Opinion by Judge FISHER; Concurrence by Judge WALLACE
FISHER, Circuit Judge:
OVERVIEW
This case arises out of a business deal ostensibly for the international trade of frozen chicken. Appellants are JR International Corp. (“JRI”), a South Korean corporation, and its wholly owned subsidiary, Jinro America, Inc. (“JAI”), which we shall collectively refer to as “Jinro” or “the Jinro Group.”1 Appellees are defendants Brian Bishop, COBBI International Food Products (“COBBI”), Landmark Forwarding Company (“Landmark”), Burnett Watkins and Secure Investments, Inc. (“Secure”), whom we shall collectively refer to as the “defendants.”
When the parties’ deal unraveled, Jinro sued the defendants to recover millions of dollars for breach of contract, fraud and racketeering, but were met with the defendants’ claim that the transaction was a sham. The defendants argued that the chicken deal was designed to cover up a high-risk investment program that circumvented Korea’s currency regulations. After the district court bifurcated the trial to address the sham contract allegation in phase one, a jury agreed with defendants’ [996]*996characterization of the transaction, whereupon the district court sua sponte entered summary judgment against Jinro on all its claims. Although Jinro raises several issues on appeal, which we shall address, we conclude the most significant is its contention that the phase one trial was prejudi-cially infected by ethnically biased, “xenophobic” expert testimony. We agree that this objectionable testimony was completely improper, and accordingly reverse the judgment against Appellants.
Jurisdiction
The district court had diversity jurisdiction under 28 U.S.C. § 1332. Appellants are South Korean and Washington corporations. Appellees are Arizona corporations and residents. The amount in controversy exceeds $75,000. We have jurisdiction under 28 U.S.C. § 1291.
I
Background
This case involves a lengthy, complicated and disputed factual background. We begin by introducing the parties and then relate their differing versions of the events.
A. The Parties
The Jinro Group is a Korean consortium comprised of nine companies, each specializing in different markets ranging from construction to food and beverages. Jinro International, a member of the Jinro Group, is involved in international trade in at least 78 countries. In 1994 and 1995, JRI traded in commodities including sugar, fish meal, cotton, cement, zinc, yarn and ball bearings. JRI’s sales from international trade in those years totaled roughly $140 million. The agreement we are concerned with here would have marked JRI’s initial entry into the trade of frozen chicken. JRI has wholly-owned subsidiary companies around the world, including Hong Kong, Japan and the United States, which is home to Jinro America.
Brian Bishop is an Arizona-based businessman, with experience in commodities trading and the international food industry. In April 1994, he formed COBBI (which stands for Corporate Offices of Brian Bishop, Inc.) and Landmark Forwarding Company, in order to enter into the chicken trading agreement with Jinro. He claims Jinro, another company, Fool Valley, and his company Landmark, actually used the chicken trading contract as a cover up for their real agreement: a joint venture to invest in a highly risky “roll program,” the nature of which we shall explain later. According to Bishop, Jinro needed the chicken deal to provide a legitimate face for the transaction, because Jin-ro’s activity in the roll program would have been forbidden by Korean law.
Burnett Watkins was an associate of Bishop’s and the owner and President of Secure Investments, Inc., an Arizona corporation. As part of the agreement between Bishop and Jinro, Secure was supposed to supply $10 million in United States Treasury notes as collateral to be held in a blocked account for the benefit of Jinro.
B. The Joint Program Agreement— Jinro’s Story
On November 16, 1994, Jinro, COBBI and Landmark entered into a written contract for the buying and selling of frozen chicken. The “Joint Program Agreement” (“JPA”) was a relatively simple contract. It recited that COBBI and Landmark were experts in trading frozen chicken, and that they desired financing of this trade. It stated that JRI and JAI were capable of financing commodity and futures trading and willing to enter into an [997]*997agreement with COBBI and Landmark for that purpose.
Under the terms of the JPA, the parties envisioned the buying and selling of chicken in large amounts, with Jinro serving as an intermediary and making its profit from the difference in price. Basically, Landmark would buy chicken in large volumes at extremely low prices and sell it to Jinro. Jinro, in turn, would immediately sell it to COBBI, which had already secured sales orders for the chicken at a higher price— chicken arbitrage, in essence. Jinro agreed under the JPA to advance Landmark $10 million to cover part of the shipment of product in the twelfth month of the JPA, which was a year-long agreement. As collateral for the advance, Landmark agreed to assign to Jinro $10 million of Treasury securities, which were to be held in a blocked account under Jinro’s name at an investment house called Saratoga Investments, Inc.
As months went by, no chicken was bought or sold. Bishop repeatedly communicated with Jinro, offering excuses for the delay. He gave various explanations, usually involving technical difficulties in obtaining letters of credit, but he always stated that the deal was on track and that he had buyers and sellers lined up. No chicken having been bought or sold by June 1995, Jinro finally declared Bishop and his companies in default on the contract and demanded repayment of its advance.
