Data-driven Multiperiod Robust Mean-Variance Optimization

Question

1. What is the focus of the multiperiod mean-variance optimization model?

2. What is the significance of choosing d as the minimum uncertainty level?

3. How can we transform the formulation of summation (3.13) to the formulation of vector product?

4. What are the conditions for the optimal solution A in the feasible region F(d, a)?

Accepted Answer

The multiperiod mean-variance optimization model focuses on transforming the optimization problem into a non-robust minimization problem with a penalty. This model extends the single-period model to a multiperiod model, allowing for the consideration of the true probability measure within a Wasserstein ball specified by empirical data and a given confidence level. The model aims to provide a tractable framework for robust mean-variance optimization in multiperiod portfolio selection. It addresses the sensitivity of the mean-variance optimization to the empirical mean and covariance of underlying stocks, which can deviate significantly from the true values. By incorporating a penalty, the model offers a more robust approach to portfolio optimization, making it competitive with various strategies in the US stock market over different 10-year intervals between 2002 and 2019.

Accepted Answer

Choosing d as the minimum uncertainty level ensures that the optimal solution of the non-robust multiperiod model under P* is within the plausible estimate of A* with a confidence level of 1-d0. This means that the investor assigns a certain level of confidence to the solution. By defining d as the minimum uncertainty level, we can determine the range of acceptable solutions for the non-robust multiperiod portfolio selection problem. This helps in making informed investment decisions and managing risks effectively. The framework provided by Blanchet et al. (2021) can be used to find the appropriate values of d and a for given data, further enhancing the decision-making process.

Accepted Answer

Next, we try to transform the formulation of summation (3.13) to the formulation of vector product. In this transformation, the p-th component of a vector (v lmn opq) l,m,n,o,p,q is v aci bdt, where t = p T 2 n 3, i = p - (t - 1)n 3 T 2, d = p - (t - 1)n 3 T 2 - (i - 1)n 2 T 2 n 2 T, c = p - (t - 1)n 3 T 2 - (i - 1)n 2 T 2 - (d - 1)n 2 T nT, and b = p - (t - 1)n 3 T 2 - (i - 1)n 2 T 2 - (d - 1)n 2 T - (c - 1)n. The vector A is defined as A = (f i t, g ci dt, h aci bdt), and the vector M is defined as M = (R i t, R c d R i t, R a b R c d R i t).

Accepted Answer

The optimal solution A in the feasible region F(d, a) has two conditions. Firstly, the investment of risky assets in each period is assumed to be n i=1 p i t = 1. This assumption allows for the numerical simulation of p i t. Secondly, the optimal solution A must satisfy the constraint min P U d (Q) [E P (A M )] >= a and max P U d (Q) [E P ((-A) M )] >= a. These constraints ensure that the optimal solution A is robust and sensitive to the underlying and historical probability measures. The optimal solution A can be represented as A i t = (f i t , g ci dt , h aci bdt ) and N i t = (1, R c d , R a b R c d ), where f i t, g ci dt, and h aci bdt are the components of the optimal solution A. The feasible region F(d, a) is defined as A : E Q [A M ] - dA 2 >= a, n i=1 (A i t ) N i t = 1. This region represents the set of all possible optimal solutions that satisfy the given constraints.

Accepted Answer

When implementing a multi-period investment strategy, rebalancing of the portfolio becomes crucial. The strategy involves dividing the portfolio into T number of subportfolios, each following a naive strategy. The target investment strategy for each subportfolio is defined based on the time and the corresponding naive strategy. This approach helps circumvent the problem of not using information from previous periods and reduces transaction costs. The algorithm outlined in the section describes the process of rebalancing the portfolio based on predefined thresholds and ensures that the investments align with the target investment strategy. By waiting for the portfolio to deviate from the target investment strategy by more than 5%, rebalancing is performed to maintain the desired allocation. This method allows for efficient rebalancing while minimizing transaction costs and optimizing the overall performance of the investment.

