Jesse Litvak appeals from his conviction, after a trial by a jury before Chief Judge Hall, on one count of securities fraud pursuant to 15 U.S.C. §§ 78j(b), 78ff. This is the second time this matter has been before us. In the first appeal, see *59 United States v. Litvak , 808 F.3d 160 (2d Cir. 2015) (" Litvak I "), appellant challenged his convictions on ten counts of securities fraud and various other counts of fraud and false statements. We reversed the convictions for fraud and false statements. We vacated the securities fraud counts and remanded them for a new trial. Id. at 190 . This appeal follows the second trial on ten counts of securities fraud. Appellant was acquitted on nine of the ten counts, each involving transactions similar to the one for which he was convicted.

He challenges his conviction on the one count principally on the ground that his misstatements were, as a matter of law, immaterial to a reasonable investor in the market for residential mortgage-backed securities ("RMBS"). We reject that argument. However, the district court erred in admitting evidence that the individual representing the counterparty in the transaction for which appellant was convicted believed appellant to be an agent acting on the counterparty's behalf. All parties agree that the belief in an agency relationship was erroneous and that appellant acted solely as a principal in the transaction underlying this appeal. Indeed, the testimony of the counterparty's representative indicated that the counterparty, his employer, had informed him that no agency relationship existed but that the representative disagreed. The district court instructed the jury that no agency relationship existed.

Because the applicable materiality test is an objective one, evidence of the idiosyncratic and erroneous belief of the counterparty's representative was irrelevant. The belief of the counterparty's representative in an agency relationship was, as discussed infra , the only evidence distinguishing appellant's count of conviction from those of the nine transactions on which he was acquitted. The evidence was, therefore, prejudicial, as our earlier opinion in this proceeding foreshadowed, id. at 187 , and we cannot conclude with fair assurance that the jury would have convicted appellant absent such evidence. See United States v. Rosemond , 841 F.3d 95 , 112 (2d Cir. 2016). Accordingly, we vacate the judgment of conviction and remand.

BACKGROUND

This case has a lengthy procedural history. In January 2013, the government filed an indictment charging appellant with eleven counts of securities fraud in violation of 15 U.S.C. §§ 78j(b), 78ff (Counts 1-11), one count of fraud against the Troubled Asset Relief Program ("TARP") in violation of 18 U.S.C. § 1031 (Count 12), and four counts of making false statements in violation of 18 U.S.C. § 1001 (Counts 13-16).

The instruments involved in the alleged fraud were RMBS. Our decision on the first appeal reversed the TARP fraud and false statements counts involving government agencies. Litvak I , 808 F.3d at 190 . It vacated the securities fraud counts on the grounds that the district court erroneously excluded evidence relevant to appellant's defense that his misstatements were not material to a reasonable investor in the RMBS market. Id. at 179-85 .

a) The RMBS Market

Understanding the RMBS market is critical to this appeal. What follows in this section of the opinion is materially undisputed on this record.

Appellant worked as a bond trader at Jefferies & Company ("Jefferies"), an investment banking firm and securities broker-dealer. Appellant bought and sold RMBS. RMBS are large and complex aggregations of residential mortgages and home equity loans. "Typically, an entity (such as a bank) will buy up a large number *60 of mortgages from other banks, assemble those mortgages into pools, securitize the pools (i.e., split them into shares that can be sold off), and then sell them, usually as bonds, to banks or other investors." City of Pontiac Policemen's & Firemen's Ret. Sys. v. UBS AG , 752 F.3d 173 , 177 n.7 (2d Cir. 2014) (quoting Litwin v. Blackstone Grp., L.P. , 634 F.3d 706 , 710 n.3 (2d Cir. 2011) ). A pool of mortgages may also be divided into tranches, sold as bonds, that have varying risk profiles and returns to investors. Holders of RMBS receive coupon payments derived from homeowner payments on mortgages. RMBS are bought and sold at very high prices. For example, the single bond involved in the sole count of conviction was conveyed for over $23 million.

RMBS are marketed to large, sophisticated financial institutions. Because RMBS are not homogeneous-each has unique features affecting its value-they are not publicly traded on an exchange like NASDAQ or the New York Stock Exchange. At any given time, therefore, there is no public list of RMBS available for sale or of the financial terms of ongoing transactions. As a result, investors do not buy and sell RMBS directly. Instead, when institutional investors are interested in buying or selling RMBS, they contact registered broker-dealers such as Jefferies to find interested buyers or sellers. Investors may also buy and sell RMBS directly with broker-dealers that hold RMBS in their own accounts.