Jinro filed suit in August 1995 to prohibit transfer of the Treasury collateral Bishop and Watkins had placed at Saratoga Investments in Jinro’s name. Ultimately, after filing a receivership action, Jinro learned there were no Treasury securities on hold for it at Saratoga. Jinro then amended its complaint to include racketeering and various fraud claims. Watkins, who had given Jinro documentation regarding the securities, including the identifying (or “CUSIP”) numbers of the Treasury notes, and his company Secure Investments were included as conspirators in the fraud. Furthermore, Jinro learned that Watkins was an undisclosed principal of Saratoga Investments.
Jinro argues it entered into a valid, written agreement for the international buying and selling of chicken, and that defendants Bishop and Watkins defrauded it out of its initial $10 million advance.
C. The Roll Program — Bishop’s Story
The defendants tell a different story. Bishop contends JRI approached him with a request to help it invest money in the United States in a high-risk, high-yield type of investment scheme known as a “roll program.” Defendants describe the “roll program” as a highly speculative investment plan involving the trade of credit and commercial paper issued by foreign state banks. According to Bishop, Jinro introduced the concept to him and asked if he would be willing to form a company that would enter into a joint venture with Jinro to pursue such an investment program. Jinro’s representatives promised substantial profits and explained that, because Jinro’s participation in such a deal was contrary to Korean laws, they needed an American company to effectuate the transaction. Bishop says he then found people who knew about such a transaction who helped him set it up. He formed his companies, COBBI and Landmark, at Jin-ro’s request so that the companies could enter into a legal trade agreement as a cover for the parties’ underlying activity. According to Bishop, Jinro instructed him never to use the words “roll program” in his correspondence with the company. After entering into the agreement, he continually updated Jinro with phony chicken orders and problems so the company [998]*998would have seemingly legitimate documentation to prove the validity of the chicken trading contract. Bishop argues Jinro wanted to avoid Korean law, which strictly regulates the investment of funds originating in Korea, and, to that end, engaged him to set up the cover transaction, the JPA.
Disappointed by the profits from the roll program, Jinro decided to sue to recover its initial $10 million investment, treating the chicken deal as though it were a legitimate agreement.
II
Discussion
The central issues on appeal, though informed by this factual background, stem from the trial itself. Jinro appeals the district court’s decision to bifurcate the trial as well as its decision to allow the defendants to introduce extrinsic evidence regarding the JPA. Jinro also contends the district court improperly arrived at its instructions on Korean law. Finally, Jinro argues the court erred by allowing the admission of ethnically biased expert testimony. We believe the district court was within its discretion in bifurcating the trial, admitting parol evidence and instructing the jury on foreign law, but that it erred by admitting the expert testimony.
Because we reverse on the issue of improper expert testimony, we need not reach Jinro’s other arguments. We do so at the outset, however, because the issues may well recur if there is another trial on remand. We then go on to address the truly salient issue in this case, the admission of biased testimony.
A. Bifurcation
We review a district court’s decision to bifurcate a civil trial for abuse of discretion. Hilao v. Estate of Marcos, 103 F.3d 767, 782 (9th Cir.1996). Rule 42(b) of the Federal Rules of Civil Procedure provides in pertinent part:
The court, in furtherance of convenience or to avoid prejudice, or when separate trials will be conducive to expedition and economy, may order a separate trial of any claim, ... or of any separate issue or of any number of claims ... or issues, always preserving inviolate the right of trial by jury....
Under Rule 42(b), the district court has broad discretion to bifurcate a trial to permit deferral of costly and possibly unnecessary proceedings pending resolution of potentially dispositive preliminary issues. See Exxon Co. v. Sofec, Inc., 54 F.3d 570, 575 (9th Cir.1995), affirmed, 517 U.S. 830, 116 S.Ct. 1813, 135 L.Ed.2d 113 (1996). Here the district court decided that, before addressing any of the parties’ claims, an initial determination should be made whether the parties had entered into a valid agreement and, if so, what that agreement entailed. As the court explained: •
A determination as to which agreement actually embodied the parties’ intent was of primary importance to all claims. Thus, I concluded that the January 1999 trial would proceed only on that issue.
The district court’s approach was a reasonable way to promote clarity and judicial economy, because the validity of the contract directly informed the resolution of the other claims. Accordingly, the court did not abuse its discretion in ordering the trial bifurcated.
B. Parol Evidence
Jinro argues the district court erred in allowing defendants to testify about an oral “true” agreement between the parties that differed from the written agreement. A district court’s application of the parol evidence rule, an issue of state [999]*999law, is reviewed under the same de novo standard applied to decisions concerning federal law. See Mastro v. Witt, 39 F.3d 238, 241 (9th Cir.1994).
In general, Arizona’s parol evidence rule prohibits the introduction of extrinsic evidence to vary or contradict, but not to interpret, an agreement. Taylor v. State Farm Mut. Auto. Ins. Co., 175 Ariz. 148, 152, 854 P.2d 1134 (1993). Defendants’ evidence, however, did not go toward varying the terms of the parties’ written contract, the JPA. Rather, the defendants argued the JPA was a sham, a cover-up contract entered into solely to allow Jinro to escape the watchful eye of Korean regulatory agencies. The JPA, simply put, was created to provide Jinro a legal face for an otherwise impermissible transaction.
