33_axarosenberg
TRANSCRIPT
-
7/27/2019 33_AXARosenberg
1/4
Topics, Issues, and Controversies in Corporate Governance and Leadership
S T A N F O R D C L O S E R L O O K S E R I E S
stanford closer look series 1
Risk Management Breakdown atAXA Rosenberg: The Curious Case of a
Quant Manager Trusted Too Much
introduction
Te need or corporate governance systems is driv-
en by problems that can occur when there is a sepa-
ration between the owners o a company and the
managers o the company. Because managers are
agents (and not sole owners themselves), they do
not always have incentive to act in the best inter-
est o shareholders. Instead, they can take actions
to improve their own situation, even when there is
a cost to those actions that is borne by sharehold-
ers. o rectiy this problem (known as the agency
problem), organizations adopt incentive and con-
trol systems that better align the interests o man-
agers and owners and improve the ability o share-
holders to monitor executive behavior.
Each company aces challenges in designing a
governance system that works best or its particularsituation and structure. In the case o public com-
panies, shareholders must overcome the challenges
o diuse ownership (which makes monitoring di-
cult) and the need to work through a board o
directors. In the case o companies with dual class
shares, shareholders must overcome the challenge
o having inerior voting rights relative to insid-
ers. Even in the case o privately held companies,
owners sometimes struggle with issues o separa-
tion and control. Te challenges can be particularly
acute when a company ounder has considerableinuence over the organization and its culture, and
third-party investors have been brought in to share
ownership. In order to be successul, the governance
system must balance deerenceto the expertise and
knowledge o the ounder with objective oversight
that allows the board o directors to intervene with
the ounders decisions when necessary. Lacking
this balance, a companys governance system can
By dv f. l B ty
My 30, 2013
invite problems that signicantly increase the pos
sibility o organizational ailure.
governance at aXa rosenberg
Rosenberg Institutional Equity Management was
private investment management rm ounded b
Barr Rosenberg in 1985. Barr, a ormer nance pro
essor at the University o Caliornia at Berkeley, i
widely renowned or his pioneering work on ris
actors that inuence stock value that is still in us
today in portolio risk assessment and perormanc
attribution. Trough his rm, Barr employed
quantitative strategy based on undamental analysi
to identiy and rank companies that were underval
ued relative to peers. He set an ambitious target o
achieving 2 to 4 percent alpha (or outperormance
relative to benchmarks, and or the most part wasuccessul in meeting this goal over a 15 year pe
riod. With his track record, Rosenberg attracted
investors in the U.S., Europe, and Asia and by th
late 1990s managed $10 billion in assets.
In 1999, French insurance company AXA, look
ing to diversiy and expand its investment manage
ment business, acquired a stake in Barrs investmen
rm. Under the agreement, AXA purchased a 50
percent ownership position and received an option
to buy an additional 25 percent. Barr remained
a signicant investor and chairman o the rmwhich was renamed AXA Rosenberg. Under thei
agreement, AXA had the right to appoint 50 per
cent o the board o directors, with Barr appointin
the remaining 50 percent.1 O note, this structur
was to remain in place in perpetuity, meaning tha
AXAs board representation would not change eve
when AXA exercised its option to increase its own
ership position to 75 percent.
-
7/27/2019 33_AXARosenberg
2/4
stanford closer look series 2
Risk ManageMent BReakdown at aXa RosenBeRg: the CuRious Case of a Quant ManageR tRusted too MuCh
Despite his prominent position as ounder and
chairman, Barr did not retain ormal leadership o
management activities. AXA Rosenberg was led by
a group CEO who was appointed by AXA. Tis in-
dividual ran the executive committee that oversaw
day-to-day operations, including portolio develop-
ment, executing and settling trades, client relations,
marketing, human relations, and legal and regula-
tory compliance work. Although Barr was the de
acto head o the research center (where the quan-
titative trading models were coded and executed),
he ceded the ormal title o director to om Mead,
who represented the centers activities on the ex-
ecutive committee.2 As a result, Barr had no or-
mal management title and no ormal management
responsibilities, despite exerting considerable inu-
ence over the rm.
Although unusual by the standards o a public
corporation, there is some precedent or this type
o governance structure in the asset management
industry. Strategic buyers (such as nancial and
insurance companies) oten have difculty retain-
ing and motivating investor talent ater signing
joint venture agreements with boutique investment
rms. o preserve the environments that allowed
these rms to be successul in the rst place, they
are willing to aord considerable autonomy to the
ounding investors. A typical arrangement is or thestrategic buyer to assume responsibility or the sup-
port unctions o the rm while giving the original
investment team ull discretion to make investment
decisions without intererence, although the seller
must typically abide by a code o ethics and submit
to periodic review by internal compliance ofcers.
