Some issues on the asset management industry The risk budgeting approach Applications Conclusion Managing Risk Exposures using the Risk Budgeting Approach 1 Thierry Roncalli ? ? Lyxor Asset Management, France & Évry University, France Séminaire de Mathématiques Appliquées, Laboratoire de Mathématiques Jean Leray, Atelier Finance & Risque, Laboratoire LEMNA, Université de Nantes, March 29, 2012 1 The opinions expressed in this presentation are those of the author and are not meant to represent the opinions or official positions of Lyxor Asset Management. Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 1 / 53
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Some issues on the asset management industryThe risk budgeting approach
?Lyxor Asset Management, France & Évry University, France
Séminaire de Mathématiques Appliquées, Laboratoire de Mathématiques Jean Leray,Atelier Finance & Risque, Laboratoire LEMNA,
Université de Nantes, March 29, 2012
1The opinions expressed in this presentation are those of the author and are notmeant to represent the opinions or official positions of Lyxor Asset Management.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 1 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Outline
1 Some issues on the asset management industry
2 The risk budgeting approachDefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
3 ApplicationsSome popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Bond portfolios managementThe risk measureEmpirical results
4 Conclusion
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 2 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some issues on the asset management industry(after the 2008 financial crisis)
Transition in the investment industry
Concentration of assets under managementRisk aversion of large institutional investors becomes higherFunding ratios are smaller (weakness of retirement systems)Pressure for more transparency and robustnessRisk management as important as performance management
Transition in the passive indexation
Robustness of market-cap indexation?Equity indexes = trend-following strategyLack of portfolio construction rules ⇒ Risk concentrationAlternative-weighted indexation = passive indexes where the weightsare not based on market capitalization
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 3 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some issues on the asset management industry(after the 2008 financial crisis)
Emergence of heuristic solutions
Strong criticism of Markowitz optimization2011: success of minimum-variance, erc, mdp/msr and risk paritystrategies ⇒ These portfolio constructions only depend on risks, noton expected returnsSpecial cases of the risk budgeting approach
⇒ Risk budgeting allocation = widely used by market practitioners(multi-asset classes, strategic asset allocation, equity portfolios)
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 4 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
Weight budgeting versus risk budgeting
Let w = (w1, . . . ,wn) be the weights of n assets in the portfolio. LetR (w1, . . . ,wn) be a coherent and convex risk measure. We have:
R (w1, . . . ,wn) =n
∑i=1
wi ·∂ R (w1, . . . ,wn)
∂ wi
=n
∑i=1
RCi (w1, . . . ,wn)
Let b = (b1, . . . ,bn) be a vector of budgets such that bi ≥ 0 and∑
ni=1 bi = 1. We consider two allocation schemes:1 Weight budgeting (WB)
wi = bi
2 Risk budgeting2 (RB)
RCi = bi ·R (w1, . . . ,wn)
2The ERC portfolio is a special case when bi = 1/n.Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 5 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
Application to the volatility risk measure
Let Σ be the covariance matrix of the assets returns. We assume that therisk measure R (w) is the volatility of the portfolio σ (w) =
√w>Σw . We
have:
∂ R (w)
∂ w=
Σw√w>Σw
RCi (w1, . . . ,wn) = wi ·(Σw)i√w>Σw
n
∑i=1
RCi (w1, . . . ,wn) =n
∑i=1
wi ·(Σw)i√w>Σw
= w>Σw√w>Σw
= σ (w)
The risk budgeting portfolio is defined by this system of equations:wi · (Σw)i /bi = wj · (Σw)j /bj
wi ≥ 0∑
ni=1wi = 1
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 6 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