Accepted Answer

The L2-norm is used in the robust optimization model to measure the robustness of the investment strategy. By considering the L2-norm, the model aims to minimize the worst-case scenario, ensuring that the investment strategy performs well under various market conditions. In the given section, the L2-norm is applied to the first-order Taylor expansion in equation (3.12) to develop the investment strategy. The performance of this strategy is then compared with well-known strategies using empirical data from the S&P 500 companies. The L2-norm helps in evaluating the effectiveness of the investment strategy by analyzing its mean, standard deviation, and Sharpe ratio of wealth returns over a period of 8 years. This comparison is made for different start dates, including periods before and after significant events like the 9/11 attacks and the financial crisis, to assess the strategy's resilience and adaptability to market fluctuations.

Accepted Answer

Our model compares with conventional methods by considering robust single and 2-period models. It evaluates the equal-weighted method, classical Markowitz model, and the maxmin utility of mean-variance (UM) method. The classical Markowitz model is adjusted to account for uncertainty in mean return, using an uncertainty set U. The model also incorporates the assumption of independent and identically distributed returns, with a covariance matrix K u calculated based on sample size d. The dual problem is formulated to maximize utility while considering a confidence interval around historical expected return. In simulations, a sample standard deviation of 1.96 and a risk-aversion coefficient of 4 are used. The comparison highlights the advantages of our model in handling uncertainties and providing robust investment strategies.

Accepted Answer

Table 4 and the subsequent plots illustrate the comparison between investment strategies with and without transaction costs. The rolling 1-year Sharpe ratios of each strategy are plotted, starting from different dates. Table 6 compares strategies with varying periods when there are no transaction costs. The change in Sharpe ratio as the number of periods increases is also depicted. These visualizations help researchers understand the impact of transaction costs on investment strategies and their performance over time.

Accepted Answer

The 2-period model and robust minimum variance methods are compared using the daily adjusted closing price of the 100 largest S&P 500 companies. The comparison involves Blanchet's single-period model, robust normconstrained (NC) model, and sparse and robust mean-variance (SP) model. The NC model controls sparsity or shrinkage of estimated asset weights, while the SP model reduces parameter uncertainty and estimation errors. The comparison is based on the sample covariance matrix, penalty parameters, and rank of absolute weights. The NC model uses a penalty parameter of 4 and risk aversion of 2, while the SP model uses a parameter t of 500 for daily data over 2 years. The comparison aims to evaluate the performance and effectiveness of these models in portfolio optimization.

Accepted Answer

In the simulations, the 2-period model performed competitively compared to equal-weighted, classical Markowitz, UM, NC, SP, and single-period model. The 2-period model showed the highest Sharpe ratio and the lowest standard deviation in four simulations starting from 2002.02.01, 2004.06.01, 2006.06.01, and 2009.06.01. The only exception was the simulation starting from 2008.08.01, where the 2-period model showed a moderately high Sharpe ratio among all strategies and consistently outperformed the equal-weighted, NC, and SP strategies. However, the rolling 1-year Sharpe ratios of the two-period model generally have smaller magnitude from a peak to a trough compared to other strategies. As the number of periods increases, the Sharpe ratio is likely to increase as well, but the computational cost also increases dramatically. Depending on the in-sample data used for parameter estimation, 3-period may perform worse than single- or 2-period model. Our simulations suggest that 2- or 3-period is desirable in terms of portfolio performance for real-world data application.

Data-driven Multiperiod Robust Mean-Variance Optimization

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What is the focus of the multiperiod mean-variance optimization model?

2. What is the significance of choosing d as the minimum uncertainty level?

3. How can we transform the formulation of summation (3.13) to the formulation of vector product?

4. What are the conditions for the optimal solution A in the feasible region F(d, a)?

5. What is the implementation of a multi-period investment strategy and how does it address rebalancing issues?

6. How does the L2-norm affect the robust optimization model?

7. How does our model compare with conventional methods?

8. How do Sharpe ratios compare for strategies with/without transaction costs?

9. How do 2-period model and robust minimum variance methods compare?

10. How does the 2-period model compare to other strategies?

Related Papers (5)

The portfolio selection problems with chance-constrained

A Fuzzy Goal Programming Model for Efficient Portfolio Selection.

Portfolio Optimization on Multivariate Regime-Switching GARCH Model with Normal Tempered Stable Innovation

Novel multiobjective evolutionary algorithm approaches with application in the constrained portfolio optimization

Prediction Portfolio Variance: Firm-Specific and Macroeconomic Factors