Bricklayers & Trowel Trades International Pension Fund v. Credit Suisse Securities (USA) LLC , 752 F.3d 82 ( 2014 )

United States Court of Appeals
For the First Circuit
No. 12-1750
BRICKLAYERS AND TROWEL TRADES INTERNATIONAL PENSION FUND,
Plaintiff, Appellant,
JAMES UPHOFF; GOODMAN FAMILY TRUST; MALKA BIRNBAUM, on behalf
of herself and all others similarly situated; NEIL MCCARTY;
RODNEY W. NARBESKY, individually and on behalf of all others
similarly situated,
Plaintiffs,
v.
CREDIT SUISSE SECURITIES (USA) LLC; CREDIT SUISSE (USA), INC.;
JAMIE KIGGEN; FRANK P. QUATTRONE; LAURA MARTIN; ELLIOT ROGERS,
Defendants, Appellees.
APPEAL FROM THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF MASSACHUSETTS
[Hon. Nathaniel M. Gorton, U.S. District Judge]
Before
Howard, Circuit Judge,
Souter,* Associate Justice
and Torresen,** District Judge.
Frederic S. Fox, with whom Kaplan Fox & Kilsheimer LLP and
Shapiro Haber & Urmy LLP were on brief, for appellant.
*
Hon. David H. Souter, Associate Justice (Ret.) of the Supreme
Court of the United States, sitting by designation.
**
Of the District of Maine, sitting by designation.
Lawrence Portnoy, with whom Daniel J. Schwartz, Jonathan K.
Chang, Dharma Betancourt Frederick, Davis Polk & Wardwell LLP,
Robert Buhlman, Siobhan E. Mee, Amanda V. Muller, and Bingham
McCutchen LLP were on brief, for appellees.
May 14, 2014
HOWARD, Circuit Judge. Alleging violations of Sections
10(b) and 20(a) of the Securities Exchange Act and of SEC Rule 10b-
5, the appellant pension fund and other America Online ("AOL")
shareholders brought this class action against Credit Suisse First
Boston ("CSFB"), former CSFB analysts Jamie Kiggen and Laura
Martin, and other related defendants. The shareholders claim that
CSFB fraudulently withheld relevant information from the market in
its reporting on the AOL-Time Warner merger, and that the
shareholders purchased stock in the new company at prices that were
artificially inflated as a result of the defendants' purposeful
omissions. This appeal concerns the admissibility of the opinion
of the shareholders' expert Dr. Scott D. Hakala, whose testimony
the district court precluded for lack of reliability. We find no
abuse of discretion in that decision. We also agree with the
district court that, without the expert's testimony, the
shareholders are unable to establish loss causation. Summary
judgment was therefore properly awarded to the defendants.
I. Background
A. Facts
On January 11, 2001, Time Warner Inc. and AOL merged into
a single media and technology company (hereinafter referred to as
"AOL"). This marriage of "old" and "new" media received extensive
coverage from both the press and the financial industry. CSFB was
among the many financial firms reporting on AOL's business and
-3-
forecasting its outlook for the future. Kiggen and Martin headed
CSFB's AOL coverage beginning the day after the merger and
continuing for about a year, through January 2002, when CSFB ceased
covering AOL (Kiggen retired in January 2002; Martin had left CSFB
a few months earlier). During the coverage period, CSFB published
the results of its research in regular reports. These contained,
in addition to observations about AOL, a buy or sell recommendation
and a price target, which was a prediction of AOL's stock price
twelve months hence. CSFB issued thirty-five such reports during
this period, and each such report recommended buying AOL stock.
CSFB initially targeted AOL's future stock price at $80, but
revised it downwardly to $75 one month later in February 2001, and
then to $45 in September 2001. Nine months later, AOL's stock was
trading at $11 per share.
The shareholders allege that Kiggen and Martin
misrepresented their true opinions in these reports, in order to
maintain a good relationship with AOL. The shareholders' theory is
that AOL had the potential to generate significant investment
banking revenue for CSFB, and Kiggen and Martin overstated AOL's
financial strength in the hopes of winning this future business
(CSFB did in fact assist AOL in managing a bond deal purportedly
generating between $750,000 and $820,000 in fees for CSFB). In a
series of internal emails among AOL team members, Kiggen and Martin
expressed doubts about their projections for AOL, yet decided not
-4-
to lower their estimates for AOL's future performance
notwithstanding these concerns. Moreover, they regularly showed
their projections to AOL and revised them based on AOL's reactions.
Even as advertising revenue, a key factor in AOL's success,
declined throughout the industry, CSFB reports continued to predict
AOL's ability to rise above the general slowdown.
In addition, the shareholders allege two instances in
which CSFB1 received non-public, material information about AOL
that CSFB did not disclose in its coverage of the company. On July
10 and 11, 2001, Anthony Lorenzo, a junior CSFB analyst not
assigned to cover AOL, emailed to Kiggen information about AOL
layoffs. Citing an unnamed source, Lorenzo wrote that AOL
"apparently . . . had some layoffs" that "were medium in terms of
severity and will not be announced publicly." The parties disagree
over the import of this tip. The shareholders claim that the
information pertained to layoffs of "up to 1,000 employees"
subsequently reported in The Wall Street Journal and The Washington
Post on August 13 and 14, 2001. CSFB counters that this unnamed
source (later identified as a low-level employee in AOL's
Interactive Marketing Group) was referring only to a small number
of layoffs that occurred within the Interactive Marketing Group on
July 10, 2001, as reported in The Washington Post the next day.