An exampleIllustration
3 assetsVolatilities are respectively 20%,30% and 15%
Correlations are set to 60% betweenthe 1st asset and the 2nd asset and10% between the first two assetsand the 3rd asset
The marginal risk for the first asset is:
limε→0
σ (w1 + ε,w2,w3)−σ (w1,w2,w3)
(w1 + ε)−w1
If ε = 1%, we have:
∂ R (w)
∂ w1' 16.72%−16.54%
1%= 18.01%
Absolute Relative
1 50.00% 17.99% 9.00% 54.40%
2 25.00% 25.17% 6.29% 38.06%
3 25.00% 4.99% 1.25% 7.54%
Volatility 16.54%
Absolute Relative
1 41.62% 16.84% 7.01% 50.00%
2 15.79% 22.19% 3.51% 25.00%
3 42.58% 8.23% 3.51% 25.00%
Volatility 14.02%
Absolute Relative
1 30.41% 15.15% 4.61% 33.33%
2 20.28% 22.73% 4.61% 33.33%
3 49.31% 9.35% 4.61% 33.33%
Volatility 13.82%
Asset WeightMarginal
Risk
Risk Contribution
Weight budgeting (or traditional) approach
Asset WeightMarginal
Risk
Risk Contribution
Risk budgeting approach
ERC approach
Asset WeightMarginal
Risk
Risk Contribution
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 7 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
The ERC portfolio of Maillard et al. (2010)Definition
The Euler decomposition gives us:
σ (w) =n
∑i=1
wi ×∂ σ (w)
∂ wi=
n
∑i=1
RCi
The ERC portfolio is the risk-balanced portfolio such that the riskcontribution is the same for all assets of the portfolio:
RCi = RCj for all i , j
The ERC portfolio is then the solution of the following non-linear system:
w2× (Σw)2 = w1× (Σw)1w3× (Σw)3 = w1× (Σw)1
...wn× (Σw)n = w1× (Σw)1w1 +w2 + · · ·+wn = 1
w1 > 0,w2 > 0, . . . ,wn > 0
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 8 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
The ERC portfolio of Maillard et al. (2010)Properties
Consider the following optimization problem:
w? (c) = argmin√w>Σw
u.c.
∑ni=1 lnwi ≥ c
1>w = 10≤ w ≤ 1
We have w? (−∞) = wmv and w? (−n lnn) = w1/n. The ERC portfoliocorresponds to a particular value of c such that −∞≤ c ≤−n lnn.
1 The solution of the ERC problem exists and is unique.2 We also obtain the following inequality:
σmv ≤ σerc ≤ σ1/n
because if c1 ≤ c2, we have σ (w? (c1))≤ σ (w? (c2)). The ERCportfolio may be viewed as a portfolio “between” the 1/n portfolio andthe minimum-variance portfolio.
3 The ERC portfolio is a form of variance-minimizing portfolio subjectto a constraint of sufficient diversification in terms of weights.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 9 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
If the correlations are the same, the solution is:
wi =σ−1i
∑nj=1 σ
−1j
The weight allocated to each component i is given by the ratio of theinverse of its volatility with the harmonic average of the volatilities.If the volatilities are the same, we have:
wi ∝
(n
∑k=1
wkρik
)−1
The weight of the asset i is proportional to the inverse of the weightedaverage of correlations of component i with other components.In the general case, we obtain:
wi ∝ β−1i
The weight of the asset i is proportional to the inverse of its beta.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 10 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
The ERC portfolio is the tangency portfolio if all the assets have thesame Sharpe ratio and if the correlation is uniform (one-factor model).Let us consider the minimum variance portfolio with a constantcorrelation matrix Cn (ρ). The solution is:
wi =−((n−1)ρ +1)σ
−2i + ρ ∑
nj=1 (σiσj)
−1
∑nk=1
(−((n−1)ρ +1)σ
−2k + ρ ∑
nj=1 (σkσj)
−1)
The lower bound of Cn (ρ) is achieved for ρ =−(n−1)−1 and wehave:
wi =∑
nj=1 (σiσj)
−1
∑nk=1 ∑
nj=1 (σkσj)
−1 =σ−1i
∑nk=1 σ
−1k→ erc
The ERC portfolio minimizes the Gini and Herfindal indexes appliedto the risk measure.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 11 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
Some analytical solutions
The case of uniform correlation3 ρi ,j = ρ
ERC portfolio (bi = 1/n)
wi (ρ) =σ−1i
∑nj=1 σ
−1j
RB portfolio
wi
(− 1
n−1
)=
σ−1i
∑nj=1 σ
−1j
, wi (0) =
√biσ
−1i
∑nj=1√
bjσ−1j
, wi (1) =biσ
−1i
∑nj=1 bjσ
−1j
The general case
wi =biβ
−1i
∑nj=1 bjβ
−1j
where βi is the beta of the asset i with respect to the RB portfolio.
