i i

i i
Cutting edge investments: Portfolio management

Stress hedging in portfolio construction

dia
Scenario stress testing is a useful and increasingly popular approach to assess portfolio performance under different market
conditions. In this article, Mehmet Bilgili, Maurizio Ferconi and Alex Ulitsky focus on how to directly incorporate stress
scenario information into portfolio construction as an additional constraint to control for potential losses and risks. To
broaden the applicability of stress testing, the authors propose a robust, constrained optimisation approach to handle
uncertainty in scenario parameters. An oil crisis event is used as a numerical illustration

Me
ecent event-driven fluctuations in financial markets have made total or active exposure to the selected stress scenario (see, for example,

R investors increasingly concerned with finding ways to assess

portfolio performance under different economic conditions. A
growing number of financial practitioners and academics have begun to
Cuffe & Goldberg 2012; Ruban et al 2010). In such a case, the resulting
portfolio construction methodology has the following form:

max Œ˛ T .h  hb /  .h  hb /T V .h  hb / (1)

explore alternative approaches to portfolio construction that go beyond h
given all constraints
conventional mean-variance analysis (Laubsch & Ulmer 1999). Scenario-

ive
based portfolio stress testing is one such methodology. It attempts to esti- h > 0; hT L 6 "1 ; .h  hb /T L 6 "2
mate the impact of extreme events and design protection against them
where ˛ denotes asset expected returns; V denotes the risk model; h
(Basel Committee on Banking Supervision 2009; Berkowitz 2000). In
denotes portfolio holdings; hb denotes benchmark holdings; L denotes
applications of stress testing, one starts by identifying a relevant scenario,
stress test loss estimates (defined to have positive values); "1 denotes
eg, an oil crisis or a Standard & Poor’s 500 drop. The next step is modelling
stress test loss threshold (defined to be positive); and "2 denotes active
that event and valuing the expected losses for individual securities and the
cis
stress test loss threshold (defined to be positive).
portfolio as a whole. This framework thus enables us to analyse portfolio
Alternatively, an investor can decide to control both stress profit and
performance subject to a concrete market event, which is making its use
loss (P&L) and stress risk simultaneously, which results in the following
increasingly widespread.
optimisation formulation:
Most commonly, stress testing is used to estimate the impact of adverse
events on investment portfolios. In this article, we focus on how to con- max Œ˛ T .h  hb /  .h  hb /T V .h  hb / (2)
t In

h
trol potential event-driven losses and risks by using a constraint-based given all constraints
framework within the conventional mean-variance portfolio construction h > 0; hT L 6 "1 ; .h  hb /T L 6 "2 ;
process. In addition, we show how to address practical situations where
.h/T VS h 6 ı1 ; .h  hb /T VS .h  hb / 6 ı2
stress events cannot be fully specified, and where only a select range of
possible values can be assigned to model parameters. This additional com- where VS denotes the stress test covariance matrix, ı1 denotes the stress
plexity has not previously been considered. We introduce a robust, con- test risk threshold and ı2 denotes the active stress test risk threshold.
strained optimisation methodology to solve for optimal stress protection
gh

However, the benefit of stress exposure control does not come for free.
in the presence of uncertainty. As a result, optimal stress hedging can now As always, the introduction of additional constraints will come at a cost, so
be applied more broadly to inform portfolio construction decisions. This one should consider using a cost-benefit analysis to assess when hedging
approach can also be extended beyond the stress hedging of investment stress exposure is indeed beneficial. Multiple metrics can be utilised. For
portfolios. Some examples of additional applications are a performance example, one can measure the cost of hedging by the amount of reduction
analysis of a loan portfolio by stress testing default correlations (Saunders
yri

in alpha exposure (see, for example, Ruban et al 2010). Alternatively, the

& Allen 2010) and a scenario analysis of a bank’s investment portfolio in amount of change in investor holdings can be considered as the cost of
the presence of common regulatory guidelines imposed as constraints in hedging. It can also be measured by either turnover or transaction costs.
portfolio construction. However, transaction cost models may not always be available, and the
The organisation of this article is as follows. Below, we describe how to turnover metric, while similar to proportional transaction costs, is not effi-
p