Most, if not all, institutional investor-buyers use computer models to establish the highest amount at which they would be willing to buy or sell a particular RMBS. These analyses can be quite complex. Factors considered include the priority level of the bond, the likely number of individuals that will prepay their mortgages, the likely number of mortgage defaults, the potential severity of the loss on a defaulted mortgage, the locations of the real property mortgaged, unemployment rates in those locations, the effectiveness of the home loan servicer in receiving payments, and the credit scores of individual homeowners.

RMBS are bought or sold through three methods, styled as: "inventory trades," "order trades," or "bids-wanted-in-competition trades" ("BWIC trades"). In an inventory trade, an investor buys a bond already held in a broker-dealer's account. In an order trade, a broker-dealer communicates with an interested buyer and seller and, if successful, effectuates a transaction in which a RMBS is transferred. A RMBS investor may approach a broker-dealer to express interest in buying or selling a RMBS in a certain price range. Or a broker-dealer may approach an investor to gauge its interest in buying or selling a bond in a certain price range. The broker-dealer owns the bond, but usually briefly, in consummating the transaction between the two investors.

A BWIC trade was involved in the sole count on which appellant was convicted. A BWIC trade is similar to an order trade, except that a BWIC trade involves an auction in which a putative seller sends a bid-list to multiple broker-dealers. These broker-dealers solicit expressions of interest and price ranges from potential buyers. The broker-dealer then places a bid in the auction of a particular security. The bid price is selected by the broker-dealer and may well differ from prices suggested by putative buyers. If the broker-dealer's bid is successful, the broker-dealer buys the bond and then offers to sell it to the interested investor. Communications are generally carried out through online instant messages.

There are various quirks to how bond prices are quoted among market participants. First, bond prices are quoted in terms of the price per $100 of face value, *61 not the total price. Second, bond prices that are not whole numbers are represented in "ticks," 1/32 of a dollar or approximately $0.03. For example, a bond priced at $79.75 is represented as 79-24. Further, a broker-dealer is compensated with reference to either the "all-in" or "on-top" price. An "all-in" price is the total price the investor is (the broker-dealer hopes) to pay a broker-dealer for a bond. It does not refer to the price at which the broker-dealer purchased the bond. An "on-top" price refers to the difference between the price the broker-dealer paid for the bond and the price at which the broker-dealer will sell the bond-the "on-top" price is the broker-dealer's profit. In "on-top" trades, the broker-dealer's profit is sometimes referred to colloquially as a "commission," "markup," or "spread." In general, inventory trades are quoted as "all-in" prices. Negotiations over order or BWIC trades may refer to the broker-dealer's acquisition cost and the "on-top" price.

An essential feature of all of these trades, however, is that the broker-dealer acts solely in its own interest as a principal. The broker-dealer continually tracks potential buyers and sellers, their interests in particular kinds of RMBS, and their ongoing acceptable price ranges. It seeks to profit from transactions in the securities by buying low and selling high. The size of the year-end discretionary bonus given to each broker-dealer's trader is largely based on the profitability of the individual's trades. The broker-dealer assumes the risk of buying the bond, and an institutional investor can refuse to purchase a bond held by the broker-dealer even when the investor caused the broker-dealer to purchase it by an expression of interest, i.e. , in an order or BWIC trade. 1 Such an investor may also have no investment purpose but may intend to resell the bond immediately, and at a higher price, to another institution it knows to be interested.

A broker-dealer is not, therefore, an agent for its counterparties in these trades and owes them no special or fiduciary duty. In a sale by a counterparty to a broker-dealer, the counterparty has no legitimate expectation that the broker-dealer will resell the bond at the price paid to the counterparty. Similarly, in a purchase, the counterparty has no legitimate expectation of purchasing a bond at the price paid by the broker-dealer. Rather, the broker-dealer and counterparty each have their own price ranges in which they will consummate their ends of the transactions. The final price is determined in an arms-length negotiation and, if agreed upon, will be somewhere in the overlap of price ranges.

Market participants protect themselves against behavior that departs from market norms of good faith largely without resort to expensive and time-consuming litigation. Institutional investors simply refuse to do business with perceived bad actors for either a period of time or indefinitely. This is referred to in the industry as placing an individual or firm in the "penalty box," Tr. at 174, and is "not out of the ordinary," id. For example, after AllianceBernstein discovered appellant's misstatements, the firm put Jefferies in the "penalty box" for approximately one month. There was also testimony that AllianceBernstein "still [doesn't ] do business with Jefferies very much at all." Id. at 176. Jefferies fired appellant shortly after AllianceBernstein's discovery.

*62 b) The First Trial and Appeal

In the first trial, appellant was convicted on all counts apart from Count Seven, which had been dismissed before trial. Litvak I , 808 F.3d at 166-69 . Appellant appealed, and we reversed the judgment of conviction on the counts alleging fraud against TARP and making false statements. Id. at 190 .