3The solution is noted wi (ρ).Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 12 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
The RB portfolio is a minimum variance (MV) portfoliosubject to a constraint of weight diversification
Let us consider the following minimum variance optimization problem:
w? (c) = argmin√w>Σw
u.c.
∑ni=1 bi lnwi ≥ c
1>w = 1w ≥ 0
if c = c− =−∞, w? (c−) = wmv (no weight diversification)if c = c+ = ∑
ni=1 bi lnbi , w? (c+) = wwb (no variance minimization)
∃c0 : w?(c0)
= wrb (variance minimization and weight diversification)
=⇒ if bi = 1/n, wrb = werc (variance minimization, weight diversificationand perfect risk diversification4)
4The Gini coefficient of the risk measure is then equal to 0.Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 13 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
Existence and uniqueness
If bi > 0, the solution exists and is unique.If bi ≥ 0, there may be several solutions.If ρi ,j ≥ 0, the solution is unique.
An example with 3 assets: σ1 = 20%, σ2 = 10%, σ3 = 5% and ρ1,2 = 50%.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 17 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
DefinitionThe ERC portfolio of Maillard et al. (2010)Main properties of the RB portfolio
Generalization to other convex risk measures
If the risk measure is convex and satisfy the Euler principle, the followingproperties are verified:
1 Existence and uniqueness2 Location between the minimum risk portfolio and the weight
budgeting portfolio3 Optimality
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 18 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Some heuristic portfolios as RB portfolios
The EW, MV, MDP and ERC portfolios could be interpreted as RBportfolios.
EW MV MDP ERCbi
βi wi wiσi 1/nPCi
MV and MDP portfolios are two limit portfolios (explaining that theweights of some assets could be equal to zero).
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 19 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
What is the problem with optimized portfolios?
“The indifference of many investment practitioners tomean-variance optimization technology, despite its theoreticalappeal, is understandable in many cases. The major problemwith MV optimization is its tendency to maximize the effects oferrors in the input assumptions. Unconstrained MV optimizationcan yield results that are inferior to those of simpleequal-weighting schemes” (Michaud, 1989).
Optimal solutions are of the following form:
w?∝ Σ−1
µ
The important quantity is then the information matrix:
I = Σ−1
The eigendecomposition of I is:
Vi (I ) = Vn−i (Σ) and λi (I ) =1
λn−i (Σ)
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 20 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
regularization of the program specification by introducing someconstraints
In the 90’s, there were some hope to save portfolio optimisation.
Today, this hope is tiny and portfolio optimization is perhaps dead.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 22 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Justification of diversified funds
Investor Profiles1 Moderate (medium risk
tolerance)2 Conservative (low risk
tolerance)3 Aggressive (high risk
tolerance)
Fund Profiles1 Defensive (80% bonds
and 20% equities)2 Balanced (50% bonds
and 50% equities)3 Dynamic (20% bonds
and 80% equities)
Relationship with portfolio theory?
Figure: The asset allocation puzzle
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 23 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
What type of diversification offer diversified funds?
Figure: Risk contribution of diversified fundsa
aBacktest with CG WGBI Index and MSCI World
Diversified funds=
Marketing idea?
Deleverage of an equityexposureDiversification in weights6= Risk diversificationNo mapping betweenfund profiles andvolatility profilesNo mapping betweenfund profiles and investorprofiles
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 24 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Risk parity funds
DefinitionA risk parity fund is an ERC startegy on multi-assets classes.
Some examples
AQR Capital ManagementBridgewaterInvescoLyxor Asset ManagementPanAgora Asset ManagementWegelin Asset Management
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 25 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
ERC diversified funds
Figure: Comparison of diversified and risk parity funds5
5Backtest with CG WGBI Index and MSCI World IndexThierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 26 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Time to rethink the bond portfolios management
Two main problems:1 Benchmarks = debt-weighted indexation (the weights are based on
the notional amount of the debt)2 Fund management driven by the search of yield with little
consideration for risk (carry position 6= arbitrage position)
⇒ Time to rethink bond indexes? (Toloui, 2010)
For the application, we consider the euro government bond portfolios. Thebenchmark is the Citigroup EGBI index6.