1
At times, we refer to the defendants, collectively, as
"CSFB".
-5-
Lorenzo's emails also mentioned "that AOL was under
investigation and has suspended some employees for inappropriate
accounting activities -- some deals booked inappropriately inflated
revenue." CSFB did not disclose this information in any of its
reports; it was eventually reported by The Washington Post in a
July 2002 article disclosing that AOL had engaged in
"unconventional" advertising deals that might have inflated
revenue. On July 24, 2002, AOL acknowledged that the SEC was
investigating its accounting practices, but denied any wrongdoing.
B. Procedural History
On the basis of these alleged material misstatements and
omissions -- overstating AOL's financial strength, not disclosing
reports of medium-severity layoffs, and not disclosing reports of
unconventional accounting -- the shareholders brought suit in
December 2005 against Kiggen, Martin, and CSFB under Section 10(b)
of the Exchange Act, 15 U.S.C. § 78j(b), and under SEC Rule 10b-5.
The complaint also alleged that CSFB, Credit Suisse First Boston
(USA) Inc. (CSFB's parent company), and CSFB executives Frank
Quattrone and Elliot Rogers violated Section 20(a) of the Exchange
Act, 15 U.S.C. § 78t(a), by failing to exercise control over their
employees' alleged misstatements and omissions.
In due course, the defendants sought summary judgment.
At the hearing occasioned by that motion, the shareholders and the
defendants each presented expert testimony to show the effect, or
-6-
lack thereof, of CSFB's omissions on AOL stock prices. The
shareholders retained Dr. Hakala, while CSFB employed Dr. René M.
Stulz. Each side subsequently moved to exclude the other's expert
opinion under Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579,
597 (1993) ("[T]he Rules of Evidence--especially Rule 702--[]assign
to the trial judge the task of ensuring that an expert's testimony
both rests on a reliable foundation and is relevant to the task at
hand."). In due course, the court held a Daubert hearing to
determine the admissibility of the proffered expert testimony on
loss causation.
C. Event Studies and Expert Testimony
Loss causation is among the six elements of a private
cause of action for securities fraud; the other five are: a
material misrepresentation or omission, scienter, a connection with
the purchase or sale of a security, reliance, and economic loss.
Dura Pharm., Inc. v. Broudo, 544 U.S. 336, 341-42 (2005). To prove
loss causation, a plaintiff "must show 'a sufficient connection
between [the fraudulent conduct] and the losses suffered . . . .'"
In re Omnicon Grp., Inc. Sec. Litig., 597 F.3d 501, 510 (2d Cir.
2010) (quoting Lattanzio v. Deloitte & Touche LLP, 476 F.3d 147,
157 (2d Cir. 2007)) (alterations in original). In other words, the
stock market must have reacted to the subsequent disclosure of the
misconduct and not to a "tangle of [other] factors." Dura Pharm.,
544 U.S. at 343.
-7-
The usual -- it is fair to say "preferred" -- method of
proving loss causation in a securities fraud case is through an
event study, in which an expert determines the extent to which the
changes in the price of a security result from events such as
disclosure of negative information about a company, and the extent
to which those changes result from other factors.2 First, the
expert selects the period in which the event could have affected
the market price.3 The expert then attempts to determine the
effect on the share price of general market conditions, as opposed
to company-specific events, using a multiple regression analysis,
a statistical means for explaining the relationship between two or
more variables. 1 David L. Faigman et al., Modern Scientific
Evidence; The Law and Science of Expert Testimony 430 (2012).
Thus, for any given day, the expert predicts the company's share
price based on the market trends on that particular day. The
expert then compares this predicted return with the actual return
in the event window in order to determine the probability that an
abnormal return of that magnitude could have occurred by chance.
2
For additional information about event studies in
litigation, see Sanjai Bhagat & Roberta Romano, Event Studies and
the Law: Part I: Technique and Corporate Litigation, 4 Am. L. &
Econ. Rev. 141 (2002), and Michael J. Kaufman & John M. Wunderlich,
Regressing: The Troubling Dispositive Role of Event Studies in
Securities Fraud Litigation, 15 Stan. J. L. Bus. & Fin. 183, 186
(2009).
3
"Stock price," "share price," "market price," "closing
price," and "return" are all used interchangeably throughout this
opinion.
-8-
If this probability is small enough, the expert can reject the
hypothesis that normal market fluctuations, as opposed to company-
specific events, can explain the movement in the share price.
Central to multiple regression analyses are variables,
which, as the term implies, can have two or more possible values.
Id. n.1. Multiple regression includes a variable to be explained
(the dependent variable) and explanatory (or independent) variables
that have the potential to be associated with changes to the
dependent variable. Id. at 430. ("[A] multiple regression analysis
might estimate the effect of the number of years of work on salary.
Salary would be the dependent variable to be explained; years of
experience would be the explanatory variable."). The third type of
variable at issue in this case is a dummy variable, which is also
known as a "binary variable" because it only has two possible
values, such as gender, or, as in this case, the existence or non-
existence of company-specific events.4 By assigning the variable
4
An opinion from the District of New Jersey provides a
succinct example of the use of a dummy variable:
[S]uppose you are investigating [United States]
consumption behavior with time series data for the period
1930 to 1950. You would expect that consumption behavior
would have been significantly different during the years
of World War II than it was before and after the war. To
take this effect into account, you can create an
artificial variable that will take the value 1 during
each of the war years and the value 0 during each of the
other years.
