would follow the precedent of Chicago Miniature and conclude that the district court clearly erred in crediting the EEOC’s statistical model as evidence of a pattern or practice of intentional discrimination by 0 & G. This flawed model failed to account for, not just the lack of an English-flueney requirement, but also the prominent use of the Spanish and Polish languages in the workplace at 0 & G.
Even though the statistical model failed to account for the lack of an English fluency requirement, the prominent use of Spanish and Polish at O & G, and O & G’s preference for skilled or experienced labor, this court reasons that the district court factored these considerations in its analysis of the statistical model because not requiring fluency in English was offset by low pay and poor working conditions (which is incidentally another logical reason why English-speaking job-seekers would not apply). One would think that only immigrants who spoke little or no English would take these jobs if working conditions and pay were bad. Since they couldn’t work elsewhere because of a language requirement, this was their only alternative. English-speaking workers with experience would go to a better workplace with higher pay. See 705 F.Supp. at 406. In fact, Dr. deVise testified to this effect.
Nevertheless, the court concludes that the lack of an English fluency requirement “could not account for the absence of any African-American hires_” Opinion at 7. The district court also ignored the major defects in the statistical model3 because O & G did not have any black employees among its thirty-five secondary workers. 705 F.Supp. at 409. This court, however, states that “O & G’s failure to hire African-Americans would be justified in statistical terms if African-American availability were 6% of the *888labor market....” Thus, by the court’s own formula, the absence of any black employees among the thirty-five who worked in the secondary department is statistically explainable. The court, however, does not think this is statistically possible — in other words, the court does not believe that black worker availability could possibly drop to 6% even if the statistical model properly defined the relevant labor market as the relevant qualified and interested labor market.
This conclusion misapprehends the significant impact the lack of an English fluency requirement, alone, has on the composition of a work force. In Chicago Miniature, 947 F.2d at 303, this court demonstrated just how significant this one factor is on the racial makeup of a work force by presenting the following hypothetical:
Assume that five employers are equidistant to 300 job-seekers: 100 blacks, 100 whites, and 100 Hispanics. Assuming that all employers are equally attractive and have identical requirements and conditions, then one would expect each employer to receive approximately 20 applications from each ethnic group. Each employer would have a racially balanced applicant flow.
However, if employer # 1 had no English fluency requirement and the other four employers did have an English fluency requirement, a different pattern would emerge. Assuming that all of the blacks and whites are fluent, as well as 50 of the Hispanics, and that the other 50 Hispanics are not fluent, then employer # 1 would receive 60% of its applications from Hispanics (all of the non-fluent Hispanics applications and its share of the fluent Hispanics), 20% from blacks, and 20% from whites. Thus, if the non-fluency characteristic of employer # 1 was ignored, it would appear to be discriminating against blacks and whites. Conversely, the other four employers would appear to be discriminating against Hispanics, because the majority of Hispanics, for nondiseriminato-ry reasons, would apply to employer # 1.
This hypothetical clearly shows that the absence of an English fluency requirement could greatly impact the percentage of interested workers. In fact, under the circumstances of this case the racial make-up of the interested labor market would be even more skewed because a percentage of the white job-seekers in this Polish immigrant neighborhood would be assumed not fluent in English. This still does not even take into account that many English-speaking job-seekers may not want to work in an environment of predominantly foreign languages (especially given the low pay and. apparently poor working conditions). 705 F.Supp. at 406. It also does not account for O & G’s preference for skilled or experienced workers.4
The district court reasoned that even though the statistical evidence was flawed in certain respects, it nevertheless supported a finding of intentional discrimination. A statistical model, however, is only as impressive as the underlying hypothesis. Chicago Miniature, 947 F.2d at 301. Where a hypothesis is faulty, it does not support its conclusion. Id.; cf. Free v. Peters, 12 F.3d 700, 705 (7th Cir.1993) (“So deficient is Professor Zeisel’s study that ... the study would not support [its] conclusion_”). As explained above, the hypothesis underlying the statistical model used in this ease was fatally flawed because it misidentified the relevant labor market by failing to identify the qualified and interested portion of the labor market. Therefore, this statistical model does not support its conclusion that O & G intentionally discriminated against black applicants.