6This index is very close to the EuroMTS index.Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 27 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Bond indexation schemes
Debt weighting
It is defined bya:
wi =DEBTi
∑ni=1 DEBTi
aTwo forms of debt-weighting areconsidered : DEBT (with the 11countries) and DEBT∗ (without Greeceafter July 2010). This last onecorresponds to the weighting scheme ofthe EGBI index.
Alternative weighting1 Fundamental indexation
The GDP-weighting is defined by:
wi =GDPi
∑ni=1 GDPi
2 Risk-based indexationThe DEBT-RB and GDP-RBweightings are defined by:
bi =DEBTi
∑ni=1 DEBTi
bi =GDPi
∑ni=1 GDPi
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 28 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Choosing the right measure of credit risk
Volatility of price returns 6= a good measure of credit riskCorrelation of price returns 6= a good measure of contagionA better measure is the yield spread, but its computation can bedifficult as it needs to first define a reference risk-free rate.
⇒ One of the best measure is the CDS spread (it does not depend on thecurrency, the yield curve or the duration).
The SABR CDS model
Let Si (t) be the spread of the i th issuer. We have:
dSi (t) = σSi ·Si (t)βi ·dWi (t)
Moreover, we assume that the correlation between the brownian motionsWi (t) and Wj (t) is Γi ,j .
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 29 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Calibration of the βi parameter
We assume that we observe spreads at some given known dates t0, . . . , tn.Let Si ,j be the observed spread for the i th country at date tj . Thelog-likelihood function for the i th country is:
` = −n2ln2π−n lnσ
Si −
12
n
∑j=1
ln(tj − tj−1)−
βi
n
∑j=1
lnSi ,j−1−12
n
∑j=1
(Si ,j −Si ,j−1)2(σS
i Sβii ,j−1
)2(tj − tj−1)
Figure: Results for the period January 2008-August 2011
ESCountry AT BE FI FR DESpread Volatility GR IE IT NL PT
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 31 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Evolution of the correlation matrix
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 32 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Computing the credit risk measure of a bond portfolio
Let w = (w1, . . . ,wn) be the weights of bonds in the portfolio. The riskmeasure is7:
R (x) =√w>Σw =
√∑
ni=1 ∑
nj=1wiwjΣi ,j
with Σi ,j = Γi ,j ·σBi ·σB
j and σBi = Di ·σS
i ·Si (t), where Di is the durationof the bond i , σS
i is the CDS volatility of the corresponding issuer, Si (t) isthe CDS level and Γi ,j is the correlation between the CDS relativevariations of issuers corresponding to the bonds i and j .
R (w) is the volatility of the CDS basket which would perfectly hedge thecredit risk of the bond portfolio.
R (w) depends on 3 “CDS” parameters Si (t), σSi and Γi ,j and two
“portfolio” parameters wi and Di .
7We have d lnBt (Di ) =−Di ·dR (t)−Di ·dSi (t) with Bt (Di ) the zero-coupon ofmaturity Di and R (t) the “risk-free” interest rate. It comes that σB
i = Di ·σSi ·Si (t).
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 33 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Defining the risk contribution
Our credit risk measure R (w) =√w>Σw is a convex risk measure. It
means that:
R (w1, . . . ,wn) =n
∑i=1
wi ·∂ R (w1, . . . ,wm)
∂ wi
=n
∑i=1
RCi
We can then break the risk measure down into n individual sources of risk.
The risk contribution RCi is an increasing function of the parameters Di ,Si (t) and σS
i .
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 34 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
⇒ Small changes in weights but large changes in risk contributions.⇒ The sovereign credit risk measure has highly increased (the largest value12.5% is obtained in November 25th 2011).⇒ If we think that the EGBI portfolio is optimal, we expect that 60% of theperformance will come from Italy and France.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 35 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Some results for the EGBI indexEvolution of the risk contributions
⇒ RB indexation is very different from WB indexation, in terms ofweights, RC and credit risk measures.⇒ The dynamics of the GDP-RB is relatively smooth (monthly turnover' 7%, max = 20%, min = 1.8%).
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 39 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 43 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Comparison with active management
Database: MorningstarCategory: Bond EURO Government216 funds8
The Academic Rule9:
Average Performance of Active Management=
Performance of the Index − Management Fees
⇒ Implied fees for Bond EURO Government: 61 bps / year10
8We don’t take into account the survivorship bias.9There is a large literature on this subject, see e.g. Blake et al. (1993).