Animal Sci. Prods., Inc. v. China Nat'l Metals & Minerals Imp. &
Exp. Corp., 702 F. Supp. 2d 320, 358 n.44 (D.N.J. 2010).
-9-
a value of zero or one in the mathematical formulae used in the
analysis, the dummy variable becomes mutually exclusive with
respect to any explanatory variables, unable to exist or affect the
outcome simultaneously. Thus, by using a dummy variable, the
projected various outcomes can reveal which explanatory variables
affect the dependent variable.
D. Dr. Hakala's Event Study
Such is the basic structure of an event study. In its
motion to exclude Dr. Hakala's testimony, CSFB alleged that his
methodology included techniques that did not meet the standards of
reliability articulated in Daubert. It challenged four elements of
Dr. Hakala's study.
1. Selection of Event Dates
The first alleged flaw in Dr. Hakala's analysis was his
selection of event dates. CSFB claimed that Dr. Hakala failed to
conform to event study methodology by selecting his event dates
after running his regression analysis. As noted previously, the
first step of an event study is identifying the relevant dates that
are the focus of the study. CSFB argued that Dr. Hakala reversed
the steps in this process, first conducting a regression analysis,
and then, after identifying fifty-seven dates with statistically
significant abnormal returns, using them as the relevant dates for
his event study.
-10-
According to CSFB, this results-driven approach produced
event dates that had "little relationship with the allegations or
facts in this case and ma[de] no sense even under [Dr. Hakala's]
own definition of 'relevance.'" For instance, Dr. Hakala
attributed some abnormal market increases in AOL stock prices to
the defendants on days when CSFB released no reports about AOL --
often when it was no longer reporting about AOL. Dr. Hakala also
characterized several of the dates in his study as corrective,
despite the fact that the complaint had labeled them as
inflationary.5 On one event date, the abnormal market return was
negative, yet Dr. Hakala classified the date as inflationary.
Finally, Dr. Hakala often identified dates as corrective when no
negative information entered the market, and other dates as
inflationary when no positive information entered the market.
2. Overuse of Dummy Variables
CSFB also asserted that Dr. Hakala's use of dummy
variables not only overstated the baseline stability of AOL's stock
prices, but also failed to satisfy the Daubert requirement of
reproducibility. The goal of a regression analysis is to create a
baseline against which the market return on event dates is
measured. Through the use of dummy variables, the event dates
5
An inflationary date occurred when misinformation or
omissions inflated AOL's stock price. A corrective date occurred
when truthful information caused AOL's stock price to return to its
normal levels.
-11-
themselves are excluded from, or "dummied out" of, the regression
analysis to indicate the presence or absence of some event. This
is designed to prevent the event days themselves from distorting
the baseline. Dr. Hakala, in addition to dummying out relevant
event dates, dummied out all dates containing material news about
AOL.6 He chose this approach to control for days when AOL's stock
price might have fluctuated due to the release of information that
was, for purposes of this litigation, irrelevant. He believed that
these material news dates could improperly influence the baseline
regression, and cited other financial economists who endorse this
methodology. Using this approach, Dr. Hakala dummied out 211 out
of 388 days in the study period -- 54% of the total number of days.
CSFB argued that Dr. Hakala's approach went too far, creating an
unrealistically stable baseline and thereby ensuring that all
relevant event dates would appear more unusual than they really
were.
CSFB also attacked Dr. Hakala's dummy selection as
arbitrary. Dr. Hakala performed three event studies relating to
the America Online-Time Warner merger. Although he used the same
6
The terms "relevant event dates" and "material news dates,"
though similar, are distinct. Relevant event dates are the fifty-
seven dates that are the focus of Dr. Hakala's event study.
Material news dates refer to the additional one hundred fifty-four
dates Dr. Hakala dummied out of his event study because they
contained material news.
-12-
criteria to select the material dates each time,7 he dummied out
more material dates in each subsequent study. CSFB argued that his
selection criteria were so vague that two economists would be apt
to pick vastly different numbers of material dates given the same
instructions. Thus, CSFB argued, Dr. Hakala's methodology cannot
be replicated. See Daubert, 509 U.S. at 593 ("Ordinarily, a key
question to be answered in determining whether a theory or
technique is scientific knowledge that will assist the trier of
fact will be whether it can be (and has been) tested."). The proof
of this flaw, according to the defendants, was the fact that even
Dr. Hakala could not select the same number of material news dates
in three separate event studies.
3. Previously Disclosed Information
In Basic Inc. v. Levinson, 485 U.S. 224 (1988), the
Supreme Court held that a plaintiff in a securities fraud suit need
not prove individual reliance on the defendants' fraudulent
statements when purchasing company stock. Id. at 247. Instead,
courts will presume reliance as long as the company's shares trade
in an efficient market, that is, one which incorporates all public
7
Dr. Hakala's criteria came from "the NASDAQ guidelines as
recognized by the SEC." See Self-Regulatory Organizations; Notice
of Filing of Proposed Rule Change by the National Association of
Securities Dealers, Inc. Relating to Issuer Disclosure of Material
Information, 67 F.R. 51,306 (Jul. 31, 2002). He also included
"third party news and reports, and analysts' reports to that list
consistent with the academic studies."