Ten transactions were involved in the remaining securities fraud counts. In each count, appellant was accused of lying about the price at which Jefferies had bought or sold the particular RMBS involved. These misstatements were made to increase Jefferies's profits. For example, by overstating to a proposed counterparty the amount he paid for a bond, appellant sought to increase the all-in price paid by the counterparty and Jefferies's profit on the transaction.

We also vacated his convictions on the securities fraud counts but remanded for a new trial on those counts. Id. In both trials, the government sought to prove the element of materiality in the securities fraud counts through the testimony of representatives of appellant's counterparties. They generally testified that they considered appellant's misstatements to be important in deciding whether to sell or buy the RMBS in question. On the first appeal, we held that the district court erred in excluding the testimony of two experts proffered by the defense as to how reasonable investment managers value RMBS, id. at 182-88 , and as to how the profit sought by the broker-dealer was so minor compared to the bond price that misstatements regarding that profit were not material to a reasonable investor, id. at 184-85 .

The excluded experts were also to testify as to the "arm's-length nature of the relationship between a broker-dealer and counterparty." Id. at 186 . This testimony was offered by the defense to rebut a purported agency relationship testified to by some representatives of the counterparties. See id. Because the exclusion of the expert testimony on how RMBS are valued required vacaturs of all the securities fraud convictions, id. at 184 , we declined to determine whether the exclusion of appellant's proferred expert testimony on agency relationships, while error, was harmless. Id. at 188 . At the time of our opinion, there was uncertainty as to whether the agency issue would even arise at a second trial because, at the oral argument of the first appeal-the proverbial eleventh hour-the government abandoned any claim of an agency relationship. See id. at 187 n.32. With regard to the agency issue, however, we noted that the existence of an agency relationship was of "great import," id. at 187 , because, if such a relationship existed, appellant's misstatements would have been more important to a reasonable investor while, if such a relationship was lacking, the misstatements would have been less so. Id.

c) The Second Trial and Appeal

Appellant was retried on the ten securities fraud counts in early 2017. The jury convicted appellant on Count Four and acquitted him on all other counts. The following evidence, viewed in the light most favorable to the government, United States v. Brock , 789 F.3d 60 , 63 (2d Cir. 2015), is relevant to appellant's conviction on Count Four.

Appellant, on behalf of Jefferies, was involved in selling a RMBS in a BWIC trade to Invesco Ltd., an investment management company with nearly $1 trillion of assets under management. On July 1, 2010, having received a bid-list of bonds from a putative seller, appellant circulated the bid-list to, inter alia , Brian Norris, who represented Invesco. Invesco's credit research team determined that Invesco *63 would make a worthwhile profit on a particular RMBS on the list, SARM 2005-21 7A1, if the purchase price were at, or under, $80 per $100 of face value. Norris then suggested to appellant that he bid 79-24 for the bond, and further told appellant that he had "some room" to bid higher than 79-24. Tr. at 748, 772.

Appellant purchased the bond and informed Norris that Jefferies won the auction, stating in an online chat that "[I] bid your level," Joint App'x at 723, that is, he bid 79-24. In fact, however, appellant had bid and purchased the bond for 79-16. Norris then proposed to buy the bond from appellant for 79-30, asking "6 ticks cool? 79-30 to me?" Id. Appellant accepted, saying "6/32s is great." 2 Id. The difference between fourteen ticks and six amounts to $73,018.53. Gov't Br. at 11. The total amount Invesco paid Jefferies for the bond was approximately $23.6 million.

Before appellant's second trial, appellant made motions in limine to bar counterparty representatives from testifying that they believed appellant was acting as their agent. Notwithstanding the concession at oral argument on the first appeal, the government proposed to offer such testimony, and the district court allowed it. The court stated that such testimony was "plausibly ... part of the government's efforts to establish the materiality of the information [appellant ] misrepresented, and could likewise contribute[ ] to the inference that [appellant ] was acting with fraudulent intent." United States v. Litvak , No. 13-cr-19, Order on Motion in Limine (D. Conn. June 28, 2016), at 4 (internal quotation marks and citation omitted ). Furthermore, the court reasoned that "because these witnesses are testifying to their own point of view- i.e. , what they understood their relationship to [appellant ] to be, based on the events leading up to these trades as they experienced them-this testimony is unlikely to confuse the jury or to be unduly prejudicial to [appellant ]." Id.

Norris then testified that he believed appellant to be his agent, and that broker-dealers "serve as an agent in between [buyers and sellers]." Tr. at 748. Joel Wollman, who worked for a different firm and was a counterparty representative to two transactions on which appellant was acquitted, also testified that appellant was "acting as [his] agent." Id. at 859.

