Producing the tangency portfolio as a corner portfolio
Summary (2 min read)
1. Introduction
- The aim of classic Mean-Variance (M-V) portfolio optimization, originated from the seminal work of Markowitz [7], is to maximize the expected return of a portfolio and minimize its variance as the measure of risk.
- The optimal CAL has the highest possible slop and is tangent to the efficient frontier of risky assets.
- Niedermayer and Niedermayer [13] provide a Matlab quadratic optimization tool based on Markowitz’s CLA.
- The authors present an algorithm, based on Tütüncü’s results, for which the tangency portfolio can be produced as a corner portfolio.
2. The tangency portfolio
- Rn, where each xi is the weight allocated to the ith asset.
- Since variance is strictly convex, x∗ is unique.
- Let ρmin be the expected return of the Global Minimum-Variance (GMV) portfolio which has the minimum variance between all feasible portfolios.
- See Theorem 1 of [17] and the paragraph following the theorem.
- The assertions of Theorem 2.3 are presented with respect to corresponding Lagrangian multiplier of any optimal portfolio.
3. The modified problem
- The authors introduce the modified problem that corresponds to Problem 1, which contains an additional variable xf so-called free variable.
- (The value of c can be interpreted as the variance of the return of the asset xf which its return is uncorrelated with the returns of the other assets and has the expected value rf ).
- Obviously x satisfies the conditions of Problem 1, if and only if, (x, 0) satisfies the conditions of Problem 2.
- Then the following theorem holds: Theorem 3.2.
- Theorem 3.2 implies that Problems 1 and 2 are equivalent for any ρ ∈ [ρ̂T , ρ̂max], in the sense that both problems have the same optimal portfolio, and consequently have the same efficient frontier (see Fig. 1).
4. The tangency portfolio as a corner portfolio
- In this section the authors discuss some advantages of applying Problem 2 and describe their algorithm.
- But, the authors can take more main steps to determine the exact location of ρT as described in the following.
- Equations are considered as constraints that are always active.
- An efficient portfolio is said to be a corner portfolio if in its vicinity on the efficient frontier, other efficient portfolios have different active sets, see [15].
- The GMV portfolio, however, is included as a corner portfolio (last corner portfolio) irrespective of its active set.
4.1. Description of the algorithm
- Now, the authors may use any relevant PQP algorithm, which produce corner portfolios, to determine the tangency portfolio as a corner portfolio.
- For this, he used the Lagrangian multipliers generated by the CLA and applied the sign of θ(Λ) during the iterations of the method.
- As mentioned before, for example, the algorithm of Niedermayer and Niedermayer [13] can compute the efficient frontier of the standard problem with 2000 assets in less than a second.
- The authors use Optimizer to compute the corner portfolios.
- Moreover, for values of c closed to zero, GMV portfolio acts like a totally riskless portfolio and that segment of the efficient frontier between the tangency portfolio and the GMV portfolio, tends to a straight line.
Did you find this useful? Give us your feedback
Citations
4 citations
1 citations
Cites background from "Producing the tangency portfolio as..."
...As Keykhaei and Jahandideh (2013) pointed out that, in fact all combinations of a risky portfolio and a riskless asset can be represented by a line - Capital Allocation Line (CAL), originating at the riskless asset and passing through the risky portfolio, in the Mean-Standard Deviation (M-SD) plane....
[...]
...As pointed out by Keykhaei and Jahandideh (2013) ‘in order to find the tangency portfolio, it is enough to find the efficient portfolios and recognize the tangency portfolios which maximizes the Sharpe ratio’....
[...]
References
6 citations
"Producing the tangency portfolio as..." refers background or methods or result in this paper
...Tütüncü [17] presented a modification of the CLA which computes the tangency portfolio as a by-product during the algorithm....
[...]
...As in [17], we define function σ : [ρmin, ρmax] → R by σ(ρ) := (x∗′Σx∗)1/2; where (x∗, Λ) ∈ Ω(ρ)....
[...]
...See Theorem 1 of [17] and the paragraph following the theorem....
[...]
...The next theorem states some results derived in [17]....
[...]
...Tütüncü [17] presented an algorithm using CLA to calculate the tangency portfolio....
[...]
4 citations
Additional excerpts
..., the points of nondifferentiability (see [2, 4, 18])....
[...]
4 citations
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the efficient frontier for a tangency portfolio?
for values of c closed to zero, GMV portfolio acts like a totally riskless portfolio and that segment of the efficient frontier between the tangency portfolio and the GMV portfolio, tends to a straight line.
Q3. What is the main aim of the above works?
The major aim of the above works is to introduce methods which calculate the corner portfolios and the efficient frontier while significantly reducing computational time.
Q4. What is the efficient frontier for a riskless asset?
Also when portfolio contains a riskless asset with rf = 0.09 and borrowing is not allowed, the efficient frontier contains x11,x 2 1,x 3 1 and x 0 1 as the corner portfolios, where x01 corresponds to the totally riskless investment, i.e., (0,0.09), on the mean axis.
Q5. What is the tangency portfolio in the modified problem?
Note that, since x∗f = 0 for any ρ ≥ ρT and x∗f > 0 for any ρ < ρT , the tangency portfolio is a corner portfolio in the modified problem.
Q6. What is the optimal CAL for Problem 2?
Since problems 1 and 2 have the same efficient frontiers on [ρ̂T , ρmax] and the efficient frontier of Problem 2 strictly dominates the efficient frontier of Problem 1 on [ρmin, ρ̂T ), the optimal CAL corresponding to Problem 2 which is tangent to the efficient frontier of the problem, is also tangent to the efficient frontier of Problem 1.
Q7. What is the way to calculate the efficient frontier of the standard problem?
their model works directly with the optimal asset weights, instead of the Lagrangian multipliers, which is more clear to express, and also can employ other suitable PQP algorithms.
Q8. What is the efficient frontier of Problem 2?
the efficient frontier of Problem 2 strictly dominates the efficient frontier of Problem 1 when ρ < ρ̂T .Corollary 3.3. ρmax = ρ̂max, where both of the problems 1 and 2 have the same optimal portfolio.
Q9. What is the optimal portfolio of Problem 1?
By the Karush-Kuhn-Tucker (K-K-T) conditions, X∗ = (x∗1, . . . , x∗n)′ is a (primal) solution or optimal portfolio of Problem 1, if and only if, there exist vectors λρ ∈ R, λb ∈ Rm and λd ∈ Rp such that:Σx∗ − λρR̄ − A′λb − C′λd = 0, R̄′x∗ = ρ,
Q10. How fast can the authors compute the efficient frontier of the standard problem?
As mentioned before, for example, the algorithm of Niedermayer and Niedermayer [13] can compute the efficient frontier of the standard problem with 2000 assets in less than a second.
Q11. What is the simplest solution to the problem?
Observe thatλ̂ρ̂T (ρ̂T − rf ) = x̂′Σx̂′ + cx∗f 2 = ρ̂T λ̂ρ̂T + b′λ̂b + d′λ̂d,where the last equation is obtained by adding the two first equations of (3.1) multiplied by x̂′ and x∗f , respectively.
Q12. How can the authors locate the tangency portfolio associated with Problem 1?
the authors can locate the expected return of the tangency portfolio associated with Problem 1, which is also the tangency portfolio associated with Problem 2 (Cor. 3.4), by looking at the value of free variable in the solution of Problem 2.