integrate optimal hedging into portfolio construction, both when a stress cient in asset differentiation. For these reasons, in this article we adopt a
test scenario is fully defined and in the presence of uncertainty. Following different metric. It is based on risk-scaled deviation to the current hold-
this, our proposed methodology is illustrated, using an oil crisis scenario ings. In practical terms, this approach will help to assess whether sufficient
Co

as an example. reduction in stress P&L and/or risk can be achieved without making sig-
nificant changes to current investor exposures and holdings. Under this
Portfolio construction with stress exposure control metric, the portfolio construction problem with loss and risk control in a
In portfolio construction, we can accommodate stress test exposure control stress scenario can be written as:
by formulating it as a constrained portfolio optimisation. First, consider
min Œ.h  hp /T V .h  hp / (3)
the case in which a scenario is described by the potential losses each asset h
given all constraints
can exhibit. In order to control aggregated portfolio loss, we extend the
conventional mean-variance framework by adding a linear constraint on h > 0; hT L 6 "; hT VS h 6 ı

risk.net 1

i i

i i
 i i

i i
Cutting edge investments: Portfolio management

where hp denotes current portfolio holdings, " denotes stress test loss while bounds on the covariance matrix can be set by making assumptions
threshold and ı denotes stress test risk threshold. about volatilities and correlations, which will translate into bounds on
The advantage of this measure is that it can be employed when there covariance. For example, when an investor wants to account for positive

dia
is no explicit information on a manager’s alpha views. As a result, this correlations between assets without specifying an explicit value (the bound
approach can also be implemented for non-optimised portfolios. In addi- on that correlation is particularly simple), it is between zero and one.
tion, the methodology described by (3) is quite flexible. It can accom- Using the robust optimisation techniques outlined in Lobo & Boyd
modate controls on alpha exposures and portfolio risk, factor in exposure (2000) and Ghaoui et al (2003), this mini-max problem can be cast into
bounds and incorporate bounds on total and/or active exposures to stress a convex optimisation form, which can be solved using any numerical
test scenarios. package for semi-definite convex programming (see appendix A). We use

Me
However, there is one important limitation to all the stress test hedging CVX in this study (Grant & Boyd 2014).
solutions discussed so far in (1)–(3). These methods can be put to work The following illustrates an application of the proposed portfolio con-
only when all model parameters that describe the scenario are available, struction framework for stress scenario hedging, both in the case of a fully
and this may not always be the case. What if an investor is not certain they defined scenario and in the case of uncertainty.
can accurately estimate asset losses or a risk model describing a stress
test? One simple reason for this may be the rarity of such events. Another Portfolio construction example with oil crisis scenario
may be that the investor tries to avoid replicating history and just assumes To illustrate our proposed portfolio construction methodology for scenario

ive
some of the information about their chosen stress scenario. As a result, exposure hedging, we use an oil crisis stress test as an example.1 This
the only information available may be the ranges of available values, or ‘crisis’ is defined by the realisation of a  12% monthly loss, as projected
perhaps the signs for some parameters, such as correlations. from daily performance in the value of crude oil futures (Nymex). We
The presence of uncertainty in scenario parameters creates a problem consider the 10 most recent years of data; based on historical events, the
that has not been addressed before. In this article, we tackle this com- probability of oil prices plunging more than this scenario is  1%.
plexity by using a robust optimisation approach. Our proposed technique To construct deterministic stress test P&L and a covariance matrix, we
cis
enables one to control the worst-case outcome under stress conditions, and employ a commonly used scenario-weighting approach similar to that
provides a practical portfolio construction solution. described in Silva & Ural (2011) and Ruban et al (2010). This methodol-
Before proceeding with our proposed robust optimisation setup and ogy relies on scenario specification using factor returns; for each historical
solution, let us consider some different types of input uncertainty, previ- observation, the distance to the selected stress scenario is computed based
ously discussed in the context of portfolio construction. Box uncertainty on the difference in risk-adjusted returns. These distances then specify the
assumes the parameter lies within a range around the point estimate for weighting scheme used to estimate the scenario model for the covariance
t In