Certainly, by the end of the second trial, if not much earlier, it was clear that appellant had never been Norris's (or Wollman's) agent, and an agency relationship did not exist in such transactions. Indeed, Invesco's compliance personnel had informed Norris (before the transaction) that broker-dealers were not agents of bond buyers in BWIC trades. During its summation, the government labeled the agency issue a "red herring." Id. at 1456.

*64 It stated that it "never claimed ... that ... Litvak was ever acting as an agent," id. , and that a "broker-dealer is not acting as someone's agent," id. at 1457. Instead, the prosecution argued that the critical point of this testimony was to show how appellant "ch[ ]ose to establish a relationship of trust to lead them all to think that he was trustworthy." Id. 3 Finally, the district court instructed the jury that appellant was not an agent of Norris or Invesco.

After a motion for a judgment of acquittal or a new trial was denied, the court sentenced appellant to 24 months' imprisonment, three years of supervised release, and a $2 million fine. This appeal followed.

DISCUSSION

Appellant challenges his conviction on two principal grounds. First, he argues that, as a matter of law, the government did not prove that his misstatements were material. Because the materiality of a defendant's misstatements to a reasonable investor is an element of the charged crime, our adoption of any variation of this argument would preclude a new trial. See United States v. Robinson , 545 F.2d 301 , 304 n.5 (2d Cir. 1976). However, we disagree. Second, he argues that the district court erred in admitting testimony on Norris's erroneous belief that appellant acted as Invesco's agent. We agree, vacate the conviction, and remand.

a) Materiality

To prove securities fraud in a civil action under Section 10(b) of the Securities Exchange Act of 1934, codified as 15 U.S.C. § 78j(b), 4 the government must prove that, "in connection with the purchase or sale of a security the defendant, acting with scienter, made a material misrepresentation (or a material omission if the defendant had a duty to speak) or used a fraudulent device. 5 In order to impose criminal liability, the government must also prove that the defendant willfully violated the law." United States v. Vilar , 729 F.3d 62 , 88 (2d Cir. 2013) (internal quotations and citations omitted ). This appeal focuses largely on the element of materiality.

A misstatement in a securities transaction is material so long as there is "a substantial likelihood that a reasonable investor would find the ... misrepresentation important in making an investment decision." Id. at 89 ; see United States v. Contorinis , 692 F.3d 136 , 143 (2d Cir. 2012). A misrepresentation is important if there is "a substantial likelihood that the disclosure of the omitted fact would have been viewed by the reasonable investor as having significantly altered the 'total mix' of information made available." Basic Inc. v. Levinson , 485 U.S. 224 , 231-32, 108 S.Ct. 978 , 99 L.Ed.2d 194 (1988).

The standard of a "reasonable investor," like the negligence standard of a "reasonable man," is an objective one. See TSC Indus., Inc. v. Northway, Inc. , 426 U.S. 438 , 445, 96 S.Ct. 2126 , 48 L.Ed.2d 757 (1976) ; Long v. Abbott Mortg. Corp. , 459 F.Supp. 108 , 116-17 (D. Conn. 1978) (citing cases). The standard may vary, *65 therefore, with the nature of the traders involved in the particular market. See Basic , 485 U.S. at 236 , 108 S.Ct. 978 ; Michaels v. Michaels , 767 F.2d 1185 , 1197 (7th Cir. 1985). The reasonable investor in a market in which many individual investors trade will be deemed to be somewhat less schooled and sophisticated than a reasonable investor in a market, like the one now before us, in which only institutions trade with the help of complex computer programs and professional traders. See, e.g. , Marini v. Adamo , 995 F.Supp.2d 155 , 184 (E.D.N.Y. 2014), aff'd , 644 Fed.Appx. 33 (2d Cir. 2016) ; In re Initial Pub. Offering Sec. Litig. , 241 F.Supp.2d 281 , 377 n.142 (S.D.N.Y. 2003) ; Lenz v. Associated Inns & Rest. Co. of Am. , 833 F.Supp. 362 , 375-76 (S.D.N.Y. 1993).

The objective nature of the "reasonable investor" standard is of particular importance in this case. To prove materiality, the government relied heavily on the testimony of counterparty traders-purported victims-that they considered appellant's misstatements to be an important factor, among others, in their investment decisions. The district court believed it important that counterparty traders testify to "their own point of view." Litvak , No. 13-cr-19, Order on Motion in Limine, at 4. This approach is permissible in a case like this, but only so long as the testimony about the significance of the content of a defendant's misstatements and each trader's "own point of view" is shown to be within the parameters of the thinking of reasonable investors in the particular market at issue. In other words, there must be evidence of a nexus between a particular trader's viewpoint and that of the mainstream thinking of investors in that market. Materiality cannot be proven by the mistaken beliefs of the worst informed trader in a market.