The question, then, is not whether Arizona’s version of the parol evidence rule allows the admission of evidence to vary the terms of a contract. Rather, the critical issue is whether the parol evidence rule allows the admission of evidence to prove that a written agreement was actually a cover-up for fraudulent or illegal activity.
The Arizona Supreme Court has not addressed this issue specifically, but we believe Arizona law would allow the admission of evidence to prove that a contract was entered into for an illegal purpose. In Arizona Cotton Ginning Co. v. Nichols, 9 Ariz.App. 493, 454 P.2d 163 (1969), for instance, the court of appeal held that evidence that a seemingly legal transaction was actually a sham, meant only to satisfy plaintiffs internal bookkeeping procedures, was admissible despite the parol evidence rule. Id. at 496, 454 P.2d 163. Moreover, unless a different rule is announced by the courts or the legislature, Arizona generally follows the Restatement. Rodriquez v. Terry, 79 Ariz. 348, 350, 290 P.2d 248 (1955). The Restatement provides:
Agreements and negotiations prior to or contemporaneous with the adoption of a writing are admissible in evidence to establish ... illegality, fraud, duress, mistake, lack of consideration, or other invalidating cause.
Restatement (Second) of Contracts § 214(d) (1981) (emphasis added). Comment “c” of § 214 explains the phrase “invalidating cause” in the following way: “What appears to be a complete and binding integrated agreement may be a forgery, a joke, a sham, ... or it may be illegal.” Id. Given the Restatement view, and Arizona’s general adherence to the Restatement, we conclude that parol evidence could be admitted, despite a seemingly valid, integrated agreement, to show the agreement was, in fact, a sham or cover-up for otherwise illegal activity.
Jinro argues the sham exception does not allow the admission of parol evidence to show only parts of the contract were a sham. This argument is not persuasive for two reasons. First, Jinro cites scant authority for the proposition itself. Second, and more importantly, the argument misstates the situation here. Bishop and his companies do not contend that parts of the JPA are valid and enforceable, but that others are not. Rather, they contend that the JPA itself was a cover-up for an entirely separate agreement. Landmark uses the term “sham” as a synonym for cover-up, or for a contract with an illegal purpose, and not to argue that its evidence fits within the so-called “sham exception” to the parol evidence rule.
In sum, parol evidence was not introduced here to vary the terms of the JPA. Rather, it was introduced to show that the contract itself was a cover-up for an aggressive investment plan to avoid Korean currency laws. See Arizona Cotton Gin[1000]*1000ning, 9 Ariz.App. at 496, 454 P.2d 163 (“We are now talking about an attack upon an apparently integrated document which establishes that the instrument is a sham and not what it purports to be at all.”). Given the circumstances and allegations, it was appropriate for the district court to allow the jury to consider evidence that the JPA was a sham contract aimed at masking Jinro’s true investment agreement.
C. Jury Instructions on Foreign Law
Jinro contends that under Universe Sales Co., Ltd. v. Silver Castle, Ltd., 182 F.3d 1036 (9th Cir.1999), the court erred in ignoring the testimony of Jinro’s legal expert in formulating a jury instruction on Korean law, even though the expert’s testimony was unrebutted. ’ We review a district court’s formulation of civil jury instructions for an abuse of discretion. Beachy v. Boise Cascade Corp., 191 F.3d 1010, 1013 (9th Cir.1999). Universe Sales noted that “[ajlthough, pursuant to Rule 44.1, courts may ascertain foreign law through numerous means, expert testimony accompanied by extracts from foreign legal materials has been and will likely continue to be the basic mode of proving foreign law.” Id. at 1038. Universe Sales then held that the district court erred in not considering the declaration of an expert on Japanese law, even though it was unrebutted by the other party.
Jinro’s argument on this front is not persuasive for two reasons. First, Fed. R. Civ. Pro. 44.1 gives wide latitude to a district court in determining issues of foreign law:
The court in determining foreign law, may consider any relevant material or source, including testimony, whether or not submitted by a party or admissible under the Federal Rules of Evidence. The court’s determination shall be treated as a ruling on a question of law.
(Emphasis added.) Second, unlike in Universe Sales, the district court did consider Jinro’s legal expert. It instructed the jury that, if the agreement between the parties was the JPA, then Jinro had received the appropriate government approval and its participation was legal under Korean law. This is precisely what Jinro’s expert testified to. In fact, it was all he testified to. He did not explain whether Jinro’s participation in the roll program would have been legal under Korean law, nor did he shed light on the relevant Korean law on that issue.
Jinro opposes the district court’s independent determination that, under Korean law, Jinro’s participation in the roll program would have been impermissible. But the court asked both parties for help in resolving that question, and neither obliged — leaving the court on its own to determine whether Jinro could legitimately have participated in the high-risk investment scheme. Accordingly, the court conducted its own independent research and reached its determinations. With all this in mind, we believe the district court acted appropriately in fashioning the foreign law jury instructions.