10This figure was only 36 bps / year for the period 01/2008 - 08/2011.Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 44 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Some popular RB portfoliosRB portfolios vs optimized portfoliosRisk parity fundsManaging Sovereign credit risk in bond portfolios
Comparison with active management
#(funds > GDP-RB) = 6
Perf. of GDP-RB Index†
=Perf. of Top 10%+ 58 bps / year
† Transaction costs = 15 bps
/ year
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 45 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Conclusion
Risk-budgeting approach = a better approach than portfoliooptimizationRisk-budgeting approach = new theoretical results (Bruder andRoncalli, 2012)
The risk-budgeting approach could be applied to:Risk-balanced allocation
Risk parity fundsStrategic asset allocation
Risk-based indexationEquity indexes (see for example Lyxor SmartIX ERC Index Series11)Bond indexes
11The web site is www.ftse.com/Indices.Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 46 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
For Further Reading
B. Bruder, T. Roncalli.Managing Risk Exposures using the Risk Budgeting Approach.SSRN, www.ssrn.com/abstract=2009778, January 2012.
B. Bruder, P. Hereil, T. Roncalli.Managing Sovereign Credit Risk in Bond Portfolios.SSRN, www.ssrn.com/abstract=1957050, October 2011.
B. Bruder, P. Hereil, T. Roncalli.Managing Sovereign Credit Risk.Journal of Indexes Europe, 1(4), November 2011.
S. Maillard, T. Roncalli, J. Teiletche.The Properties of Equally Weighted Risk Contribution Portfolios.Journal of Portfolio Management, 36(4), Summer 2010.
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 47 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Weights constraints and Portfolio TheoryMain result
We consider a universe of n assets. We denote by µ the vector of theirexpected returns and by Σ the corresponding covariance matrix. Wespecify the optimization problem as follows:
min12w>Σw
u.c.
1>w = 1µ>w ≥ µ?
w ∈ Rn⋂C
where w is the vector of weights in the portfolio and C is the set ofweights constraints. We define:
the unconstrained portfolio w? or w? (µ,Σ):
C = Rn
the constrained portfolio w̃ :
C(w−,w+
)={w ∈ Rn : w−i ≤ wi ≤ w+
i
}Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 48 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Weights constraints and Portfolio TheoryMain result
Theorem
Jagannathan and Ma (2003) show that the constrained portfolio is thesolution of the unconstrained problem:
w̃ = w?(
µ̃, Σ̃)
with: {µ̃ = µ
Σ̃ = Σ + (λ +−λ−)1>+1(λ +−λ−)>
where λ− and λ + are the Lagrange coefficients vectors associated to thelower and upper bounds.
⇒ Introducing weights constraints is equivalent to introduce some relativeviews (similar to the Black-Litterman approach).
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 49 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Proof for the global minimum-variance portfolio
We define the Lagrange function as f (w ;λ0) = 12w>Σw −λ0
(1>w −1
)with λ0 ≥ 0. The first order conditions are Σw −λ01 = 0 and1>w −1 = 0. We deduce that the optimal solution is:
w? = λ?0 Σ−11 =
11>Σ1
Σ−11
With weights constraints C (w−,w+), we have:
f(w ;λ0,λ
−,λ +)
=12w>Σw−λ0
(1>w −1
)−λ
−> (w −w−)−λ
+> (w+−w)
with λ0 ≥ 0, λ−i ≥ 0 and λ
+i ≥ 0. In this case, the first-order conditions
becomes Σw −λ01−λ−+ λ + = 0 and 1>w −1 = 0. We have:
Σ̃w̃ =(
Σ +(λ
+−λ−)1>+1
(λ
+−λ−)>) w̃ =
(2λ̃0− w̃>Σw̃
)1
Because Σ̃w̃ is a constant vector, it proves that w̃ is the solution of theunconstrained optimisation problem with λ ?
0 =(2λ̃0− w̃>Σw̃
).
Thierry Roncalli Managing Risk Exposures using the Risk Budgeting Approach 50 / 53
Some issues on the asset management industryThe risk budgeting approach
ApplicationsConclusion
Weights constraints and Portfolio TheoryExamples
Table: Specification of the covariance matrix Σ (in %)