Download presentation

Presentation is loading. Please wait.

Published byGertrude Clarke Modified about 1 year ago

1
Portfolio Diversity and Robustness

2
TOC Markowitz Model Diversification Robustness Random returns Random covariance Extensions Conclusion

3
Introduction & Background The classic model S - Covariance matrix (deterministic) r – Return vector (deterministic) Solution via KKT conditions

4
Introduction & Background The efficient frontier

5
Problems and Concerns Number of assets vs. time period Empirical estimate of Covariance matrix is noisy Slight changes in Covariance matrix can significantly change the optimal allocations Sparse solution vectors Without diversity constraints the optimal solution allows for large idiosyncratic exposure

6
Outline Diversity Constraints L1/L2-norms Robust optimization via variation in returns vector Variation in Covariance Estimators via Random Matrix theory Results Further developments

7
Original problem : extension of Markowitz portfolio optimization Diversity Extension

8
Adding The L-2 norm constraint

9

10

11
L-1 norm constraint:

12

13
Robust optimization The classic model Robust: letting r vary i.e. adding infinitely many constraints

14
Robust Model The robust model E is an ellipsoid

15
Robust Model (cont’d) Family of constraints: it can be shown that The new Robust Model:

16
Robust Optimization (cont’d)

17
Robust Optimization Ellipsoids Ellipsoids Fact iff

18
Random Matrix Theory Covariance Matrix is estimated rather than deterministic The Eigenvalue/Eigenvector combinations represent the effect of factors on the variation of the matrix The largest eigenvalue is interpreted as the broad market effect on the estimated Covariance Matrix

19
Random Matrix Implementation compute the covariance and eigenvalues of the empirical covariance matrices Estimate the eigenvalue series for the decomposed historical covariance matrices Calculate the parameters of the eigenvalue distribution Perturb the eigenvalue estimate according to the variability of the estimator

20
Random Matrix Confidence Interval Confidence interval

21
Random Matrix Formulation Problem to solve

22
Markowitz and Robust Portfolio Return is assumed to be random r~N(m,S) Robust portfolio also lies on efficient frontier

23
Efficient Frontier Perturbed Covariance The worst case perturbed Covariance matrix shifts the entire efficient frontier

24
Further Extensions Contribution to variance constraints Multi-Moment Models Extreme Tail Loss (ETL) Shortfall Optimization

25
Contribution to Variance Model

26
QQP Formulation Add artificial :

27
We’d Like To Thank

Similar presentations

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google