Jinro also argues the court erred by not providing the parties with the sources it used in reaching its conclusions on Korean law. The Advisory Committee Notes to Rule 44.1 state: “Ordinarily the court should inform the parties of material it has found diverging substantially from the material which they have presented; and in general the court should give the parties an opportunity to analyze and counter new points upon which it proposes to rely.” By their terms, then, the Notes do not require a district court to disclose its sources. Aso, here the parties presented no materials from which the court’s [1001]*1001view could have diverged. Although it might have been better for the court to have disclosed its sources, perhaps enabling the parties to respond and allowing for a more meaningful appeal of possibly erroneous decisions on foreign law, we do not believe the district court erred by not doing so here.
D. Improper Expert Testimony
Lastly, we address the district court’s admission of expert testimony, which we review for an abuse of discretion. De Saracho v. Custom Food Machinery, Inc., 206 F.3d 874, 879 (9th Cir.2000). As a key component of their defense, defendants relied on David Herbert Pelham as an expert witness to testify about Korean law and the business practices of Korean companies — particularly their alleged propensity to engage in fraudulent activity, including the avoidance of Korean currency laws. Since 1994, Pelham had been the general manager of the Pinkerton Detective Agency’s office in Korea. Although he claimed to have familiarized himself with Korean business practices as part of his job duties and as “a hobby of his,” Pelham had no formal education or training in business or as a cultural expert. He was not a lawyer. Rather, he was a private investigator who had been providing “commercial security” for over four years for various non-Korean corporations doing business in Korea. Prior to his employment by Pinkerton, Pelham had been trained as an investigator with the United States Air Force Office of Special Investigations, and had commanded the office that investigated Korean companies doing business with the Air Force. Pelham’s other purported qualifications were that he had “served five tours of duty in Korea,” lived there for about 12 years and was married to a Korean woman.
Although Pelham had not investigated Jinro itself, defendants offered his testimony ostensibly to educate the jury about the “modus operandi” of Korean businesses. On appeal, however, defendants no longer attempt to justify Pel-ham’s testimony on this basis. Rather, they raise two other arguments: absence of bias and Jinro’s waiver. First, they deny that Pelham’s testimony was ethnically biased given that his wife is Korean. For example, they state: “Mr. Pelham, married to a Korean, is far from a racist,” and “Mr. Pelham is far removed from one who is racist against Koreans or biased against the relationships developed among the Korean people, for he is married to a Korean wife.” We consider this argument both astonishing and irrelevant. Pelham’s marital status does not neutralize the content of his testimony, which is what we are called upon to evaluate in this appeal.
Second, the defendants argue Jin-ro waived review of Pelham’s testimony by failing to object to the improper statements. Again, we disagree. Before Pel-ham testified, Jinro filed motions in limine arguing that, based on his declaration, he (1) was not qualified as an expert under Fed.R.Evid. 702 and (2) would make ethnically biased generalizations about Korean businessmen and culture in violation of Fed.R.Evid. 403 and 404 and the Constitution.2 Specifically, Jinro argued Pelham was not qualified as an expert on Korean law or business practices: “[H]is expertise evidently lies in the areas of investigation and security.” Moreover, it protested that Pelham’s expected generalized testimony that Korean businesses routinely engage in various unlawful or unethical activities was a “blatant appeal to racism and xenophobia.” It contended that even if Pelham was a qualified expert on business prac[1002]*1002tices, his testimony would be unduly prejudicial — begging the jury to infer that because some Korean businesses are corrupt, Jinro is too. In a minute order, and without explanation, the district court denied Jinro’s motions as to Pelham’s general testimony on Korean culture and business practice, although it ruled that Pelham could not testify as an expert in Korean law. Thereafter, during trial, Jinro repeatedly and unsuccessfully objected to Pelham’s testimony and, following his testimony, renewed its pretrial objections in an oral motion to strike the testimony.3 The court denied this motion as well. Given Jinro’s numerous specific objections, we believe it adequately preserved its right to appellate review of Pelham’s testimony. We turn to that testimony.
At trial, Pelham testified about how, in his experience, “Korean businessmen” behave in general. Little of this testimony was directly connected to Jinro itself, and none was based on Pelham’s personal knowledge of Jinro or its affiliates. Rather, Pelham addressed himself to generalizations about Korean business attitudes and behavior.
Thus, Pelham was asked on the subject of Korea’s currency restrictions how “Korean businessmen relate to these laws or react to them.”
A: Well, Korean businessmen don’t like laws that restrict their freedom to conduct business and make money. And they certainly don’t like these laws.
Q: Why is that?
A: Well, it — it—it restricts their ability to make investments that they may want to make and just makes it difficult to do business.
This generalization, of course, hardly defines Korean businessmen as unique and prone to illegal behavior. One can fairly surmise that those in business, regardless of nationality or ethnicity, might not “like” laws that “restrict their ability to make investments” or make it “difficult to do business.” That, however, does not mean they will violate the law; and it certainly does not justify a jury concluding that Jinro, as a Korean business, would be likely to do so.
Continuing his generalizations, Pelham was asked whether there “are attempts made in Korea to get around these [currency] regulations?” “Yes,” he replied; it’s “common knowledge.” Asked for examples of such currency evasion, he evoked images of crime — “just outright theft of [ ] putting money in a suitcase and leaving the country” — smuggling, conspiracy or, in an apparent attempt to focus implicitly on Jinro, “through some kind of a phony contract to get it approved through the foreign exchange bank as a trade agreement to deceive it.” 4
When asked where the money from these schemes went, he replied:
A: Well, the United States is a favorite safe haven.