The district court recognized this flaw, noting that “the relevant labor market for O & G was never clearly established by plaintiff.” 732 F.Supp. at 73. The district court and this court, however, attempt to remedy the flawed hypothesis by modifying the results of the hypothesis. We are not statisticians, however, and cannot know the actual impact a change in the underlying hypothesis can have on the conclusion that hypothesis sup*889ports. If the EEOC is going to rely solely on statistical data to prove a case of intentional discrimination, it is imperative that the model be valid; no amount of judicial intuition can rectify a defective statistical model. Therefore, contrary to the court’s position, the absence of black employees in the secondary department could have been accounted for by the statistical model had it considered the lack of an English fluency requirement, the use of the Polish and Spanish languages at 0 & G, and 0 & G’s preference for skilled or experienced workers. Chicago Miniature, 947 F.2d at 292. See also Sears, 839 F.2d at 334-35 (“interest [in the job] alone can account for the disparities computed under EEOC’s analysis”). The failure of this model to account for these variables rendered the model unreliable and it was clear error for the district court to credit it as evidence of intentional discrimination. Chicago Miniature, 947 F.2d at 302.5
What this case really comes down to is the fact that from 1979 to 1985, none of the thirty-five secondary employees of 0 & G were black. In this court’s opinion (and that of the district court), this “inexorable zero” seals 0 & G’s fate. The concept of “inexorable zero” first arose in Teamsters, 431 U.S. at 342 n. 23, 97 S.Ct. at 1858 n. 23. Since then this court has recognized that an “inexorable zero” is evidence of intentional discrimination. Loyd v. Phillips Bros., Inc., 25 F.3d 518, 524 (7th Cir.1994). In fact, dicta in Loyd indicated that a 100% segregated work force “is sometimes alone sufficient to support judgment for the plaintiff.” Id.
But every zero is not “inexorable.” Craik v. Minnesota State Univ. Bd., 731 F.2d 465, 494 (8th Cir.1984) (Swygert, J., dissenting). See, e.g., Johnson v. Transportation Agency, Santa Clara Cty., Cal., 480 U.S. 616, 656, 107 5.Ct. 1442, 1464, 94 L.Ed.2d 615 (1987) (O’Connor, J., concurring) (no discrimination in fact existed even though there were no women in skilled craft positions). To say otherwise would go against Congress’ caution to “employers not to engage in hiring quo-tas_” Chicago Miniature, 947 F.2d at 296 (quoting 42 U.S.C. § 2000e-2).
In this case, a small work force is involved — only thirty-five positions. In contrast, in Teamsters the work force included over 1800 positions. Teamsters, 431 U.S. at 337, 97 S.Ct. at 1855. Where a zero may be inexorable when such a large work force is involved, it is not when the work force is so small.6 Moreover, where, as here, the 100% work force is statistically explainable, the absence of a certain race or gender is alone insufficient to support a finding of intentional discrimination.7 In other words, the “zero” in this case is not inexorable and, therefore, it cannot constitute sufficient evidence that O & G engaged in intentional discrimination.
