Posted: December 19, 2018

SSA Swindling? – Part III – Repleading After Dismissal – Is Plaintiffs’ Statistical Analysis Enough to Save Plaintiffs’ Claims?

This week, we return to In re SSA Bonds Antitrust Litigation, No. 1:16-cv-03711-ER (SDNY) (“In re SSA“), an action first introduced in our June 27, 2018, post, which gives a full account of the alleged collusion in the Consolidated Amended Complaint. In this post, we revisit Judge Ramos’ August 24, 2018, Opinion and Order granting the Motion to Dismiss Plaintiffs’ Consolidated Amended Complaint, previously covered in our September 4, 2018 post, and look to the Second Consolidated Amended Class Action Complaint (“SCAC” or “Second Amended Complaint”) filed November 13, 2018, after Plaintiffs were granted leave to replead and shore up deficiencies in their pleading of injury-in-fact.

Brief Overview of the Alleged Collusion

In brief, Plaintiffs are buy side funds, such as pension and retirement funds and asset management companies. They alleged that the Defendants, large dealer banks including but not limited to Bank of America, Barclays, and Credit Suisse, used their position as major players in the supranational, sub-Sovereign, and agency bonds (“SSA”) market to manipulate the bid-ask spread on SSA bonds. Plaintiffs based their allegations primarily on about 150 chats, phone calls, and other correspondence among individuals employed at the dealer-banks regarding certain deals. These 150 communications were produced by Bank of America and Deutsche Bank, who settled for a combined $65.5 million in August of 2017.

The Southern District Dismisses Without Prejudice and with Leave to Replead

In late August of 2018, Judge Ramos granted Defendants’ motion dismissing Plaintiffs’ Consolidated Amended Complaint for failure to state a claim because Plaintiffs failed to allege injury-in-fact sufficient to establish antitrust standing. Judge Ramos however granted Plaintiffs leave to replead until later in the fall of 2018. Judge Ramos based his decision primarily on the fact that, as Plaintiffs had not allegedly purchased any of the deals in question, the 150 communications on which Plaintiffs relied were not sufficient to plausibly allege a widespread conspiracy which harmed Plaintiffs, and therefore Plaintiffs had not alleged injury-in-fact sufficient to establish standing. In reaching his conclusion, Judge Ramos noted that deficiencies could potentially be shored up with, “statistical analysis of market prices and quotes or allegations based on government enforcement actions [which] may suffice to allege the expected impact of a manipulative tactic on a given market and the expected frequency of manipulation.” In re SSA Bonds Antitrust Litig., No. 16 CIV. 3711 (ER), 2018 WL 4118979, at *7 (S.D.N.Y. Aug. 28, 2018) (Citing In re London Silver Fixing, Ltd., Antitrust Litig., Nos. 14 MDL 2573, 14 Misc. 2573 (VEC), 2018 WL 3585277, at *27 n.36 (S.D.N.Y. July 25, 2018)). However, Plaintiffs only pleaded generalized academic literature, and did not plead statistical analysis, applying this literature to the facts of their case. Moreover, while Plaintiffs cited media reports that the government was investigating collusion in the SSA bond market, the reports did not go into sufficient specifics to help Plaintiffs plead injury-in-fact. Thus, “[b]ecause ‘Plaintiffs [did] not even present evidence that they traded at ‘artificial prices,’’ they have alleged ‘no actual injury …, let alone a connection between Defendants’ unlawful conduct and that non-injury.’” Id. (Citing Harry v. Total Gas & Power N. Am., Inc., 889 F.3d 104, 116 (2d Cir. 2018)).

The Second Consolidated Amended Class Action Complaint Provides Statistical Analysis, but Is it Enough?

While Judge Ramos indicated that statistical analysis may be enough to plead injury-in-fact, he stopped short of discussing at length what that statistical analysis should say in order to be sufficient. Judge Ramos at best indicated that the absence of an alleged analysis of spreads paid during and after the period of collusion was fatal. Plaintiffs revisited statistical analysis in the Second Amended Complaint noting that estimating the extent to which prices were effected by collusion could be quantified using a comparison of bid-ask spreads, as previously noted, or using an analysis of the profit margins and spreads on similar types of bonds or investment vehicles. SCAC ¶ 507.