Q: Any particular place in the United States?
[1003]*1003A: Well, if you had to pick one state over another, California would be the most likely.
Q: Why California?
A. California has the largest concentration of Korean people, ethnic Korean people, Korean Americans, especially L.A.
Eventually, over objection, Pelham was permitted to opine that the written chicken-trading agreement between Jinro and the defendants was “the sort of agreement” that could have been approved by Korean banking authorities, but that the “roll program” — the defendants’ alleged “real” but unwritten agreement — would never have been approved, thus implying that Jinro had entered into a phony written contract just like the other unethical Korean businesses that populate Pelham’s universe.
In a strange twist, however, defense counsel then asked Pelham to opine on the merits of entering into oral contracts “with Koreans.”5 The logic of his answer — that oral agreements should be avoided — seemingly undercut the defendants’ theory that the oral “roll program” was the true deal rather than the written chicken-trading agreement. But the meat of his answer — ■ and its unsubtle purpose — continued Pel-ham’s theme that Korean businessmen are not to be trusted:
Q: Would you recommend to your clients that they enter into oral agreements with Koreans?
A: I would never recommend anybody rely on oral agreements.
Q: Why is that?
A: Well, I don’t think oral agreements are a very safe way to do business anyplace, but particularly in Asia and probably more particularly in Korea.
Q: Why is that?
A: Well, because of the culture, dealing with Korean businessmen can end up with some pretty sorry results if you haven’t safeguarded yourself.
On cross-examination, counsel for Jin-ro — in an unsuccessful attempt at damage control6 — elicited from Pelham his explicit view of the ethics of Korean businesses, implicit throughout his direct testimony reviewed above.
Q: You aren’t telling this jury that you believe that all Korean businesses operate corruptly, are you?
A: All Korean businesses? I don’t think I would say all Korean business. I would say that the prevalence of corruption and fraud in the Korean business community is very great and very extensive.
When asked to explain the basis for this conclusion, Pelham stated that “newspaper articles” were one indicator. Indeed, his declaration revealed newspaper articles to be a principal source for his opinion on the alleged prevalence of corruption in the Korean business community. There he referred specifically, for instance, to a Wall Street Journal article and a public opinion poll in the Korean Herald as sources. He [1004]*1004also based his opinion on unspecified information from his office staff. By the end of his testimony, Pelham had cited no research or study, nor any empirical data, and had made only generalized, anecdotal references to his personal experience. This was consistent with his declaration, which had described the particular type of currency evasion tactic alleged here, a “phony trade contract,” not a scheme about which he had personal experience, but simply one he had “heard of.”
We conclude that Pelham should not have been allowed to testify as he did for two reasons. First, his qualifications as an expert were suspect at best, and his testimony was extremely unreliable; it should not have been admitted under Rule 702. Second, even assuming his expert testimony was admissible, it was far more prejudicial than probative and should have been excluded under Rule 403.7
1. Expert testimony: Rule 702
Pelham came before the jury cloaked with the mantle of an expert. This is significant for two reasons: First, it allowed him to testify based on hearsay information, and to couch his observations as generalized “opinions” rather than as firsthand knowledge about Jinro and its activities in particular. Second, as the opinion of a purported “expert” on Korean business practices and culture, his statements were likely to carry special weight with the jury. For these reasons, care must be taken to assure that a proffered witness truly qualifies as an expert, and that such testimony meets the requirements of Rule 702.
That Rule provides:
If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise.
The Supreme Court interpreted Rule 702 and articulated general guidelines for its application in Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), and Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999). Daubert, the seminal opinion, focused on scientific testimony; Kumho made clear that DaubeH’s principles apply to “technical and other specialized knowledge” as well. Kumho, 526 U.S. at 141, 147-49, 119 S.Ct. 1167. Pelham’s testimony was admissible, if at all, as “other specialized knowledge,” and we shall review it as such.
To begin, several general principles are important to our analysis here. First, Rule 702 is applied consistent with “the ‘liberal thrust’ of the Federal Rules and their ‘general approach of relaxing the traditional barriers to “opinion” testimony.’ ” Daubert, 509 U.S. at 588, 113 S.Ct. 2786 (quoting Beech Aircraft Corp. v. Rainey, 488 U.S. 153, 169, 109 S.Ct. 439, 102 L.Ed.2d 445 (1988)). An expert witness— unlike other witnesses — “is permitted wide latitude to offer opinions, including those that are not based on firsthand knowledge or observation,” so long as the “expert’s opinion [has] a reliable basis in the knowledge and experience of his discipline.” Id. at 592, 113 S.Ct. 2786; Kumho, 526 U.S. at 148, 119 S.Ct. 1167.