-13-
statements about the company -- including the defendants'
fraudulent statements -- into its share price. Id. "An investor
who buys or sells stock at the price set by the market does so in
reliance on the integrity of that price." Id. Consequently,
investors must also implicitly rely on the integrity of the
information affecting the stock price. Investors who avail
themselves of the fraud-on-the-market theory recognized in Basic,
however, must be consistent. If it is assumed that the market
reacts to the fraud, it must also be assumed that it reacts to the
truth. Accordingly, once a misstatement or corrective disclosure
is publicly known in an efficient market, courts will assume that
the stock price reacts immediately, and any claim that an event
moved the stock price when the event was not actually a new
disclosure will necessarily fail.8
CSFB argued that Dr. Hakala's event study included some
"new disclosures" that were not in fact new to the market. CSFB
pointed to several instances in Dr. Hakala's event study when he
attributed the rise or fall of AOL's stock price to the disclosure
of "stale" information. Consequently, CSFB averred, this
information could not form the basis of a proper event date, and
8
A case pending before the Supreme Court has raised the issue
of the continuing viability of the fraud-on-the-market theory. See
Halliburton Co. v. Erica P. John Fund, Inc., No. 13-317 (U.S.
argued March 5, 2014).
-14-
Dr. Hakala's rejection of the efficient market hypothesis rendered
his study inadmissible.
4. Failure to Control for Confounding Factors
The purpose of an event study, as noted, is to isolate
the impact of an alleged misstatement, omission, or disclosure on
the stock price. A recurring problem in event studies is the
presence of "confounding factors" -- news stories, statements, or
events that coincide with relevant event dates and that themselves
potentially affect the company's stock price. CSFB claimed that
Dr. Hakala made no attempt to control for the many confounding news
stories that emerged at the same time as CSFB reports and other
relevant events, and therefore that his event study did not show
that CSFB's statements, as opposed to some other news story, moved
the stock price on any given day.
E. The District Court's Opinion
In January 2012, the district court issued an order
precluding Dr. Hakala's testimony, relying on the four factors
argued in CSFB's motion to preclude. While it gave specific
examples for each factor, the court explained that these were
illustrative of pervasive problems.
With respect to the event date selection, the district
court determined that
"[r]ather than study the market's reaction to the
misrepresentations alleged in the complaint, Dr. Hakala
cherry-picked unusually volatile days and made them the
focus of his study. If the stock price increased
-15-
sharply, he attributed it to the defendants (even if no
CSFB reports were released on that day). If the stock
price decreased sharply, he called it a corrective
disclosure (even if the news released was positive). The
Court concludes . . . that, quite simply, Dr. Hakala's
theory does not match the facts.
Bricklayers & Trowel Trades Int'l Pension Fund v. Credit Suisse
First Boston, 853 F. Supp. 2d 181, 188 (D. Mass. 2012) (citations
omitted) (internal quotation marks and alterations omitted),
reconsideration denied, published in 853 F. Supp. 2d 181, 195 (D.
Mass. May 17, 2012).
Next, and with little independent analysis, the district
court followed the reasoning of two other district courts in
concluding that Dr. Hakala's use of dummy variables was also
unreliable. See In re Northfield Labs., Inc. Sec. Litig., 267
F.R.D. 536, 548 (N.D. Ill. 2010); In re Xcelera.com Sec. Litig.,
No. 00-11649-RWZ, 2008 WL 7084626, at *1 (D. Mass. Apr. 25, 2008).
In those cases, courts had criticized the high percentage of
dummied-out dates in Dr. Hakala's studies, finding that the
practice artificially stabilized the baseline regression. Here,
the district court noted that Dr. Hakala had dummied out a higher
percentage of days in this study than in either of those cases. It
concluded that "[i]f those courts were correct in excluding his
event studies . . . , as this Court believes they were, it follows
a fortiori that his event study should be excluded here."
Bricklayers, 853 F. Supp. 2d at 189.
-16-
The district court also found that Dr. Hakala's study
"repeatedly ignores the efficient market principle" by attributing
price fluctuations to previously disclosed information. Id. The
shareholders' attempt to "presume an efficient market to prove
reliance and an inefficient market to prove loss causation,"
according to the district court, was tantamount to "hav[ing] their
cake and eat[ing] it too." Id. at 190.
Finally, the district court rejected Dr. Hakala's attempt
to disaggregate confounding information on the event dates. It
acknowledged that AOL received near uninterrupted coverage during
the Class Period, making Dr. Hakala's task difficult. But it
concluded that Dr. Hakala did not use an accepted means for
separating the impact of the relevant event from the impact of
confounding information. As an example of one method that Dr.
Hakala could have used, the district court discussed intra-day
trading analysis. This analysis requires tracking the stock price
throughout the day to see whether its daily highs or lows
correspond with the relevant event, or with the release of some
other information. Dr. Hakala did not do this. Instead, he
"either attributed a rough proportion of the movement to each
report or blamed it all on the defendants." Id. at 191. The
district court considered this approach unreliable and
unscientific.
-17-
The court ultimately concluded that Dr. Hakala's event
study lacked sufficient reliability to be presented to a jury. It
indicated that "[h]ad Dr. Hakala's event study suffered from only
one of the four methodological defects identified by this Court, or
suffered from those flaws jointly but to a lesser degree, today's
ruling might have been different," id. at 191, but, given the
extent of Dr. Hakala's errors, preclusion was necessary.