each component (Tütüncü 2004). Ellipsoid uncertainty sets define a surface matrix, which, in turn, is utilised to compute asset level P&L by propa-
around the point estimate (centre), with axes (parameters) determining the gating scenario shocks to other factors. In this article, we are focused on
size of the set (Goldfarb & Iyengar 2003). Both approaches can be applied illustrating our proposed methodology for optimal hedging, so we select
to handle incomplete information in linear terms, such as stress P&L. A a single control variable – an oil price shock – and do not distinguish, for
more complicated case is how to handle uncertainty in a stress scenario risk example, whether the driver for the scenario can be further classified as a
model. In particular, what should be done when a user can specify only the supply- or demand-driven event (Kilian & Park 2009).
gh

signs for correlations? While seemingly more complex, the last example Also for illustration, let us assume a long-only multi-asset invest-
can be explicitly formulated as a special case of box-type uncertainty (see, ment fund that has established equal-weighted positions in 10 different
for example, Lobo & Boyd 2000). In that paper, uncertainty was present exchange-traded funds. Table A contains strategy correlations under nor-
in the contemporaneous risk model used in portfolio construction. Here, mal market conditions (see the upper triangle) and current holdings. The
we allow for uncertainty in the description of the stress scenario, but we initial portfolio allocation decision is made given the current correlation/
yri

assume the risk model corresponding to current market conditions is fully volatility structure in the market. Alpha expectations and other portfolio
specified. manager conditions result in an ex ante portfolio risk of 10:7% per year.
In this article, we aim to consider a rather general case of input uncer- Table A also provides the results for each strategy in crisis mode. The
tainty. Our assumption is that both the stress test covariance matrix (ie, lower triangle of the correlation matrix contains stressed values.
the covariance matrix associated with the stress scenario) and the stress First, consider a deterministic case in which we only hedge against
p

test loss estimates are not explicitly known. The only available informa- stress P&L. The resulting portfolio construction approach is described by
tion is that they are subject to a box-type uncertainty with lower and upper (3), with only a linear constraint providing the upper bound for possible
bounds (C ; CN ) and (L; L),
N respectively. In that case, portfolio construction scenario total loss. With fully defined stress scenario parameters, we can
Co

with stress exposure hedging can be formulated as a worst-case scenario easily compute portfolio P&L and show that, with current holdings, the
optimisation: fund loss will be 163 basis points if the oil crisis scenario materialises.
h i Figure 1 shows how portfolio risk changes when potential stress test loss
min maxŒ.h  hp /T V .h  hp / C hT VS h (4)
h VS is reduced from 163bp to 80bp. Interestingly, our results indicate that stress
given all constraints
protection results in lower overall portfolio risk. The green line shows that
h > 0; hT L 6 "; C 6 VS 6 CN ; N
L6L6L

where  controls the aversion to uncertain stress scenario risk. In practice, 1 Any of the approaches described by (1)–(3) can be further extended to
bounds on losses can be explicitly selected based on scenario expectations, simultaneously control exposure to multiple stress events.

2 risk.net June 2017

i i

i i
 i i

i i
Cutting edge investments: Portfolio management

A. Inputs for oil crisis stress test scenario

Oil crisis
Allocation Volatility volatility Oil crisis

dia
Strategy ID (%) (%) (%)* P&L (bp) Asset correlations (current and oil crisis)**
Strategy 1 10 16.0 19.5 280 1.00 0.12 0.89 0.84 0.62 0.09 0.90 0.96 0.95 0.82
Strategy 2 10 3.0 4.1 36 0.22 1.00 0.06 0.08 0.11 0.77 0.24 0.18 0.18 0.15
Strategy 3 10 22.1 24.8 329 0.87 0.12 1.00 0.84 0.62 0.10 0.75 0.80 0.76 0.66
Strategy 4 10 5.9 9.5 71 0.55 0.19 0.65 1.00 0.69 0.03 0.71 0.78 0.75 0.68
Strategy 5 10 7.5 6.5 50 0.60 0.04 0.59 0.62 1.00 0.33 0.54 0.60 0.54 0.42
Strategy 6 10 2.3 4.5 33 0.12 0.58 0.07 0.19 0.19 1.00 0.13 0.11 0.08 0.26