B. Remedy
Even if I were to agree with the court that sufficient evidence existed to support a finding of intentional discrimination, I would nonetheless conclude that the district court erred in ordering O & G to pay more than *890$378,000 in back pay to some 451 claimants. 790 F.Supp. at 780. The district court calculated the amount of back pay by first determining O & G’s hiring shortfall of black workers. However, in determining O & G’s hiring shortfall, the district court relied on the 22.5% labor market availability presented by Dr. deVise. But as both the district court and this court recognize, this percentage failed to account for the prominent use of the Polish and Spanish languages at O & G, the lack of an English fluency requirement, and O & G’s preference for skilled or experienced workers. Even assuming that the court is correct that these factors would not reduce the availability of black applicants to 6%— the percentage required to support a finding of no discrimination — they clearly reduce the percentage of blacks in the labor pool by something. How much they reduce the availability of interested black applicants, however, is unclear. As the earlier hypothesis demonstrated, the lack of an English fluency requirement by itself substantially skewed the racial composition of the interested labor market. A significant reduction in the interested labor market would have a correspondingly significant reduction in the amount O & G must pay in back pay. This is especially true when a small company is involved.
As troubling as the actual award and the amount of back pay, is the EEOC’s publication of the following advertisement in the Chicago Tribune, Chicago Sun Times, Chicago TV Guide and Chicago Defender:
If you are black and you were looking for a kick and/or punch press operator job between 1979 and 1985 at O & G Spring & Wire Forms Specialty Company, located at Division and Kostner in the city of Chicago, or if you were looking for that kind of work in the surrounding area, you may be eligible for a monetary payment as a result of a lawsuit filed by the federal government, EEOC v. O & G Spring & Wire Forms Specialty Company, No. 85 C 9966, pending in federal court.
The ad then encouraged the reader to send for an EEOC claim form.
Not surprisingly, 451 claims were submitted. This result can be compared to incidents where city buses are involved in traffic accidents. Jump-ons or ghost riders get on the stalled bus, then claim injury.8 Apparently in this case some claimants were in prison during the period when they supposedly would have applied. Others held better and higher-paying jobs. O & G elected not to engage in the expensive process of challenging claims of individual class members because its already substantial financial burden had a fixed amount. Even so it had to pay $8000.00 to run the ads. Since the EEOC did not bother to clean out the seemingly fraudulent claims, any windfall went to the payees. See Hunter v. Allis-Chalmers Corp., Engine Div., 797 F.2d 1417, 1429-30 (7th Cir.1986). The only likely remaining deterrent to further fraud on the court would be the likelihood that successful claimants would not tell others because additional claims would reduce their cut.
Even where liability is without question, this process flies in the face of common sense. Tedious as it may be, the EEOC should carefully screen claimants before allocating compensation for the alleged wrong.
II. Attorney’s Fees
The EEOC’s suit against O & G was not limited to its racial disparate treatment claim; it also filed suit against O & .G for intentional discrimination based on age. The district court found no violation of the Age *891Discrimination in Employment Act, 29 U.S.C. § 626 (“ADEA”), but nonetheless concluded that 0 & G was not entitled to recover attorney’s fees from the EEOC.
This court holds that while the Equal Access to Justice Act (“EAJA”) applies to actions brought under the ADEA, 0 & G is not entitled to fees under this Act. I agree that the EAJA applies to age discrimination suits. However, I conclude that 0 & G was entitled to recover attorney’s fees from the EEOC under the EAJA.
“Under the EAJA the government bears the burden of proving that its position was substantially justified.” Marcus v. Shalala, 17 F.3d 1033, 1036 (7th Cir.1994). The position of the United States is substantially justified if it is “justified in substance or in the main — that is, justified to a degree that could satisfy a reasonable person.” Pierce v. Underwood, 487 U.S. 552, 565, 108 S.Ct. 2541, 2550, 101 L.Ed.2d 490 (1988). “EAJA fees may be awarded if either the government’s prelitigation conduct or its ligation position are not substantially justified.” Marcus, 17 F.3d at 1036. “Thus, fees may be awarded in cases where the government’s prelitigation conduct was not substantially justified even though its litigating position may have been substantially justified and vice versa.” Id.
In this case, the district court concluded that the EEOC’s position alleging age discrimination was substantially justified because it had retained an expert to testify on its behalf. Just because a litigant can obtain an expert to testify on its behalf, however, does not mean that the litigant’s position is reasonable. Litigants often attempt to submit irrelevant and unreliable “expert” testimony to support their case. Cf., Daubert v. Merrell Dow Pharm., Inc., — U.S. —, —, 113 S.Ct. 2786, 2796, 125 L.Ed.2d 469 (1993).