The district court awarded summary judgment to CSFB sua
sponte, deciding that Dr. Hakala's event study, "even if it were
admitted," did not raise a triable issue of loss causation.9 Id.
at 191-92. The district court's reasoning largely restated the
problems that persuaded it to preclude Dr. Hakala's event study.
After their motion for reconsideration was denied, the
shareholders appealed both the preclusion of Dr. Hakala's event
study and the grant of summary judgment.
II. Analysis
A. Expert Testimony
We review a district court's decision to exclude an
expert witness's testimony for abuse of discretion. Milward v.
Acuity Specialty Prods. Grp., Inc., 639 F.3d 11, 13 (1st Cir.
2011). "This standard is not monolithic: within it, embedded
findings of fact are reviewed for clear error, questions of law are
9
The defendants' original motion for summary judgment had
been denied earlier, subject to being revisited if the court
determined that Dr. Hakala's testimony should be excluded.
-18-
reviewed de novo, and judgment calls are subjected to classic
abuse-of-discretion review." Ungar v. Palestine Liberation Org.,
599 F.3d 79, 83 (1st Cir. 2010).
Since the Supreme Court's decision in Daubert, trial
judges have acted as gatekeepers of expert testimony, assessing it
for reliability before admitting it. See Milward, 639 F.3d at 14.
Expert testimony comes in many different forms, but certain
non-exclusive factors can assist a trial court in its task: "(1)
whether the theory or technique can be and has been tested; (2)
whether the technique has been subject to peer review and
publication; (3) the technique's known or potential rate of error;
and (4) the level of the theory or technique's acceptance within
the relevant discipline." United States v. Mooney, 315 F.3d 54, 62
(1st Cir. 2002) (citing Daubert, 509 U.S. at 593-94). Moreover, an
expert's opinion must be relevant "not only in the sense that all
evidence must be relevant, but also in the incremental sense that
the expert's proposed opinion, if admitted, likely would assist the
trier of fact to understand or determine a fact in issue."
Ruiz-Troche v. Pepsi Cola of P.R. Bottling Co., 161 F.3d 77, 81
(1st Cir. 1998) (citations omitted).
As the district court observed, no single factor is
dispositive in determining the admissibility of Dr. Hakala's expert
testimony. Consequently, we will address the four factors from the
-19-
district court's opinion individually before analyzing the overall
admissibility of Dr. Hakala's testimony
1. Selection of Event Dates
The district court committed no abuse of discretion in
concluding that Dr. Hakala selected event dates based on unreliable
criteria. Event selection should not be difficult to understand,
yet Dr. Hakala's event study leaves us guessing as to how he chose
the fifty-seven dates included in his study. He certainly did not
rely on the shareholders' complaint. Not only did Dr. Hakala
include many dates that bear no relationship to the allegations in
the complaint, in some instances he has turned the complaint on its
head, treating certain events as corrective when the complaint
labeled them inflationary. This complete disconnect between the
event study and the complaint nullifies the usefulness of Dr.
Hakala's work; from all appearances, the event study is more
concerned simply with identifying abnormal market movement than in
supporting the shareholders' causation allegations. Thus, we agree
with the district court's negative assessment of Dr. Hakala's
selection of event dates.
On appeal, the shareholders argue that the district court
could only arrive at this conclusion by rejecting Dr. Hakala's
testimony, and that by so doing it interposed itself as a
fact-finder. It is true that a trial court should not "determine
which of several competing scientific theories has the best
-20-
provenance." Id. at 85. If an expert has reached her conclusion
"in a scientifically sound and methodologically reliable fashion,"
id., the differences "should be tested by the adversarial process,"
Milward, 639 F.3d at 15. Moreover, the court should not rely on
credibility determinations to resolve a disagreement between
experts. See Seahorse Marine Supplies, Inc. v. P.R. Sun Oil Co.,
295 F.3d 68, 81 (1st Cir. 2002) ("The ultimate credibility
determination and the testimony's accorded weight are in the jury's
province.").
Here, Dr. Hakala stated on several occasions that he
pre-selected relevant event dates without reference to the stock
price, yet the district court specifically found, to the contrary,
that Dr. Hakala had "cherry-picked unusually volatile days and made
them the focus of the study." Bricklayers, 853 F. Supp. 2d at 188.
The shareholders claim that the district court impermissibly
discredited Dr. Hakala's testimony on this issue. This argument
misses the point. The problem is not whether Dr. Hakala selected
his event dates with reference to AOL's stock price. The problem
is that the indisputably volatile dates that Dr. Hakala selected
were often unrelated to the shareholders' allegations, and
therefore do not "help the trier of fact to understand the evidence
or to determine a fact in issue." Fed. R. Evid. 702(a). The
district court focused on this deficiency, and not on the mechanics
of how Dr. Hakala selected these event dates. Consequently, we see
-21-
no reason to address whether the district court made an
impermissible credibility determination.
2. Overuse of Dummy Variables
We turn next to the district court's conclusion that Dr.
Hakala overused dummy variables, which, according to the court,
"artificially deflated the baseline volatility of AOL's stock price
during the Class Period." Bricklayers, 853 F. Supp. 2d at 189.