A finding of materiality does not require proof of "actual reliance." Vilar , 729 F.3d at 89 . Materiality requires proof only that a reasonable investor would deem the content of a misstatement a substantial factor to be considered in the making of the particular investment decision. Unlike common law fraud or a civil action for damages where an investor seeks compensatory damages, proof of harm is not necessary in a criminal prosecution under Section 10(b), so long as materiality, intent to defraud, and a connection to a securities transaction are shown.

b) Materiality and Value

Appellant challenges the government's proof of materiality as legally insufficient. We review the sufficiency of the evidence de novo . United States v. Coplan , 703 F.3d 46 , 62 (2d Cir. 2012) (citing United States v. Yannotti , 541 F.3d 112 , 120 (2d Cir. 2008) ). "As a general matter, a defendant challenging the sufficiency of the evidence bears a heavy burden, as the standard of review is exceedingly deferential." Brock , 789 F.3d at 63 (citing Coplan , 703 F.3d at 62 ). The evidence must be viewed "in the light most favorable to the [g]overnment, crediting every inference that could have been drawn in the [g]overnment's favor, and deferring to the jury's assessment of witness credibility and its assessment of the weight of the evidence." Id. (citing Coplan , 703 F.3d at 62 ). A conviction will be upheld "if any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt." Id. (internal quotation marks and citations omitted ).

Appellant contends that, as to the count of conviction, the government presented evidence showing only that Norris agreed to a price for the bond that was based on Invesco's internal analysis of value completed before he spoke to appellant. According *66 to appellant, Norris "was primarily interested in obtaining the bond, not in minimizing Jefferies'[s] profit." Appellant Br. at 43.

However, on Count Four, there was sufficient evidence for a rational jury to find appellant's misstatements material beyond a reasonable doubt, whether or not Norris paid "the price he proposed to pay." Id. at 44. As noted, the determination of the materiality of a misstatement or omission "is an objective one, involving the significance of an omitted or misrepresented fact to a reasonable investor." TSC Indus. , 426 U.S. at 445 , 96 S.Ct. 2126 ; see Amgen Inc. v. Conn. Ret. Plans & Trust Funds , 568 U.S. 455 , 459, 133 S.Ct. 1184 , 185 L.Ed.2d 308 (2013) ("materiality is judged according to an objective standard ..."). A finding of materiality, therefore, requires a showing only of importance, not "actual reliance." Vilar , 729 F.3d at 89 .

In Litvak I , we held that the testimony of various counterparty representatives was sufficient to sustain a finding of materiality on the counts of securities fraud. See 808 F.3d at 175-76 ("The trial record includes testimony from several representatives of Litvak's counterparties that his misrepresentations were important to them in the course of the transactions on which the securities fraud charges were predicated ...") (internal quotation marks and citations omitted ). In that regard, we also held that exclusion of appellant's experts' testimony was not harmless because the government was relying on the subjective views of "purported victims," id. at 184 , and the expert testimony was intended to show that, under an objective materiality test, the misstatements would not have been deemed significant by a reasonable investor, id.

In the first trial, the district court excluded expert testimony that minor price variations would be irrelevant to sophisticated investors. We held that this exclusion was also error, although we did not reach the harmlessness issue. Id. at 185 . Therefore, some misstatements may be so "obviously unimportant" as to be immaterial as a matter of law. 6 Feinman v. Dean Witter Reynolds, Inc. , 84 F.3d 539 , 540-41 (2d Cir. 1996) (quoting Goldman v. Belden , 754 F.2d 1059 , 1067 (2d Cir. 1985) ); see also TSC Indus. , 426 U.S. at 450 , 96 S.Ct. 2126 (applying "obvious" standard to definition of materiality for purposes of Section 14(a) of the Securities Exchange Act ).

In Feinman , brokers added transaction fees of a few dollars that may have "exceed[ed ] the broker[s'] actual handling charges." 84 F.3d at 541 . However, the brokers "did not mislead their customers as to what portion of the total transaction cost was going toward purchasing securities." Litvak I , 808 F.3d at 176 . Here, unlike the market in Feinman , "competition among the firms in the labeling and pricing of their services," id. at 177 (quoting Feinman , 84 F.3d at 541 ), is not available. Further, the Feinman transaction occurred in an open market, but a misrepresentation in the RMBS market is "virtually undiscoverable." Id. 7

*67 Appellant also argues that his misstatements cannot, as a matter of law, be material because they were not relevant to the intrinsic value of the bond, at best affecting only "the negotiation over price." Appellant Br. at 30. This argument appears to have been considered and rejected in Litvak I . See Litvak I , 808 F.3d at 175-78 . Nothing in the record of the second trial causes us to reach a different result.