[1005]*1005Second, although the district judge has “the discretionary authority, reviewable for its abuse, to determine reliability in light of the particular facts and circumstances of the particular case,” Kumho, 526 U.S. at 158, 119 S.Ct. 1167, the proper exercise of that gatekeeping function is critically important “to ensure the reliability and relevancy of expert testimony.” Id. at 152, 119 S.Ct. 1167; see also Daubert, 509 U.S. at 590, 594-95, 113 S.Ct. 2786. “[I]t is not discretion to perform the [ga-tekeeping] function inadequately. Rather, it is discretion to choose among reasonable means of excluding expertise that is fausse and science that is junky.” Kumho, 526 U.S. at 159, 119 S.Ct. 1167 (Scalia, J„ concurring).
In Daubert, the Court addressed the reliability requirement in the context of “scientific knowledge”; those principles are equally relevant in addressing the kind of “other specialized knowledge” at issue here. As the Court observed:
[T]he word “knowledge” connotes more than subjective belief or unsupported speculation.... [I]n order to qualify as “scientific knowledge,” an inference or assertion must be derived by the scientific method. Proposed testimony must be supported by appropriate validation — i.e., “good grounds,” based on what is known. In short, the requirement that an expert’s testimony pertain to “scientific knowledge” establishes a standard of evidentiary reliability.
Daubert at 590, 113 S.Ct. 2786; see also id. at n. 9 (equating “evidentiary reliability” with “trustworthiness”). Subsequently, the Court in Kumho expressly, extended Daubert’s standard of “evidentiary reliability” to all experts, not just scientific ones. Kumho, 526 U.S. at 147-48, 119 S.Ct. 1167. That standard:
requires a valid ... connection to the pertinent inquiry as a precondition to admissibility. And where such testimony’s factual basis, data, principles, methods, or their application are called sufficiently into question, ... the trial judge must determine whether the testimony has a reliable basis in the knowledge and experience of [the relevant] discipline.
Id. at 149, 119 S.Ct. 1167 (first and last alteration in original) (internal quotations and citations omitted).
Finally, Daubert invoked an important constraint on expert testimony that is particularly relevant here:
Rule 403 permits the exclusion of relevant evidence “if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury....” Judge Weinstein has explained: “Expert evidence can be both powerful and quite misleading because of the difficulty in evaluating it. Because of this risk, the judge in weighing possible prejudice against probative force under Rule 403 ... exercises more control over experts than lay witnesses.”
Daubert, 509 U.S. at 595, 113 S.Ct. 2786, quoting 138 F.R.D. 631, 632 (1991).
Applying these principles, we conclude that the district court abused its discretion in admitting Pelham’s testimony for two reasons: it was not reliable and, in any event, it was unduly prejudicial and should have been excluded under Rule 403. We address the reliability ground first.
2. Reliability
Pelham purported to be an expert on Korean business culture and practices, particularly with regard to Korean currency laws and the propensity of Korean businesses to evade them through various illegal schemes. In actuality, his qualifications to render such opinion testimony were glaringly inadequate, amounting to [1006]*1006little more than the limited perspective of a professional investigator whose work experience had exposed him to instances of corrupt business behavior. He did not have the legal, business or financial expertise to evaluate the substance of the Jinro transaction. He had no education or training as a cultural expert generally, or as an expert on Korean culture specifically. He was not a trained sociologist or anthropologist, academic disciplines that might qualify one to provide reliable information about the cultural traits and behavior patterns of a particular group of people of a given ethnicity or nationality.
Rather, Pelham offered his impressionistic generalizations about Korean businesses based on his personal investigative experiences, his “hobby” of studying Korean business practices, unspecified input from his office staff and his marriage to a Korean woman — hardly an adequate foundation for the type of expert opinion he offered the jury. Moreover, he provided no empirical evidence or studies to support his sweeping indictment of the Korean business community — other than to cite newspaper articles and a few anecdotal examples, some of them clearly hearsay. We recognize that persons experienced in a particular field may have a “practical” expertise or specialized knowledge that might qualify them to provide relevant and reliable information to a lay jury. For example, Pelham perhaps might have been qualified to testify based on his experience as a professional commercial investigator (his “discipline”) about the structure of the Korean governmental and banking systems — to illuminate for the jury how that country’s regulatory system worked. This might have helped jurors understand and contextualize Jinro’s interaction (or lack thereof) with that system.8
That was not the nature or purpose of his testimony, however. Instead, relying on his training and work as an investigator whose experience necessarily focused on persons or businesses suspected of corrupt behavior, Pelham culturally stereotyped Korean businesses — leading the jury to believe that the subset of Korean companies and individuals Pelham dealt with or “had heard of’ typified most such companies and individuals. Pelham’s sweeping generalizations, derived from his limited experience and knowledge — plainly a skewed sample — were unreliable and should not have been dignified as expert opinion.
3. Rule 403
Even if Pelham’s testimony might have been admissible as expert testimony, it was so tinged with ethnic bias and stereotyping that it should have been excluded under Rule 403’s balancing test. Daubert, 509 U.S. at 595, 113 S.Ct. 2786. “Otherwise admissible expert testimony may be excluded under Fed.R.Evid. 403 if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or undue delay.” United States v. Hoac, 990 F.2d 1099, 1103 (9th Cir.1993).