While our review of the record lends some support to the district
court's assessment, there are countervailing factors suggesting
that Dr. Hakala's exclusion of various dates during the Class
Period affects only the weight, and not the admissibility, of his
event study.
CSFB argued, and the district court agreed, that "Dr.
Hakala's event study uses a much higher percentage of dummy
variables than is considered acceptable in the financial
econometric community." Id. at 188. We think, however, that in
arriving at this conclusion, the court may have given insufficient
weight to the shareholders' proffer. The shareholders offered
scholarship, see, e.g., Robert B. Thompson, II, et al., The
Influence of Estimation Period News Events on Standardized Market
Model Prediction, 63 Acc. Rev. 448, 466 (1988) ("[T]he distribution
of security returns during periods in which Wall Street Journal
news is released appears to differ systematically from the
distribution of non-release period returns. This 'news-release'
-22-
effect can be incorporated in models of the returns generating
process by conditioning on news releases."), as well as expert
testimony from Dr. M. Laurentius Marais10 that supported Dr.
Hakala's approach.
CSFB has identified articles that describe event study
methodologies without mentioning the option of controlling for
material news. See, e.g., A. Craig MacKinlay, Event Studies in
Economics and Finance, 35 J. Econ. Literature 13, 17-19 (describing
various market models for event studies without mentioning a news-
conditioned model, but noting that "[t]he use of other models is
dictated by data availability"). The shareholders, meanwhile,
point to academic event studies that do control for material news
dates using a definition of "material news" narrower than Dr.
Hakala's. See, e.g., Richard Roll, R2, 43 J. Fin. 541, 558 (1988)
(selecting material news from the Dow-Jones news service and The
Wall Street Journal).
Ultimately, Dr. Hakala's approach may not be inconsistent
with the methodology or goals of a regression analysis. A
regression analysis seeks to isolate the effect that one variable
has on another. Dr. Hakala's event study sought to isolate the
effect of the general market conditions on AOL's stock price. He
believes that "material news dates" have the potential to distort
10
Dr. Marais submitted testimony rebutting Dr. Stulz's
criticism of Dr. Hakala's use of dummy variables.
-23-
this relationship, and therefore excludes them from his analysis.
Other market economists may disagree with the efficacy of this step
or with the way that he defines materiality, but it is hard to see
how it fails to follow the logic of regression studies. Indeed,
CSFB's event study excludes certain dates for precisely the same
reason. Nor do we consider the percentage of dummied-out dates
dispositive of the issue. The district court was troubled by the
fact that Dr. Hakala excluded 211 of the 388 dates in the study
period. Bricklayers, 853 F. Supp. 2d at 188. That fact alone,
however, does not negate the reliability of his study. The
remaining 177 dates provided enough data to conduct a robust
regression analysis. As Dr. Marais (the shareholders' expert on
the issue of dummy variables) noted, the important factor is not
"the mechanistic and superficial percent of some universe of
observations" that Dr. Hakala dummied out, but the "valid technical
principles concerning the validity of the exercise."
The district court noted two previous court opinions that
disapproved of Dr. Hakala's use of dummy variables. We have held,
however, that "the question of admissibility must be tied to the
facts of a particular case." Milward, 639 F.3d at 14-15 (citations
omitted) (internal quotation marks omitted). The importance of
that counsel is manifest here. Based on the record before us, Dr.
Hakala's event studies in those two cases differed from this one in
at least one key respect: in the other cases he dummied out dates
-24-
on which "any news" about the company appeared. Northfield Labs.,
267 F.R.D. at 548; see also Xcelera, 2008 WL 7084626, at *1. This
is not a frivolous distinction, and the district court in Xcelera
highlighted its importance: "Although the academic literature
supports the use of dummy variables for events in which significant
company-specific news is released, no peer-reviewed journal
supports the view that dummy variables may be used on all dates on
which any company news appears." Xcelera, 2008 WL 7084626, at *1.
No one contends that Dr. Hakala dummied out every day in which AOL
appeared in a news story, yet that was precisely the problem in
Northfield and Xcelera.11 Given that Dr. Hakala employed a
different methodology for this case, Northfield and Xcelera are of
limited value in assessing it.
In Bazemore v. Friday, 478 U.S. 385, 400 (1986) The
Supreme Court observed that, "Normally, failure to include
variables will affect the analysis' probativeness, not its
admissibility." Thus, while Dr. Hakala's use of dummy variables
may, as defendants contend, have artificially deflated the baseline
volatility of AOL's stock in his regression analysis, it may be a
dispute that should be resolved by the jury.
11
CSFB argues that Dr. Hakala employed the same "material
news" standard in his event studies in Northfield and Xcelera as he
did here. That may be true, but it does not address the fact that
the courts in those cases specifically found that Dr. Hakala
excluded any date containing company-specific news. No such
finding exists in this case.
-25-
CSFB launches one more assault on Dr. Hakala's use of
dummy variables. It contends that his methodology fails under
Daubert because it cannot be replicated. Dr. Hakala has performed
three separate event studies related to the AOL merger, and each
time he has dummied out more material news dates than before.
Consequently, CSFB argues, his selection of material news dates is
arbitrary and could not be replicated by another economist. We are
not so sure.
Daubert suggests that a key question in determining
whether a particular technique is scientific knowledge that will be
useful to a jury is "whether it can be (and has been) tested."