In this case, the EEOC had retained Dr. deVise to present a statistical model in support of its position that O & G discriminated against persons over forty-five. This statistical model was not scientifically valid, however. Or as the district court stated, it was like comparing “apples to oranges.” 705 F.Supp. at 407. A look at the model offered by Dr. deVise demonstrates the obvious flaw. Dr. deVise created a statistical model which compared the percentage of persons over forty-five hired at O & G to the percentage of persons over forty-five already working in other facilities in similar positions in Chicago and Cook County. 705 F.Supp. at 400. This statistical model was more than unscientific, it was completely misleading. In fact, the district court found Dr. deVise’s reliance on this model so incredible that he called into question Dr. deVise’s credibility concerning his other testimony. It was unreasonable for the EEOC to attempt to rely on this obviously inappropriate statistic.
The EEOC’s reliance on this faulty statistic borders on bad faith. This is especially true given the fact that the EEOC had access to a valid comparison, but apparently ignored this statistic because it disproved its case. The proper scientific comparison would have been the percentage of people over age forty-five then working at O & G, •with the percentage of workers at other similar facilities over age forty-five. 705 F.Supp. at 407. This proper comparison demonstrates that in 1986, over 50% of O & G’s work force was over age 45, while only 37% of comparable employees at other facilities were over forty-five. 705 F.Supp. at 407. This comparison not only does not prove the EEOC’s ease, it disproves it. It is troubling to think that an agency of our federal government would retain an expert to generate an invalid statistical comparison to prove discrimination where none exists. “Mark Twain said the same thing perhaps more memorably: ‘There are three kinds of lies — lies, damned lies and statistics.’” Griffin v. Board of Regents of Regency Univ., 795 F.2d 1281, 1289 (7th Cir.1986).
The EEOC’s misuse of its power is further evidenced by its modification of its age discrimination claim. The EEOC’s original complaint alleged that O & G discriminated against five people. At trial the EEOC presented only three of the original five people9 *892and the court did not find that any of them had applied for a vacant position at 0 & G. In. fact, in the case of two of the three witnesses, the next persons hired by 0 & G were ages forty-five and fifty-five. Instead of dropping the suit after determining that these five individuals were not the victims of discrimination, the EEOC somewhere along the line changed the emphasis of its suit from individual acts of discrimination to a pattern or practice of discrimination. The EEOC then placed reliance on an invalid and unscientific statistical model. Worse yet, the EEOC’s complaint originally asserted that 0 & G discriminated against persons forty and over. 30.4% of 0 & G’s new hires were, however, forty years old or older. 705 F.Supp. at 403. This high percentage of new hires apparently did not support a claim of discrimination because somewhere, along, the line, the EEOC modified its claim to involve persons forty-five or older, thus reducing 0 & G’s percentage of new hires to 17%. Again, instead of dropping its suit against 0 & G, the EEOC forged forward. (Moreover, statistics aside, if the EEOC, wants to rely on the so-called “inexorable zero” as proof -of discrimination, it cannot ignore the prominent panorama of older workers — more than twenty-six workers of the approximately fifty-three workers were over forty-five. 705 F.Supp. at 403, 406.)
The EEOC is “an arm of the federal government with authority to subject its citizens to the burdens of litigation. With this authority comes a responsibility....” United States v. Hodgekins, 28 F.3d 610, 614 (7th Cir.1994). The facts of this case demonstrate that the EEOC not only neglected its responsibility, it abused its power. Its behavior goes beyond the realm of reasonableness. “The Equal Access to Justice Act was intended, one might have thought, for just such a case as this, where a groundless ... suit is brought by the mighty federal government against a tiny firm..-..” E.E.O.C. v. Consolidated Serv. Sys., 989 F.2d.233, 238 (7th Cir.1993).