In the case o AXA Rosenberg, however, the de-
gree o autonomy was extreme. Barr not only over-
saw all activities in the research center, but AXA was
given little visibility into the details o the research
process. Barr did not report to the group CEO,and the management team o AXA Rosenberg did
not audit the models that the research team devel-
oped. Te group CEO had no authority to hire or
re personnel in the research center, and did not
always interview candidates who applied to work
there. Furthermore, although Barr retained the title
o chairman, he routinely delegated the acilitation
o board meetings to ellow board members that
he had appointed. Barr attended the meetings, bu
oten by telephone rather than in person. Finally
even though Barr was not majority owner and wa
only one member o the board, he had eectiv
control over many organizational decisions. I Bar
opposed a change avored by othersor example
a decision to adjust the prot-sharing programi
was typical that the board and the executive com
mittee would deer to his position. Barr also sat on
a specially created governing board that acted a
a tie-breaker i the AXA Rosenberg board dead
locked on any issue.3
Following the joint venture agreement, the as
sets under management at AXA Rosenberg in
creased signicantly. In 2002, they doubled rom
$10 billion to $20 billion, owing largely to an inu
sion o cash rom AXA and the clients o its wealth
management division. By 2005, assets exceeded
$69 billion and by 2007 reached $135 billion. T
investors in AXA Rosenberg were primarily insti
tutional investors, pension unds (both oreign
and domestic), sovereign wealth unds, and retai
investors. Also during this time, AXA exercised it
option and increased its ownership position to 75
percent.
governance breakdown
Te governance breakdown that led to the unraveling o AXA Rosenberg had its roots in a technolog
migration project spearheaded by Barr and man
aged by research director om Mead. In the mid
1990s, Barr decided to transition the code under
lying the research centers models rom its origina
programming language o FORRAN to an ob
ject-oriented language called Eiel. Te migration
which was intended to take no more than a ew
years, ended up stretching out over a decade and e
ectively became a continuous work in process. Par
o the problem was the way the project was managed; rather than upgrade the model in its totality
all at once, the model was upgraded piecemeal. A
the research center was constantly rewriting cod
to optimize its investment strategy and to keep up
with rapid growth and new product development
the migration was unable to keep pace with th
continuous revisions. Another problem was due to
the choice o programming language. While Eie
-
7/27/2019 33_AXARosenberg
3/4
stanford closer look series 3
Risk ManageMent BReakdown at aXa RosenBeRg: the CuRious Case of a Quant ManageR tRusted too MuCh
is considered a high-quality language, it is also very
obscure. Te PhDs hired into the research center
were oten not experienced working with it and re-
quired a steep learning curve, and the AXA audit
teams could not eectively audit the technology
transer. Finally, because Barr closely guarded access
to the model, a small set o programmers had to
balance model reprogramming with model optimi-
zation. As a result, the system migration was long
and expensive.
Te board o AXA Rosenberg was rustrated
with the pace o progress. Tey were particularly
concerned that the intellectual talent in the research
center was spending too much time on reprogram-
ming and not enough on innovation and under-
standing market dynamics. Te CEO arranged or
a successul technology leader to be seconded to
the research center to support the project manage-
ment o the migration to Eiel; however, he and
the board were unable to enorce accountability or
meeting deadlines and completing the migration.
In 2007, the board agreed that the research mod-
els be adjusted to assume more market risk. Tis
was done in response to client requests to assume
more risk and improve returns, which had been
lackluster in the low volatility market ollowing the
bursting o the technology bubble. Unbeknownst
to the board, as the changes to the risk model wereimplemented, an error was introduced. A program-
mer incorrectly programmed the risk model so that
some o its calculated risk elements were too small
by a actor oten thousand. As a result, the model
sometimes grossly undercalculated the riskiness o
the rms investment decisions. Even though the re-
search center had quality control measures in place
to check the code revisions, these measures did not
detect the error because the model was producing
higher risk on a simulated basis as the board expect-
ed. Te error remained in the code or two yearsbeore it was detected.
Starting in 2007 and through 2008, the volatil-
ity o the market increased. Most unds that traded
on a quantitative basis perormed poorly during
this time. AXA Rosenberg, however, experienced
a considerable deterioration in perormance rela-
tive to other active quantitative managers. Many o
AXA Rosenbergs unds slipped into the ourth and
th quintiles.
In June 2009, a programmer identied the er
ror in the course o updating the risk model. H
notied his boss, om Mead, who in turn inormed
Barr. Te chie investment ofcer, who was respon
sible or implementing a portolio strategy based
on the models outputs, was also notied. However
Barr made the decision not to inorm the group
CEO, the head o compliance, or the board o
directors. Instead, he proposed that his team wai
and correct the error during the next round o cod
updates that were scheduled to take place a ew
months later. Known to him or not, this decision
was in clear violation o SEC regulations which re
quire that an investment company notiy clients i
its policies dier materially rom those disclosed in
its marketing materials.4
Although the error was corrected in Septembe
2009, the CEO was not notied that the error had
existed until November 2009. Shortly thereater, th
board launched an internal investigation to evaluat
the matter. Barr and om Mead were recused rom
this process. In March, the company inormed th
SEC o the error, and the SEC launched a sepa
rate investigation. When clients were inormed o
the error and the SEC investigation in April, man
demanded the return o their capital. By the tim
the investigations were complete nine months laterassets under management had declined to $20 bil
lion.