Daubert, 509 U.S. at 593. There, the Court was encouraging trial
courts to limit expert testimony to falsifiable theories, meaning
those "capable of empirical test." Id. (quoting Carl Hempel,
Philosophy of Natural Science 49 (1966)). Testing a particular
theory will either reproduce consistent results, thus confirming
the theory, or inconsistent results, thus casting doubt on it. In
this case, Dr. Hakala has theorized that, given certain
assumptions, AOL's stock experienced abnormal returns on fifty-
seven event dates. One would test that theory by repeating his
event study under the same conditions that he did. This would not
be a difficult task, since Dr. Hakala has provided all of the
necessary guidelines to recreate his event study.
-26-
Rather than put Dr. Hakala to the test, CSFB has simply
argued that Dr. Hakala's techniques are unreproducible because of
differences in the number of material news dates that he has
dummied out in successive event studies. But CSFB here is not
comparing apples to apples. Only one of the three studies to which
they refer is at issue here. One of the others was created in
support of class action certification; the other was in connection
with a different lawsuit. That fact alone could be enough to
neuter CSFB’s argument and leave such matters as fodder for cross-
examination, not exclusion.
Ultimately, both the number of dates Dr. Hakala excluded
from consideration and the methods he employed to select those
dates create close questions. And while, as noted, appellant's
arguments raise credible questions, we need not resolve this
particular sub-issue because, as the district court concluded, the
other three bases for excluding Dr. Hakala's testimony are sound.
3. Prior Disclosures
We have described an efficient market for the purpose of
class action securities litigation as "one in which the market
price of the stock fully reflects all publicly available
information." In re PolyMedica Corp. Sec. Litig., 432 F.3d 1, 14
(1st Cir. 2005). We have also explained that the relevant inquiry
is whether the market is informationally efficient, id. at 16,
meaning that "all publicly available information is impounded in
-27-
[the] price" rapidly after it is disseminated. Id. at 14. The
district court correctly applied this standard to Dr. Hakala's
event study. Having established that AOL stocks traded in an
efficient market in order to obtain class certification, the
shareholders could not abandon that factual premise when proving
loss causation. Yet several of the relevant events in Dr. Hakala's
study are based on published references to information previously
disclosed that, under an efficient market theory, would have
already been incorporated into AOL's share price. The lag between
the original disclosure and the event date ranged from one day to
roughly a month. The majority of these disclosures occurred at
least a week before the event dates; thus, the event dates occurred
long after an efficient market would have processed the news.
The shareholders respond that the event dates included
new information that was not contained in the original disclosures.
We conclude, however, that while the disclosures made on the event
dates did not merely parrot previously released information, they
did no more than to provide gloss on public information, and thus
permitted the district court to find that they would not have moved
AOL's share price in an efficient market. See In re Omnicon Grp.,
597 F.3d at 512 (holding that the "negative characterization of
already-public information" does not constitute a corrective
disclosure of new information). For instance, Dr. Hakala included
a February 2001 Lehman Brothers report on AOL. While this report
-28-
downgraded its January 2001 buy or sell recommendation for AOL, it
based this downgrade on information that was known the previous
month. That Lehman Brothers reconsidered its initial appraisal of
AOL’s business, or lost confidence in AOL from one month to the
next, does not demonstrate corrective information entering the
market. See id. ("A negative journalistic characterization of
previously disclosed facts does not constitute a corrective
disclosure of anything but the journalists' opinions."). The
district court did not abuse its discretion in determining that
this recurring problem affected the admissibility of Dr. Hakala's
event study.
4. Confounding Factors
When proving loss causation in a securities fraud suit,
plaintiffs "bear[] the burden of showing that [their] losses were
attributable to the revelation of the fraud and not the myriad
other factors that affect a company's stock price." In re Williams
Sec. Litig., 558 F.3d 1130, 1137 (10th Cir. 2009); Dura, 544 U.S.
at 343 (holding that a plaintiff does not show loss causation if
the lower share price reflects "not the earlier misrepresentation,
but changed economic circumstances, changed investor expectations,
new industry-specific or firm-specific facts, conditions, or other
events, which taken separately or together account for some or all
of that lower price"). Thus, when conducting an event study, an
-29-
expert must address confounding information that entered the market
on the event date.
This case deals with a highly publicized merger that
captured the attention of the entire financial industry. There is
no doubt that Dr. Hakala faced a "herculean task" in sorting
through the continuous flow of information about AOL. Bricklayers,
853 F. Supp. 2d at 190. We agree with the district court, however,
that Dr. Hakala did not establish any reliable means of addressing
this problem. Instead, he seemingly made a judgment call as to
confounding information without any methodological underpinning.
In support of Dr. Hakala's treatment of confounding
factors, the shareholders correctly point out that "even a
statistical event study involves subjective elements." In re Xerox
Corp. Sec. Litig., 746 F. Supp. 2d 402, 412 (D. Conn. 2010)
(citations omitted) (internal quotation marks omitted).