A number of cases have dealt with the problem of testimony that either directly or indirectly seeks to link a defendant’s conduct to that which is said to be typical of a particular racial, ethnic group or nationality. Although most such cases involve criminal prosecutions, which im[1007]*1007plicate constitutional and other considerations that are not wholly applicable in civil litigation, their principles are nonetheless relevant here. Thus, as we recently observed in Bird v. Glacier Elec. Coop., Inc., 255 F.3d 1136, 1151 (9th Cir.2001):
[F]airness to parties and the need for a fan1 trial are important not only in criminal but also in civil proceedings, both of which require due process. Racial stereotyping cannot be condoned in civil cases.
See also United States v. Cabrera, 222 F.3d 590, 597 (9th Cir.2000) (“The fairness and integrity of criminal trials are at stake if we allow police officers to make generalizations about racial and ethnic groups in order to obtain convictions. People cannot be tried on the basis of their ethnic backgrounds or national origin.”)
Allowing an expert witness in a civil action to generalize that most Korean businesses are corrupt, are not to be trusted and will engage in complicated business transactions to evade Korean currency laws is tantamount to ethnic or cultural stereotyping, inviting the jury to assume the Korean litigant fits the stereotype. In stark terms, Pelham’s syllogism reduced to this: (a) Korean businesses generally are corrupt; (b) Jinro is a Korean business; (c) therefore, Jinro is corrupt. Our case-law, and that of other circuits, establishes that this is an impermissible syllogism.
Cabrera, for instance, involved the prosecution of two defendants, both natives of Cuba, for conspiracy to distribute crack cocaine. At trial, a government witness, Detective Brooks, testified, making repeated references to the defendants’ nationality. He explained that the police had been “working a lot of Cubans in the area.” Id. at 591. He also testified that the round, flat wafers of cocaine he had purchased from Cabrera were the “typical way that a lot of Cubans do their drugs.” Id. at 592. When asked if the semblance of the cocaine indicated it came from the Cuban community, Brooks stated, “Well, that’s the way the Cubans cook.... ” Id. Strikingly similar to Pelham’s generalizations here was the following exchange:
Q: So you’ve never seen the non-wafer shape cocaine in a non-Cuban case?
A: No I have not. To say it’s impossible, no, but it’s indicative of the Cuban community....
Id. at 593.
Analyzing Brooks’ references to “Cubans” under the conventional standards of relevance and prejudice, as dictated by Fed.R.Evid. §§ 401 and 403, we held that his testimony was largely irrelevant, but in any event was unfairly prejudicial under Rule 403. “[I]t was unnecessary to inject [defendants’] national origin into the trial.” Id. at 596. Highlighting the ethnicity of other Cuban drug dealers under investigation “merely made it seem more likely in the eyes of the jury that [defendants] were drug dealers because of their ethnicity.” Id. In sum, Detective Brooks’ “repeated references to their Cuban origin and his generalizations about the Cuban community prejudiced [the defendants] in the eyes of the jury.” Id.
Similarly, in Bains v. Cambra, 204 F.3d 964 (9th Cir.2000), we held that although testimony about Sikh religious attitudes toward marriage and divorce was proper to explain the defendant’s motive to kill, the state prosecutor committed constitutional error by arguing such evidence showed “that all Sikh persons (and thus Bains by extension) are irresistibly predisposed to violence when a family member has been dishonored ... and also are completely unable to assimilate to and to abide by the laws of the United States.” Id. at 975. The prosecutor’s “inflammatory” argument was “more a statement about the [1008]*1008stereotypical ‘nature’ of a particular group rather than an explanation of the beliefs followed (to different degrees and in different ways) by some members of that group,” thus inviting the jury impermissi-bly to infer that Bains was compelled to kill simply because he was a Sikh. Id.
Our sister circuits, too, have condemned the inappropriate injection of race or ethnicity into a trial. The Eighth Circuit, for example, in United States v. Vue, 13 F.3d 1206 (8th Cir.1994), held it was error to admit the testimony of a customs officer that suggested persons of Hmong descent were likely to be involved in opium smuggling. After noting such evidence was either irrelevant or, if arguably relevant, unfairly prejudicial, the court concluded that “the introduction of such evidence is highly improper.” Id. at 1213. Moreover, “the error here was indeed of constitutional dimension, because the injection of ethnicity into the trial clearly invited the jury to put the Vues’ racial and cultural background into the balance in determining their guilt.” Id.
The Second Circuit addressed this issue in United States v. Cruz, 981 F.2d 659 (2nd Cir.1992). That case involved a drug prosecution where the principals were all Hispanic and the defendant was repeatedly referred to as “the Dominican.” The government’s expert witness described the area of Manhattan in which the alleged drug transactions took place as has having “a very high Hispanic population,” including “Dominicans.” The court held such testimony was highly improper and prejudicial. “Injection of a defendant’s ethnicity into a trial as evidence of criminal behavior is self-evidently improper and prejudicial for reasons that need no elaboration here.” Id. at 663-64; see also United States v. Doe, 903 F.2d 16, 20-21 (D.C.Cir.1990) (holding that expert’s testimony that the retail crack and cocaine market in Washington, D.C. “has been taken over basically by Jamaicans” was unfairly prejudicial under Rule 403, because it “strongly suggested that appellants were guilty because two of them are Jamaican”); cf. United States v. Socony-Vacuum Oil Co., Inc., 310 U.S. 150, 239, 60 S.Ct. 811, 84 L.Ed. 1129 (1940) (noting that “appeals to class prejudice are highly improper and cannot be condoned and trial courts should ever be alert to prevent them”).