We cannot agree that, because statements about the price paid by the broker-dealer for a RMBS are not intended, or understood, as relevant to the intrinsic value of the bond, such assertions are not material as a matter of law. 8 When the broker-dealer seeks a profit for its role in procuring and selling a security desired by a buyer, the profit becomes part of the price paid by the buyer. The value of the security may be the most important factor governing the decision to buy, but the price must be considered in determining whether the purchase is deemed profitable. 9 The broker-dealer's profit is part of the price and lies about it can be found by a jury to "significantly alter[ ] the 'total mix' of information ... available." Basic , 485 U.S. at 231-32 , 108 S.Ct. 978 (internal citation omitted ).

(c) Evidentiary Rulings

Appellant argues that the district court erred by admitting Norris's testimony that appellant was Invesco's agent, and that this evidence may have tipped the balance in favor of the jury's conviction on that count. We agree.

(i) Error

"We review a district court's evidentiary rulings under a deferential abuse of discretion standard, and we will disturb an evidentiary ruling only where the decision to admit or exclude evidence was 'manifestly erroneous.' " United States v. McGinn , 787 F.3d 116 , 127 (2d Cir. 2015) (quoting United States v. Samet , 466 F.3d 251 , 254 (2d Cir. 2006) ). Even if a decision was "manifestly erroneous," we will affirm "if the error was harmless." Id. (citing United States v. Miller , 626 F.3d 682 , 688 (2d Cir. 2010) ).

In the first trial, the issue of whether appellant was acting as a principal or as an agent in the various charged transactions was disputed. Only at oral argument of the first appeal did the government concede that appellant was acting as a principal. Nevertheless, in the second trial and at the government's urging, the district court again admitted testimony from Norris that appellant was acting as Invesco's agent during the trade. The district court held that the testimony was relevant to the materiality of appellant's misstatements and to appellant's fraudulent intent. The court also noted that the testimony was unlikely to confuse or prejudice the jury because it went only to Norris's "own point of view." 10 Litvak , No. 13-cr-19, Order on Motion in Limine, at 4.

*68 In Litvak I , we noted the importance of an agency relationship to the materiality issue. We stated that testimony about whether appellant was the buyer's agent was relevant because it might cause a jury to

construe [Litvak's misstatements] as having great import to a reasonable investor if coming from the investor's agent . If, on the other hand, Litvak was acting on his own behalf ( i.e. , as a principal ), and not as the purported victims' agent, a jury may well construe that relationship as providing a distance between Litvak and a reasonable investor, which may tend to show that his statements could not have been reasonably viewed as important in the course of a transaction.

Litvak I , 808 F.3d at 187 (citation omitted ) (emphasis in original ).

It is now undisputed that appellant never acted as Invesco's or Norris's agent during the trade in question. Appellant acted as a principal: Jefferies bought the bond from one party, held it for a brief period of time, and then sold it to Invesco. The transaction was at arms-length, and appellant owed no fiduciary duties to the buyer. As the owner of the bond, Jefferies did not have to convey the bond to Invesco and assumed the risk that Invesco would walk away from the deal after Jefferies had bought the bond.

To repeat, the lack of an agency relationship is undisputed. The district court instructed the jury that Litvak was not Invesco's agent. Indeed, the government argued in summation that the agency issue was a "red herring" and that it had never claimed that Litvak was Invesco's agent. Tr. at 1456.

Evidence is relevant if "(a) it has any tendency to make a fact more or less probable than it would be without the evidence; and (b) the fact is of consequence in determining the action." Fed. R. Evid. 401. Relevancy is a "very low standard." United States v. White , 692 F.3d 235 , 246 (2d Cir. 2012) (quoting United States v. Al-Moayad , 545 F.3d 139 , 176 (2d Cir. 2008) ). "To be relevant, evidence need not be sufficient by itself to prove a fact in issue, much less to prove it beyond a reasonable doubt." United States v. Abu-Jihaad , 630 F.3d 102 , 132 (2d Cir. 2010).

At the government's urging, the district court admitted Norris's testimony as relevant to the materiality issue. It believed that any prejudice was unlikely because Norris would be "testifying to [his] own point of view." Litvak , No. 13-cr-19, Order on Motion in Limine, at 4. On appeal, the government argues that Norris's belief in an agency relationship was relevant to whether the "victims trusted Litvak." Gov't Br. at 48. It adds that "[e]vidence as to whether and why victims trusted Litvak or believed his representations was probative as to what an objectively reasonable investor would have thought, and thus relevant to the disputed issue of materiality." Id. at 49-50. If trust exists, the government argues, a jury could interpret misstatements by appellant as more likely to be material.