The EEOC argues that its position was reasonable based on two letters. The first letter, one from O & G’s attorney, stated: “Many of the machines do require a younger person because of the extensive effort required to operate the machines and move the necessary part.”' The second letter, written by O & G’s controller, stated: “Because of these demanding aspects of Secondary Work, we prefer to hire younger males between 20-45 years old.” The district court in concluding that the EEOC’s position was reasonable did not rely in any way on these letters, likely because the letters were unauthorized and incorrect statements of O & G’s policy. The EEOC nonetheless claims that these letters provided a reasonable basis for bringing an age discrimination suit. These letters may have originally provided a reasonable basis for investigating a claim of age discrimination. Once the EEOC began an investigation, however, it would be immediately apparent from the make-up of O & G’s work force that no such discrimination occurred. At that point, the EEOC would no longer be justified in continuing its claim against O & G. Since EAJA fees are appropriate where either “the government’s prelitigation conduct or its ligation position are not substantially justified,” Marcus, 17 F.3d at 1036, any initial justification provided by the letters does not protect the EEOC from an award of fees for its unjustified litigation position. Id.
III. Conclusion
This entire case is founded on a very flawed statistical analysis. Thus the evidence was insufficient to prove racial discrimination. The same statistics resulted in an award that was too high coupled with a distribution remedy that invited fraud. And the EEOC’s pursuit of its non-existent ADEA claim justifies attorney fees to the defendant. As I stated at the outset, this case troubles me. No doubt the EEOC has a broad range of potential victims to look after. Among those vulnerable are iion-En-glish-speaking Polish and Hispanic immigrants who are exposed to discrimination because of their race and/or their national origin. By not taking the language factor into consideration the EEOC has in effect put a quota on one vulnerable group at the expense of another.
*893In the recent case of Consolidated Service, 989 F.2d at 237, the EEOC brought suit against a company owned by a Korean immigrant, charging the company with discrimination in favor of persons of Korean origin. The district court found the immigrant company liable, but this court reversed. In authoring that opinion, now Chief Judge Posner wrote:
In a nation of immigrants, this must be reckoned an ominous case despite its outcome. The United States has many recent immigrants, and today as historically they tend to cluster in their own communities, united by ties of language, culture, and background. Often they form small businesses composed largely, of relatives, Mends, and other members of their community, and they obtain new employees by word of mouth. These small businesses— grocery stores, furniture stores, clothing stores, cleaning services, restaurants, gas stations — have been for many immigrant groups, and continue to be, the first rung on the ladder of American success. Derided as clannish, resented for their ambition and hard work, hated or despised for their otherness, recent immigrants are frequent targets of discrimination, some of it violent. It would be a bitter irony if the federal agency dedicated to enforcing the antidiscrimination laws succeeded in using those laws to kick these people off the ladder by compelling them to institute costly systems of hiring.
Id. at 237-38.
Judge Posner’s prophecy has come to pass in this ease. I respectfully dissent.
. The court recognizes that the plaintiff failed to show that any vacancies existed at the time that the seven people applied, opinion at 875, n. 3, but nonetheless credits this evidence as evidence of individual acts of intentional discrimination.
. Contrary to the court's portrayal of bias, opinion at 877, 885, it is not the presence of Polish and Spanish workers at 0 & G that would dissuade anyone, black or white, from applying. Rather, it is the language barrier and the lack of “comfort in the language setting" that would lessen the interest of English-speaking workers — • of any race — to work where Polish and Spanish are routinely spoken by co-workers. 705 F.Supp. at ‘403, 406. Just as Polish- and Spanish-speaking workers would be attracted to a workplace where they could communicate with other workers, 705 F.Supp. at 406, people who do not speak those languages well might not apply because of their inability to communicate with a substantial portion of the existing work force.