Me
Strategy 7 10 16.2 20.4 269 0.93 0.24 0.74 0.49 0.57 0.11 1.00 0.96 0.94 0.82
Strategy 8 10 17.2 20.2 272 0.96 0.26 0.74 0.47 0.55 0.12 0.97 1.00 0.96 0.84
Strategy 9 10 16.6 19.4 229 0.95 0.25 0.73 0.46 0.54 0.11 0.97 0.98 1.00 0.84
Strategy 10 10 16.0 20.7 201 0.84 0.10 0.65 0.51 0.52 0.02 0.89 0.88 0.88 1.00
*Lowest volatility under oil crisis stress test conditions is set to be 90% of stress test risk volatility level, while highest stress test volatility is set to be 110% of stress test risk
volatility for each strategy. **Upper triangle shows the strategy correlations with current risk model, while lower triangle shows the scenario-weighted correlations under oil
crisis conditions. **Lower bounds on correlation uncertainty under oil crisis are calculated by subtracting 10% (50%) of absolute value of correlation oil crisis correlation if oil
crisis scenario correlation is higher (lower) than current correlation for each entry. **Upper bounds on correlation uncertainty under oil crisis are calculated by adding 50%
(10%) of absolute value of correlation oil crisis correlation if oil crisis scenario correlation is higher (lower) than current correlation for each entry

ive
1 The current risk and deviation from initial holdings across 2 The current risk and stress test risk when stress test risk is
various stress test P&L thresholds for a stress test P&L controlled with a constraint in the stress test risk reduction
reduction model model, achieving stress test P&Ls of 150bp and 140bp

12 Current risk (%) with stress test loss = 150bp

–75
cis
Current risk (%) with stress test loss = 140bp
–85 10 Stress test risk (%) with stress test loss = 150bp
Scenario loss (bp)

–95 Stress test risk (%) with stress test loss = 140bp
Current risk (%)

–105 8 12.5
–115 6 12.0
–125 11.5
Portfolio risk (%)

–135 4 11.0
t In

–145 10.5
Stress test loss (bp) 2
–155 10.0
Current risk (%)
–165 0
0 1.0 2.0 3.0 4.0 9.5
Deviation from initial portfolio (%) 9.0
8.5
(a) Portfolio A: 23bp loss reduction at a cost of 92bp deviation from
8.0
initial portfolio. (b) Initial portfolio 0 0.5 1.0 1.5 2.0 2.5
gh

Deviation from initial portfolio (%)

the risk of portfolio A decreases by 23bp if the impact of the oil crisis sce- (a) Portfolio B: 13bp less loss, 87bp less stress test risk. (b) Portfolio C:
23bp less loss, 142bp less stress test risk
nario is hedged to 140bp. The purple line shows that the cost of hedging,
as measured by portfolio tracking error with respect to current holdings,
yri

increases by 92bp to achieve the hedged portfolio A. A portfolio man-

ager needs to balance the cost of portfolio adjustments against the benefit risk by 142bp and oil crisis loss by 40bp, with a 1% deviation from the
of hedging the stress test scenario to select the appropriate reallocation current holdings. Compared with portfolio C, portfolio B achieves a 10bp
solution. greater reduction in ‘oil crisis’, with 55bp less stress test risk, 38bp less
Alternatively, an investor can control simultaneously for stress P&L and current risk and 39bp more deviation from initial holdings at the same
p

stress risk (see (3)). The constraint on stress test risk can be equivalently level of stress test risk aversion.
modelled as a penalty term in the objective function. Optimisations at While actual numbers depend on model details, the overall shape of
different levels of the stress-risk aversion parameter  will result in efficient these efficient frontiers can be explained qualitatively. Indeed, the simul-
Co