We do not suggest that expert testimony relating to certain aspects of a defendant’s race, ethnicity or nationality can never be admissible. Testimony about cultural traits or behavior, for instance, is not inherently prejudicial. The cases we have just reviewed acknowledge as much. But as reflected in these decisions, the risk of racial or ethnic stereotyping is substantial, appealing to bias, guilt by association and even xenophobia. That is why Pelham’s testimony — his ethnic syllogism— was fatally flawed here. It invited the jury to distrust Jinro by invoking an ethnic, national stereotype.
Defendants cite Vang v. Xiong, 944 F.2d 476 (9th Cir.1991), to support the admissibility of Pelham’s cultural testimony. Not only is Vang readily distinguishable, it illustrates the impropriety of allowing Pel-ham’s ethnic generalizations as “expert” testimony. In Vang, two Hmong women brought a civil suit under 42 U.S.C. § 1983 accusing a city refugee counselor of repeatedly raping them. To explain why the plaintiffs would have submitted to repeated rapes and remain silent about them, the district court permitted an epidemiologist and anthropologist named Hurlich to testify as an expert about Hmong culture and to explain the behavior of the plaintiffs within that cultural context.
[1009]*1009At trial, Hurlich explained that Hmong women are generally submissive, and are raised to respect and obey men. He described the role of Hmong women in marriage; their attitudes towards sex, discussion of sex, and extramarital affairs. Most significantly, Hurlich explained that upon fleeing from Laos, Hmong refugees were reliant on government officials for their needs and would not survive in the United States without government assistance. Because of this rebanee on government assistance, the Hmong have developed an awe of persons in government positions.
Id. at 481.
On appeal, we held this cultural testimony was properly admitted in the district court’s discretion. We made particular note, however, of several circumstances— not present here — that informed our decision. First, the district court bmited the scope of the testimony, both in its pretrial ruling on admissibility and in sustaining defendant’s objection when plaintiffs tried to expand upon this limited scope at trial. Id. Second, Hurlich was the only expert either side had been able to locate who could “explain to the trier of fact who these people are, where they came from, and why they have responded the way they have in these various functions and various relationships.” Id. He could help the jury understand behavior that might otherwise be confusing, and explain the cause, effect and nature of Hmong reb-anee on governmental agencies for support. Id. at 481-82. Third, although we acknowledged the testimony was prejudicial to defendant because it supported the rape allegations, we concluded it was “not ... unduly prejudicial because of its bmit-ed scope and its direct relevance to the issues in the case.” Id. at 482. Finally, and most significant here, “Hurlich’s testimony derived from his expertise as an anthropologist and his study of the Hmong.” Id. (emphasis added).
Here the district court made no assessment of Pelham’s qualifications to opine generaby as to rampant corruption amongst Korean businesses “because of the culture.” Nothing in the record shows Pelham to be a cultural expert by education, profession or experience. Nor did the court limit the scope of such testimony, either before or during trial. Moreover, the nature and purpose of the cultural testimony in Vang was quite different from Pelham’s. There the expert provided an academic, noninflammatory explanation of the Hmong culture in order to explain the seemingly inexplicable behavior of the two women plaintiffs in repeatedly submitting to rape by a government official. Pelham, on the other hand, spoke from the perspective of a professional investigator about the assumed motives of Korean businesses, repeatedly implying that Korean businesses were a corrupt lot. But this was not the informative testimony of the anthropologist in Vang; this was the kind of guilt-by-ethnic-association testimony condemned by this and other courts.
The manner in which Jinro’s status as a Korean business was exploited here begged the jury to draw an inference adverse to Jinro based entirely on its ethnic identity or national origin. See London Guar. & Accident Co. v. Woelfle, 83 F.2d 325, 340 (8th Cir.1936) (noting that “[r]e-marks tending to create an atmosphere of hostility toward foreign corporations are condemned as an appeal to sectional or local prejudice”). As the Eighth Circuit observed in Vue, “Formal equality before the law is the bedrock of our legal system.” 13 F.3d at 1213. This principle — at least with regard to prejudicial, unreliable, ethnically biased testimony — must apply to civil litigants as well.
[1010]*1010III
Conclusion
This was without doubt a complicated trial. The district court acted appropriately in all but one of its trial decisions: it erroneously allowed the admission of improper expert testimony. Pelham was unqualified to give the kind of testimony he did; it simply was unreliable. Moreover, the testimony was inflammatory and appealed to ethnic bias. Because its admission — under the aegis of expert testimony — “clearly invited the jury to put [Jinro’s] ... cultural background into the balance,” id., we believe the jury may have improperly arrived at its ultimate finding that the operative agreement between the parties was for the roll program, as defendants alleged. The district court’s sua sponte grant of summary judgment in favor of the defendants was based entirely on this finding and, therefore, cannot stand. Accordingly, we reverse the judgment against Jinro and remand for proceedings consistent with this opinion.
REVERSED and REMANDED.