Nevertheless, a subjective analysis without any methodological
constraints does not satisfy the requirements of Daubert. As the
district court noted, "[i]t would be just as scientific to submit
to the jurors evidence of defendants' alleged fraud and AOL's stock
fluctuations and let them speculate whether the former caused the
latter." Bricklayers, 853 F. Supp. 2d at 190; cf. Milward, 639
F.3d at 17-19 (admitting expert testimony based on a subjective
"weight of the evidence" methodology, but identifying the
established steps in this analysis and the factors used in
-30-
analyzing the causal relationship). Dr. Hakala had tools at his
disposal, such as intra-day trading analysis, to guide his analysis
of confounding information.12
5. Bottom Line
Ultimately, we conclude that the district court did not
abuse its discretion in excluding Dr. Hakala's testimony. While we
may question its analysis with respect to dummy variables, the
court's treatment of the remaining three issues is more than
sufficient to satisfy our deferential review. See Ruiz-Troche, 161
F.3d at 83 ("[W]e will reverse a trial court's decision if we
determine that the judge committed a meaningful error in judgment."
(citations omitted) (internal quotation marks omitted)).
Even conceding the aforementioned problems with Dr.
Hakala's event study, however, the shareholders contend that the
event study identified abnormal market movement, on certain key
dates, that did not suffer from any methodological infirmities.
Therefore, they claim, the district court abused its discretion by
throwing out the good with the bad. True enough, some reviewing
courts have found abuses of discretion where trial courts rejected
12
Dr. Hakala could also have used content analysis. See,
e.g., David Tabak, Making Assessments About Materiality Less
Subjective Through the Use of Content Analysis (2007), available at
http://www.nera.com/67_5197.htm; Esther Bruegger & Frederick C.
Dunbar, Estimating Financial Fraud Damages with Response
Coefficients, 35 J. Corp. L. 11, 25 (2009) ("'[C]ontent analysis'
is now part of the tool kit for determining which among a number of
simultaneous news events had effects on the stock price.").
-31-
mostly salvageable expert testimony for narrow flaws. See City of
Tuscaloosa v. Harcros Chems., Inc., 158 F.3d 548, 563 (11th Cir.
1998) (reversing the exclusion of expert testimony in its entirety
where only "a small portion of [the] data and testimony [was]
fundamentally flawed"). Here, however, we confront the reverse
situation -- pervasive problems with Dr. Hakala's event study that,
allegedly, still leave a few dates unaffected.
The district court was not obligated to prune away all of
the problematic events in order to preserve Dr. Hakala's testimony.
Out of fifty-seven event dates, the shareholders list five "key
disclosures" that should survive the district court's order. The
district court did not abuse its discretion in treating the entire
event study as inadmissible given the overwhelming imbalance
between unreliable and reliable dates. The burden of proof falls
on the party introducing expert testimony. Moore v. Ashland Chem.
Inc., 151 F.3d 269, 276 (5th Cir. 1998) ("The proponent need not
prove to the judge that the expert's testimony is correct, but she
must prove by a preponderance of the evidence that the testimony is
reliable."). Requiring judges to sort through all inadmissible
testimony in order to save the remaining portions, however small,
would effectively shift the burden of proof and reward experts who
fill their testimony with as much borderline material as possible.
We decline to overturn the district court's ruling on this specious
logic.
-32-
We also reject the shareholders' argument that CSFB
ambushed them with new arguments at the Daubert hearing.13 CSFB
presented no new arguments at the Daubert hearing. Instead, it
made a thorough presentation of the alleged problems of each event
date. This should not have caught the shareholders off guard. The
Daubert hearing occurred over three years after CSFB first
challenged Dr. Hakala's expert testimony. During those three
years, CSFB reiterated its arguments in expert reports,
depositions, and briefings. It cited numerous examples of specific
dates, but never claimed that those dates constituted the entirety
of Dr. Hakala's flaws. The shareholders knew how CSFB would attack
Dr. Hakala's event study, and they could have anticipated the scope
of the attack.
B. Summary Judgment
Our review of a grant of summary judgment is de novo,
interpreting the record in the light most favorable to the
nonmoving party. See Henry v. United Bank, 686 F.3d 50, 54 (1st
Cir. 2012). "Under [Federal Rule of Civil Procedure 56(a)],
summary judgment is proper if the pleadings, depositions, answers
to interrogatories, and admissions on file, together with the
affidavits, if any, show that there is no genuine issue as to any
material fact and that the moving party is entitled to a judgment
13
The parties argue over which standard of review we should
apply to this issue. We need not answer that question, as the
outcome is the same under any standard of review.
-33-
as a matter of law." Celotex Corp. v. Catrett, 477 U.S. 317, 322
(1986) (internal quotation marks omitted).
Although the district court awarded summary judgment to
CSFB "even if [Dr. Hakala's event study] were admitted,"
Bricklayers at 191-92, we need not engage in such counterfactual
analysis, see Peguero-Moronta v. Santiago, 464 F.3d 29, 34 (1st
Cir. 2006) ("We can affirm [the district court] on any basis
available in the record . . . ."). To sustain this suit, the
shareholders needed to show a connection between CSFB's deceptive
practices and the drop in AOL's stock price. The shareholders
relied solely on Dr. Hakala's event study to satisfy this element.
Without it, they cannot show a genuine dispute as to this issue.
The district court did not need to tunnel into Dr. Hakala's event
study for any evidence favorable to the shareholders' claim. The
district court excluded Dr. Hakala's testimony in its entirety. We
uphold that ruling. Thus, there is no evidence to sort through,
and this complete lack of evidence compels a grant of summary
judgment to CSFB.
III. Conclusion
For the foregoing reasons, we affirm the district court's
exclusion of the shareholders' expert testimony and consequently
affirm its award of summary judgment to CSFB.
It is so ordered.
-34-