frontiers, as shown in figure 2. On this plot, we present risk profiles of taneous improvement in stress P&L and stress risk exhibited by portfolio B
hedged portfolios against the deviation from initial holdings for varying versus portfolio C is in line with the expectation that stress test risk and
levels of stress test risk. stress test loss reduction are positively correlated. The hedging opportu-
The purple and red lines in figure 2 show the stress test risks of different nities increase with a higher loss reduction level (blue line versus green
hedged solutions that achieve 150bp and 140bp losses under the oil crisis line), leading to higher benefits of risk reduction per one unit of deviation
scenario. When we compare these with the green and blue lines, we can from initial holdings. However, there is a limit to risk reduction for a given
see that scenario-weighted stress test risk is 14% more than the current level of stress test loss. The curvature of the blue line shows the stress test
risk on average. Portfolio B shows a hedged solution that reduces oil crisis reduces at diminishing rates with further deviation from initial holdings. In

risk.net 3

i i

i i
 i i

i i
Cutting edge investments: Portfolio management

B. The upper and lower bounds on volatility and correlations for the oil crisis scenario
Lower Upper
volatility volatility

dia
Strategy ID bound (%) bound (%) Asset correlations (upper & lower bound for oil crisis)***
Strategy 1 17.6 23.4 1.00 0.20 0.96 0.61 0.66 0.11 1.00 1.00 1.00 1.00
Strategy 2 3.6 4.9 0.33 1.00 0.11 0.29 0.06 0.64 0.12 0.23 0.23 0.09
Strategy 3 22.3 29.8 0.44 0.18 1.00 0.72 0.65 0.04 0.81 0.81 0.80 0.72
Strategy 4 8.6 11.4 0.28 0.17 0.33 1.00 0.68 0.29 0.54 0.52 0.51 0.56
Strategy 5 5.8 7.8 0.30 0.04 0.30 0.31 1.00 0.29 0.86 0.61 0.81 0.78
Strategy 6 4.0 5.3 0.18 0.29 0.08 0.17 0.17 1.00 0.06 0.11 0.10 0.02
Strategy 7 18.4 24.5 0.84 0.26 0.37 0.25 0.51 0.12 1.00 1.00 1.00 1.00

Me
Strategy 8 18.2 24.2 0.86 0.39 0.37 0.24 0.28 0.18 0.87 1.00 1.00 1.00
Strategy 9 17.4 23.2 0.86 0.38 0.37 0.23 0.49 0.17 0.87 0.88 1.00 1.00
Strategy 10 18.6 24.8 0.76 0.15 0.33 0.26 0.47 0.03 0.80 0.79 0.79 1.00
***Upper triangle shows the upper bound on strategy correlations under oil crisis scenario, while lower triangle shows the lower bound on correlations for oil crisis scenario

practice, computing these efficient hedging frontiers will inform investors

3 How portfolio risk and deviation from initial holdings change
on the desired level of hedging.
when stress test risk trade-off is varied for the stress test

ive
To illustrate our approach for a scenario with uncertainty, we decided to hedging with the robust optimisation model for a fixed stress test
use point estimates for stress P&L, while allowing box-type constraints for loss threshold of 140bp
elements of the stress covariance matrix. Here, we use the following rules
to specify the range for correlations and volatilities, so they encompass Current risk (%)
observed variability across all periods identified by the stress conditions.2 Worst-case risk (%)
Worst-case risk of initial portfolio
The correlation upper bound is defined as the stress-based estimated value, 17 Deviation from initial holdings (%) 3.0

Deviation from initial holdings (%)

cis
increased (decreased) by 50% of the absolute value of stress regime cor- 16
relation if the stress regime correlation is higher (lower) than the normal 15 2.5
regime correlation. The lower bound is defined as the estimated stressed
Portfolio risk (%)

14
2.0
value, reduced (increased) by 10% of the absolute value of the stress regime 13
correlation if the stress regime correlation is higher (lower) than the normal 12 1.5
regime correlation. This asymmetry reflects the expectation that correla- 11
t In