However, the materiality standard is an objective one and centers on the views of a hypothetical, reasonable investor in the market at issue. See Amgen , 568 U.S. at 467 , 133 S.Ct. 1184 ("[t ]he question of materiality ... is an objective one, involving the significance of an omitted or misrepresented fact to a reasonable investor") (quoting TSC Indus. , 426 U.S. at 445 , 96 S.Ct. 2126 ); Chris-Craft Indus., Inc. v. Piper AircraftCorp. , 480 F.2d 341 , 363 (2d Cir. 1973) ("The materiality test is concerned *69 only with whether a prototype reasonable investor would have relied."). It can hardly be the law that the "point of view," Litvak , No. 13-cr-19, Order on Motion in Limine, at 4, of an investor who is admitted to be wrong-the government, in putting the best face on it, describes Norris as "confused" and "incorrect"-is relevant to prove what a reasonable investor, neither confused nor incorrect, would have deemed important. The agency issue raised by the government's proffer of Norris's testimony was indeed an irrelevant "red herring." Tr. at 1456.

While the individual views of a counterparty trader may usually be relevant to the nature of the market involved and as to the beliefs of a reasonable investor, a reasonable investor would not misperceive the role of a broker-dealer in the RMBS market. See Litvak I , 808 F.3d at 187 n.32. In fact, Norris received material from Invesco's legal and compliance department stating that, in transactions such as the one here, broker-dealers act as principals. His disagreement with that advice was unreasonable. Norris's indisputably idiosyncratic and unreasonable viewpoint is not, therefore, probative of the views of a reasonable, objective investor in the RMBS market. Even if Norris's testimony had relevance, 11 the district court abused its discretion under Rule 403. Relevant evidence may be excluded "if its probative value is substantially outweighed by a danger of one or more of the following: unfair prejudice, confusing the issues, misleading the jury, undue delay, wasting time, or needlessly presenting cumulative evidence." Fed. R. Evid. 403. At best, Norris's testimony about the supposed agency relationship had a high probability of confusing the jury by asking it to consider as relevant the perception of a counterparty representative that was entirely wrong. At worst, Norris's testimony on an agency relationship would mislead the jury based on the government's argument that a perceived relationship of trust showed materiality. This argument itself is flawed because it equates an indisputably incorrect personal belief with an objective test of materiality.

In its summation to the jury, the government sought to cabin the effect of Norris's testimony by emphasizing that it never claimed that appellant acted as an agent. 12 It further labeled the defense's emphasis on the agent-principal distinction a "red herring," a remark that all but concedes irrelevance. Tr. at 1456. It argued in summation that appellant created the perception of acting as an agent and that he aimed to establish a "relationship of trust." Id. at 1457. But the government points to no evidence that appellant was the source of Norris's mistaken belief in an agency relationship or took steps sufficient to cause Norris to disregard the common knowledge of the market, much less the information provided Norris by Invesco's legal and compliance department. 13 The *70 government's concept of subjective trust as evidence of materiality became a back door for the jury to apply the heightened expectations of trust that an agency relationship carries. The government's argument similarly ignores the settled law that those at either end of an arms-length transaction are acting only in their own self-interest. See In re Mid-Island Hosp., Inc. , 276 F.3d 123 , 130 (2d Cir. 2002) ("[W]hen parties deal at arms length in a commercial transaction, no relation of confidence or trust sufficient to find the existence of a fiduciary relationship will arise absent extraordinary circumstances.") (quoting Pan AmCorp. v. Delta Air Lines, Inc. , 175 B.R. 438 , 511 (S.D.N.Y. 1994) ) (alteration in original ).

(ii) Harmlessness Analysis

The jury's verdict prevents us from viewing the admission of the agency testimony as harmless because we cannot "conclude with fair assurance that the error[ ] did not substantially influence the jury." Rosemond , 841 F.3d at 112 (internal quotation marks and citations omitted ). The following factors are weighed in determining whether the admission of evidence was harmless: "(1) the overall strength of the prosecutor's case; (2) the prosecutor's conduct with respect to the improperly admitted evidence; (3) the importance of the wrongly admitted testimony; and (4) whether such evidence was cumulative of other properly admitted evidence." McGinn , 787 F.3d at 127-28 (quoting UnitedStates v. Gomez , 617 F.3d 88 , 95 (2d Cir. 2010) ). Applying this test, we conclude that: (1) the government's case on materiality was not overwhelming and was vigorously contested; (2) the government proffered the testimony in question-although admitting that the testimony was based on factual and legal error and was a "red herring"-and argued to the jury that a "relationship of trust" existed; (3) the error was of great importance in view of the verdict, as discussed infra ; and (4) the testimony was not cumulative of properly admitted testimony.