For their analysis, Plaintiffs turned to the publicly available data from Bloomberg, which provides market-wide pricing data of the US SSA market but not data at the trade or quote level. SCAC ¶ 509. First, Plaintiffs performed a regression analysis, which, according to the Plaintiffs, included a “Collusion Indicator” or “a variable indicating whether or not the pricing information is being drawn from the core conspiracy period . .. to detect if bid-ask spreads were higher (or lower) during the alleged core conspiracy period than before or after, after controlling for [factors] that can legitimately cause spreads to vary across bonds and over time.” Those factors included: “(a) the default risk of the bond; (b) the coupon rate of the bond; (c) the time since issuance of the bond; (d) the time to maturity for the bond; (e) the issue size of the bond; (f) the total size of other issues outstanding from the same issuer; and (g) the inverse of the bond’s price. These factors are consistent with what other studies have found to be important drivers of the bid-ask spread for bonds.” This regression analysis found to a statistically significant degree, that the “Collusion Indicator was positively associated with spreads, which indicates that the alleged presence of the conspiracy is associated with higher bid-ask spreads, while its comparative absence is associated with lower bid-ask spreads.” SCAC ¶¶ 512-522.

Using this Bloomberg data, Plaintiffs also allegedly performed regression analysis to create a predictive model during the alleged conspiracy period of what bid-ask spreads “should” be based only on “legitimate economic factors” and compared that predictive model with real life market-wide data for that same period, finding that bid-ask spreads were “always higher during the core conspiracy period—and only during that period— than what can be explained by legitimate economic factors.” SCAC ¶¶ 523-525. A second predictive model was also allegedly created to predict the yields on SSA bonds rather than the spreads. While this model and actual market data were allegedly synchronized before and after the conspiracy period, with the predictive model explaining 96% of spreads and movements during those years, this yield-based predictive model was worse at predicting actual outcomes during the alleged conspiracy period by an allegedly statistically significant degree. SCAC ¶¶ 526-527. Third, Plaintiffs allegedly created a predictive regression model to predict the volatility of yields that were not explainable by legitimate economic factors, and measured this “excess” volatility, finding that when compared to real world data this excess volatility was low during the pre and post conspiracy time frame, but higher during the core conspiracy period, “consistent with Defendants pushing yields artificially low when selling, then pressing yields artificially high when buying, causing yields to bounce around more than what the economic model can account for during the core conspiracy period.” SCAC ¶¶ 528-529. Fourth, Plaintiffs allegedly created a volatility based predictive model to determine whether bid-ask spreads were higher. This again allegedly showed abnormally high bid-ask spreads during the core-conspiracy period. SCAC ¶¶ 530-531.

Plaintiffs allegedly also compared pricing behavior to determine whether bid-ask spreads were higher during the conspiracy period. They allegedly were. SCAC ¶ 533. Similar spread-based analysis was allegedly performed for US Treasury Bonds and foreign sovereign debt, again allegedly showing inflated bid-asks spreads during the core conspiracy period when compared to these other debt instruments. SCAC ¶¶ 534-537. Variation based analysis on bid-ask spreads in SSA was also performed allegedly showing that bid-ask spreads were most predictable day to day during the core-conspiracy period. This test was also run using the U.S. Treasuries as a control, and even controlling for change in the treasury market, the bid-ask spreads in USD SSA bonds showed greater diversity of spreads after the end of the core-conspiracy period when compared to before that date. SCAC ¶¶ 538-543. On the flip side, while the conspiracy was allegedly keeping bid-asks spreads consistently higher, the yields of SSA were allegedly more volatile during the core-conspiracy period based on analysis of market data. SCAC ¶¶ 544-548. Finally, Plaintiffs allegedly analyzed multiple bonds from the same issuer, which allegedly should have had a high level of yield correlation, and while yield correlation was allegedly high during the class period, it was lower when compared to the pre and post-class period. SCAC ¶¶ 549-551.

Plaintiffs thus have now allegedly analyzed many different metrics with the intent of showing that they have suffered an injury-in-fact as a consequence of Defendant’s alleged conspiracy. It will be interesting to see whether, if the Second Amended Complaint is again challenged, these metrics and “legitimate market forces controlled” models will be enough to state a claim on which relief can be granted, or if Plaintiffs still have failed to allege Defendants causation of injury-in-fact.

This post was written by Lee J. Rubin.

We welcome your feedback. If you have questions or comments about this post, please e-mail John M. Lundin, the Manipulation Monitor’s editor, at or Lee J. Rubin at or call John or Lee at (212) 344-5400.

Click here to subscribe to this or another of Schlam Stone & Dolan’s blogs.

View posts