AXA Rosenberg hired Cornerstone Research to
calculate the economic cost o the error. Corner
stone determined that the coding error had aected
600 client portolios and caused $217 million in
losses (approximately 22 basis points on average
over the two years it was in place. O note, it oun
that 57 percent o client portolios either had no
been aected or had beneted rom the increased
risk taking caused by the error.5
In June 2010, AXA agreed to purchase the re
maining 25 percent o AXA Rosenberg that it did
not own. AXA Rosenberg became a wholly owned
subsidiary o AXA Investment Managers. It con
tinues to exist today and continues to rely on th
same quantitative investment models originally
developed by Barr. However, its product ocus ha
shited to lower risk investment strategies with
-
7/27/2019 33_AXARosenberg
4/4
stanford closer look series 4
Risk ManageMent BReakdown at aXa RosenBeRg: the CuRious Case of a Quant ManageR tRusted too MuCh
lower targeted alphas.
In July 2010, both Barr Rosenberg and om
Mead resigned rom the company. In February
2011, AXA Rosenberg settled charges o mislead-
ing investors with the SEC and agreed to pay over
$240 million, including $217 million to reimburse
clients or investment losses and $25 million in
penalties. In a press release, SEC Director Robert
Khuzami criticized AXA Rosenberg or an organi-
zational structure that did not allow the companys
executive committee to have clear insight into the
operating activities o the research center:
o protect trade secrets, quantitative investment
managers oten isolate their complex computer
models rom the frms compliance and risk man-
agement unctions and leave oversight to a ew
sophisticated programmers. Te secretive struc-ture and lack o oversight o quantitative invest-
ment models, as this case demonstrates, cannot
be used to conceal errors and betray investors.6
In September 2011, the SEC charged Barr Rosen-
berg with securities raud or ailing to disclose a
signicant error in the companys investment mod-
el. Barr agreed to pay $2.5 million in penalties. As
part o the settlement, Barr was banned rom the
securities industry or lie.7
why this Matters
1. Risk management is a undamental duciary
duty o the board o directors (including the
subsidiary board o AXA Rosenberg). How does
a board satisy itsel that risks are known and ap-
propriately monitored within an organization?
2. Te case o AXA Rosenberg involves a widely
renowned investor with extremely specialized
nancial knowledge. Is it really possible or a
board to monitor such an executive and engage
in rigorous risk management?3. Barr Rosenbergs relation with the board and
the executive committee o AXA Rosenberg was
guarded, and he was not particularly orthcom-
ing with inormation. How does an executives
personality aect the implementation o risk
management by the board?
4. Te joint venture agreement between AXA and
Rosenberg involved several curious eatures.
Could the deal structure have been modied to
mitigate the economic impact on AXA share
holders? I so, how? Would this have improved
the outcomes or all parties?
1 Te board was a subsidiary board, responsible or compliance wit
the rms obligations under the Investment Company Act o 1940Te mutual unds managed by AXA Rosenberg were overseen by separate board o trustees comprised o independent directors.
2 Portolio development activities were separate and distinct rom th
activities o the research center. Te research center calculated anranked the relative value o potential stock investments. Tese wer
then conveyed to the chie investment ofcer whose team created
portolio that was diversied in terms o industry, size, geographyetc. From an organizational perspective, the chie investment ofce
was not a member o the research center. Both the chie investmen
ofcer and the director o research were members o the executiv
committee and the board o directors.3 In the Matter o Barr M. Rosenberg, administrative proceedin
number 3-14559, in the U.S. Securities and Exchange Commission4 SEC, Securities Act o 1933; Investment Advisors Act o 1940.
5 During the period in question, AXA Rosenberg managed an averago $100 billion in assets.
6 SEC, SEC Charges AXA Rosenberg Entities or Concealing Erroin Quantitative Investment Model, (Feb. 3, 2011).
7 SEC, SEC Charges Quant Manager with Fraud, (Sep. 22, 2011)
dv l h Mg sy d h c
lhp dvpm rh h s
Gu sh Bu uy mmb
h r c cp Gv s
Uvy. B ty h wh s c
lhp dvpm rh. thy
uh h b Corporate Governance Matters
th uh wu h Mh e. Gum
h h pp h m.
th s c l s h
u h xp p, u, v
p gv hp. th c l
s pubh by h c lhp dv
pm rh h s Gu sh
Bu h r c cp Gv
s Uvy. f m m, v
hp://www.gb..u/.
cpygh 2013 by h B tu h l
s Ju Uvy. a gh v.