. In fact, the district court recognized that these exact factors would cause a disparity in the racial composition of O & G's workforce, stating:
Given the background of O & G, i.e., its founder being a Polish immigrant speaking fluent Polish, and the fact that O & G had its start-up with Polish employees taken from American Spring, it is obvious that Polish immigrants would be drawn to work at O & G in numbers in excess of Polish immigrants in the work force at large. It is certainly reasonable to expect that employees who do not speak English would be more comfortable working for a company that has a large number of Polish-speaking employees. It is also to be expected that immigrant populations would have a network to communicate the availability of jobs in Polish-speaking surroundings. The same would be true for Hispanic immigrants who also would be expected to be drawn to an employer that has a substantial number of Hispanic-speaking employees.
It is also reasonable to expect that recent immigrants would be willing to work at lower pay and under poorer conditions as an offset for comfort in the language setting.
It is also reasonable for an employer such as O & G to have a preference for employees with experience or a technical background because of its lack of a training program and its desire to promote to skilled jobs from within rather than to incur the expense of advertising for skilled employees.
705 F.Supp. at 406.
. O & G's preference for skilled or experienced workers also explains why approximately half of its workforce is Polish. In Poland, technical training is taught in high school. This includes training the student how to read rulers and micrometers and how to operate lathes, milling machines, and presses.
.The EEOC's failure to present any evidence of individual acts of intentional discrimination confirms the weakness of the statistics. "This court has recognized that ‘examples of individual discrimination are not always required, but we think that the lack of such proof reinforces the doubt arising from the questions about validity of the statistical evidence.’ ” Sears, 839 F.2d at 311. See also Rossini v. Ogilvy & Mather, Inc., 798 F.2d 590, 604 (2d Cir.1986) ("In evaluating all of the evidence in a discrimination case, a district court may properly consider the quality of any anecdotal evidence or the absence of such evidence.”). Moreover, this court has recognized that "strong statistics may prove a case on their own, while shaky statistics may be insufficient unless accompanied by additional evidence.” Op. at 876 (citing Teamsters, 431 U.S. at 340, 97 S.Ct. at 1857); Chicago Miniature, 947 F.2d at 300-01. Thus “when the statistical evidence does not adequately account for ‘the diverse and specialized qualifications necessary for [the positions in question]' strong evidence of individual instances of discrimination become vital to the plaintiff’s case.” Sears, 839 F.2d at 311 (internal quotations omitted). This vital evidence is totally lacking in this case.
. The EEOC also points out that small sample sizes also provide misleading results in standard deviation analysis and unless the total number in the sample is at least thirty, such an analysis is of no statistical significance.
. O & G’s workforce was not completely without black employees, but at the time in question the thirty-five secondary workers did not include any.
. See, e.g., Peter Kerr, "Ghost Riders” Are Target of an Insurance Sting, N.Y. Times, Aug. 18, 1993, at A1 (video cameras inside a bus involved in a staged accident filmed seventeen people scrambling onto the bus after the accident occurred but before police arrived); Peter Kerr, Insurance Fraud Sting Catches "Ghost Riders” Who Jump on the Bandwagon in New Jersey Bus Crashes, Guardian Newspapers Limited, Aug. 19, 1993, at 18 (in one case, twenty-seven bus passengers filed claims even though the bus was not in an accident — the crash the passengers heard came from behind the bus and involved only a car and a lorry, not the bus); Conor O'Cleiy, Bus Firm Exorcises Its Ghosts, The Irish Times, Aug. 19, 1993, at l (one videotape made during a staged accident showed a man jumping on board and declaring to the passengers: "All you people who want to get paid, you stay right there, stay down. Wait for the ambulance to come. Your neck hurts, your legs hurt, all of that. You’ll get some money. Stay there. They pay.").
. In fact, one of the five testified during a deposition that he did not file a claim with the EEOC and he had no idea how his name was placed on *892the EEOC’s complaint. Another failed to appear altogether.