1.0
tions tend to increase during crisis periods. In addition, for each strategy 10
we allowed volatility to be within 90–120% of the oil crisis values. The 9
0.5
resulting risk model (table B) combines uncertainty coming from both 8
sources. 7 0
0.005 0.010 0.050 0.100 0.500 1.000 2.000
The effect of robust optimisation was measured by changing the stress Worst-case risk trade-off (θ)
risk aversion parameter to determine the impact of the worst-case scenario
(a) Initial portfolio. (b) Portfolio D: 23bp less loss, 198bp less worst-case
gh

on portfolio allocations. This is similar to our analysis for a fully defined risk
stress risk model, which allows us to compare both approaches.
In figure 3, we depict efficient stress hedging frontiers with box-type
C. Upper bounds on stress test covariance matrices are not always
uncertainty in a stress test risk model. This figure shows how the current
achieved with the stress test hedging with robust optimisation model for
risk (green line), the worst-case risk (blue line) and the deviation from portfolio D
yri

the initial portfolio (purple line) change with the increasing impact of Strategy 1 Strategy 2 Strategy 4 Strategy 9
stress test risk uncertainty (horizontal axis), while the expected stress test Strategy 1 1 0.20 0.61 1.00
scenario loss is kept constant at 140bp. As expected, worst-case risk values Strategy 2 0.23 1 0.29 0.23
can be significantly higher than the current portfolio risk estimate. In fact, Strategy 4 0.61 0.28 1 0.51
Strategy 9 0.98 0.23 0.51 1
in this example, the worst-case volatility associated with the crisis scenario Red indicates upper bound. Green indicates upper bound achieved. Blue indicates
p

is almost 50% higher than the value based on a regular model. upper bound not achieved
Stress test hedging with robust optimisation provides both a worst-case
scenario covariance matrix and optimal holdings. When the upper bounds the same direction under the oil crisis scenario, and the upper bounds on
Co

on the covariance matrix yield a positive semi-definite matrix, we can asset correlations are not always achieved. Some asset correlations under
simply use the upper bound as the worst-case risk covariance. However, the worst-case scenario for portfolio D are shown in table C. For exam-
when the upper bounds on the covariance matrix are not positive semi- ple, the correlation between strategy 1 and strategy 9 cannot be 1.0 in the
definite, the robust optimisation approach guarantees the worst-case sce- worst-case scenario.
nario risk model is a valid risk model. Asset correlations do not move in To illustrate the efficiency of robust optimisation in reducing worst-case
scenario risk, we compare two portfolios that have the same level of loss
2 The number of observations was not sufficient to model and explore reduction. Portfolio A is based on loss hedging in a deterministic scenario
variability in risk model uncertainty across different stress periods. (see figure 1). Portfolio D achieves the same level of loss protection and is

4 risk.net June 2017

i i

i i
 i i

i i
Cutting edge investments: Portfolio management

a solution for robust optimisation for scenario risk control in the presence After using the techniques outlined in Lobo & Boyd (2000) and Gold-
of uncertainty. We can estimate the worst-case risk for both portfolios; farb & Iyengar (2003), this problem can be cast into a convex robust
as expected, including risk control results in a lower value of worst-case optimisation model as follows:

dia
volatility.
min Œ.h  hp /T V .h  hp / C .hQ;
N VN i  hQ; V i/
N
h;hC ;h ;Q;Q

Summary hT L0 C T .hC C h / 6 "; h D hC  h ; hC > 0; h > 0;

" #
Stress test analysis provides an efficient framework with which to identify QN  Q h
the potential impact of a market event on an investment portfolio. In this Q  0; N
Q  0; M D 0
hT 1