Returning to McGinn factor (3), introduction of the evidence was evidently of great importance to the jury. The acquittals were almost certainly based on the jury's acceptance of the defense's materiality theory that was largely suppressed in the first trial. On Count Four, Norris's testimony about a perceived agency relationship was the only rational reason for the jury to have convicted appellant on that count of securities fraud while acquitting him on all other counts.

The amount of money lost was not a differentiating factor between the counts. Four counts on which appellant was acquitted involved gains that were much larger than the $73,000 plausibly at issue in Count Four. Count Two involved a plausible gain of $602,000. Count Six involved a gain of $212,000. Count Eight involved a gain of $229,000. Count Ten involved a gain of $180,000.

The government's "relationship of trust" argument, based largely on Norris's testimony, contributed to the prejudice on Count Four by stressing the personal view of Norris. In seven of the charged trades, the counterparty representative either did not testify, or stated that, in general, they viewed statements by broker-dealers with suspicion and did not consider appellant to be their agent. For example, a counterparty representative, Michael Canter, an employee of AllianceBernstein, stated that information *71 volunteered by traders should be viewed with skepticism. He also noted that he understood that appellant was trading as a representative of a principal. His testimony as to any trust of appellant was admittedly personal. Katherine Corso of York Capital abandoned plans to set up an auction to sell a bond in favor of selling it to appellant for what she considered to be an agreeable price, i.e. , a better deal than an auction. Corso further understood that appellant was acting as a representative of a principal in the transaction with all the accompanying risk that status entailed. 14 Vladimir Lemin, representing Magnetar Capital, testified that information in the market "could be less than truthful." Tr. at 736. According to him, such negotiations had to be approached with care. Three other counterparty representatives in charged trades failed to testify. In acquitting on these counts, the jury appears to have credited the substantial evidence that investment managers view with suspicion statements by broker-dealers in the context of arms-length transactions.

The acquittals on Counts Nine and Ten-which involved two trades in addition to the seven charged trades discussed above-call for particular scrutiny because they involved trades with Wollman, the only counterparty representative apart from Norris to testify that he believed appellant was acting as his agent. The government argues that these two acquittals show that the jury was not swayed by the testimony on agency in convicting appellant. Instead, it argues, if the jury was swayed, "one would have expected convictions on every trade in which either one [Norris or Wollman] was the victim." Gov't Br. at 52.

However, Wollman's credibility in that regard was severely weakened. Wollman admitted that he viewed negotiations with broker-dealers with skepticism. He also admitted that, in particular, he viewed appellant's statements during one of the transactions with suspicion. There was also evidence that, during that same transaction, Wollman concealed relevant information from appellant. Wollman did not tell appellant that he intended to immediately resell half of the bond to Royal Bank of Scotland Securities ("RBS Securities") for a higher price. While negotiating, Wollman told appellant that the bond would have an acceptable return at one purchase price, yet he had agreed with RBS Securities that the bond would have the same return at a higher purchase price. On redirect, Wollman admitted that this behavior may have misled appellant: "If [appellant ] knows that I, then, went ahead and sold it to [RBS Securities], he likely misses out on that future transaction. And he would have reason to be upset about not getting to partake in that future transaction." Tr. at 1007. This conduct is not consistent with the view that appellant was Wollman's agent.

*72 The defense also introduced online chats claiming that Wollman engaged in other deceptive practices -if true, likely criminal acts under the government's theory of this case-not involving appellant. Wollman was said by the defense-Wollman denied the inference-to have told a broker-dealer to invent a seller-who was presumably interested in selling a bond to Wollman at an attractive price-to coax an individual into selling a bond. And Wollman, when trying to sell a bond in a BWIC auction, may have made up a bid to stoke interest in the sale.

While Norris acknowledged that transactions in the RMBS market generally had to be approached with skepticism, his credibility was not similarly impeached. A jury invited by the government's "relationship of trust" argument to consider the subjective views of counterparty representatives might well convict on the count involving Norris while acquitting on the counts involving Wollman.

In light of the jury's verdict, therefore, we cannot "conclude with fair assurance," Rosemond , 841 F.3d at 112 , that Norris's testimony did not "substantially influence the jury," id. , in convicting appellant. We conclude that the district court's error in admitting testimony on agency was not harmless and requires a vacatur.

CONCLUSION

Accordingly, we vacate and remand the judgment of conviction. Because appellant is currently incarcerated, we order that he be released on appropriate bond pending further proceedings.