Me
article, we describe how to control that negative impact by constraining the
loss and/or risk associated with the stress scenario in portfolio construc- The robust model includes auxiliary variables (hC , h , Q, Q)
N in order
tion. Furthermore, we propose a novel, robust constrained optimisation to choose a worst-case risk model from the covariance uncertainty set.
methodology that directly addresses the challenges posed by uncertainty Here, hA; Bi represents the trace of the matrix product AB. The trade-off
when modelling stress test scenarios. Using the computational example of between the deviation from the initial holdings and the worst-case stress
an oil crisis, we illustrate the application of our proposed methodologies test risk is controlled by . The inequality constraints () on matrices
and the effectiveness of the resulting solutions. V ; VN ; Q; QN define the positive semi-definite structure of these matrices.
Finally, the stress test hedging framework presented is quite flexible. It To illustrate this approach, we include only the uncertainty in the stress

ive
can be extended to multiple scenarios by simultaneously controlling the scenario risk model described above. The resulting convex semi-definite
different drivers of risks. In addition, the proposed robust optimisation optimisation problem was solved using Grant & Boyd (2014). 
approach can not only handle uncertainty in asset-by-asset risk models
Mehmet Bilgili is a vice president and Maurizio Ferconi and
but also be applied to partially defined factor-based structural risk models. Alex Ulitsky are managing directors and at BlackRock in San
Francisco. The authors report no conflicts of interest. The
cis
Appendix A: robust optimisation approach for stress authors alone are responsible for the content and writing of
this article. The views expressed here are those of the authors
test hedging
alone. This material is not the property of BlackRock and is
We consider the same problem formulation described in (4). The box-
not representative of the views of BlackRock, its officers or
type uncertainty in stress covariances is set by its lower and upper bounds directors. This note is intended to stimulate further research
(V ; VN ), which can be driven by, for example, uncertainty in correlations, and is not a recommendation to trade particular securities or
while stress test loss is represented with the following set:
t In

of any investment strategy. The authors thank Stephen Boyd,

Ronald Ratcliffe,Vivek Gupta and anonymous referees for their
L D fL W L0 C ; ji j 6 i ; i D 1; : : : ; ng valuable input and comments.
Email: Mehmet.Bilgili@blackrock.com,
where L0 shows the point estimate for scenario loss and  is the uncertainty Maurizio.Ferconi@blackrock.com,
around each scenario loss estimate. Alex.Ulitsky@blackrock.com.
gh

REFERENCES
Basel Committee on Banking Ghaoui LE, M Oks and Kilian L and C Park, 2009 Saunders A and L Allen, 2010
Supervision, 2009 F Oustry, 2003 The impact of oil price shocks on Credit Risk Management In and
Principles for sound stress Worst-case value-at-risk and the US stock market Out of the Financial Crisis: New
yri

testing practices and robust portfolio optimization: a International Economic Review Approaches to Value at Risk and
supervision conic programming approach. 50(4), pages 1267–1287 Other Paradigms, volume 528
Report, Bank for International Operations Research 51(4), John Wiley & Sons
Settlements pages 543–556 Laubsch AJ and A Ulmer, 1999
Risk Management: A Practical Silva A and C Ural, 2011
Berkowitz J, 2000 Goldfarb D and G Iyengar, Guide Stress testing of portfolios
p

A coherent framework for stress 2003 RiskMetrics Group Report, November, Barclays
testing Robust portfolio selection Capital Index, Portfolio and Risk
Journal of Risk 2(2), pages 5–15 problems Lobo MS and S Boyd, 2000 Solutions
The worst-case risk of a portfolio
Co

Mathematics of Operations
Cuffe SL and LR Goldberg, Research 28(1), pages 1–38 Unpublished Manuscript, available Tütüncü RH, 2004
2012 at http://bit.ly/2qV7x5c Robust asset allocation
Allocating assets in climates of Grant M and S Boyd, 2014 Annals of Operations Research
extreme risk: a new paradigm for CVX: Matlab software for Ruban OA, D Melas and 132(1-4), pages 157–187
stress testing portfolios disciplined convex programming, MSCI Inc, 2010
Financial Analysts Journal 68(2), version 2.1 Stress testing in the investment
pages 85–108 Available at http://cvxr.com/cvx process
Research Paper 2010-28, August,
MSCI Barra

risk.net 5

i i

i i