Statistical Estimation of Portfolios for Dependent Financial Returns
View this Special IssueResearch Article  Open Access
Hiroyuki Taniai, Takayuki Shiohama, "Statistically Efficient Construction of αRiskMinimizing Portfolio", Advances in Decision Sciences, vol. 2012, Article ID 980294, 17 pages, 2012. https://doi.org/10.1155/2012/980294
Statistically Efficient Construction of αRiskMinimizing Portfolio
Abstract
We propose a semiparametrically efficient estimator for αriskminimizing portfolio weights. Based on the work of Bassett et al. (2004), an αriskminimizing portfolio optimization is formulated as a linear quantile regression problem. The quantile regression method uses a pseudolikelihood based on an asymmetric Laplace reference density, and asymptotic properties such as consistency and asymptotic normality are obtained. We apply the results of Hallin et al. (2008) to the problem of constructing αriskminimizing portfolios using residual signs and ranks and a general reference density. Monte Carlo simulations assess the performance of the proposed method. Empirical applications are also investigated.
1. Introduction
Since the first formation of Markowitz’s meanvariance model, portfolio optimization and construction have been a critical part of asset and fund management. At the same time, portfolio risk assessment has become an essential tool in risk management. Yet there are wellknown shortcomings of variance as a risk measure for the purposes of portfolio optimization; namely, variance is a good risk measure only for elliptical and symmetric return distributions.
The proper mathematical characterization of risk is of central importance in finance. The choice of an adequate risk measure is a complex task that, in principle, involves deep consideration of the attitudes of market players and the structure of markets. Recently, value at risk (VaR) has gained widespread use, in practice as well as in regulation. VaR has been criticized, however, because as a quantile is no reason to be convex, and indeed, it is easy to construct portfolios for which VaR seriously violates convexity. The shortcomings of VaR led to the introduction of coherent risk measures. Artzner et al. [1] and Föllmer and Schied [2] question whether VaR qualifies as such a measure, and both find that VaR is not an adequate measure of risk. Unlike VaR, expected shortfall (or tail VaR), which is defined as the expected portfolio tail return, has been shown to have all necessary characteristics of a coherent risk measure. In this paper, we use risk as a risk measure that satisfies the conditions of coherent risk measure (see [3]). Variants of the risk measure include expected shortfall and tail VaR. The riskminimizing portfolio, introduced as a pessimistic portfolio in Bassett et al. [3], can be formulated as a problem of linear quantile regression.
Since the seminal work by Koenker and Bassett [4], quantile regression (QR) has become more widely used to describe the conditional distribution of a random variable given a set of covariates. One common finding in the extant literature is that the quantile regression estimator has good asymptotic properties under various data dependence structures, and for a wide variety of conditional quantile models and data structures. A comprehensive guide to quantile regression is provided by Koenker [5].
Quantile regression methods use a pseudolikelihood based on an asymmetric Laplace reference density (see [6]). Komunjer [7] introduced a class of “tickexponential’’ distribution, which includes an asymmetric Laplace density as a particular case, and showed that the tickexponential QMLE reduces to the standard quantile regression estimator of Koenker and Bassett [4].
In quantile regression, one must know the conditional error density at zero, and incorrect specification of the conditional error density leads to inefficient estimators. Yet correct specification is difficult, because reliable shape information may be scarce. Zhao [8], Whang [9], and Komunjer and Vuong [10] propose efficiency corrections for the univariate quantile regression model.
This paper describes a semiparametrically efficient estimation of an riskminimizing portfolio in place of an asymmetric Laplace reference density (a standard quantile regression estimator), by using any other quantile zero reference density , based on residual ranks and signs. A consistent and asymptotically normal onestep estimator is proposed. Like all semiparametric estimators in the literature, our method relies on the availability of a consistent firstround estimator, a natural choice being the standard quantile regression estimator. Under correct specifications, they attain the semiparametric efficiency bound associated with .
The remainder of this paper is organized as follows. In Section 2, we introduce the setup and definition of an riskminimizing portfolio and present its equivalent formation under quantile regression settings. Section 3 contains theoretical results for our onestep estimator, and Section 4 describes its computation and performance. Section 5 gives empirical applications, and Section 6 our conclusions.
2. RiskMinimizing Portfolio Formulation
“risk’’ can be considered a coherent measure of risk as discussed in Artzner et al. [1]. The risk of , say , is defined as where and denote the quantile function of a random variable with distribution function . Here, we recall the definition of expected shortfall and the relationship among the tail risk measures in finance. The expected shortfall defined for as can be shown to be a risk measure that satisfies the axioms of a coherent measure of risk. It is worth mentioning that the expected shortfall is closely related but not coincident to the notion of conditional value at risk defined in Uryasev [11] and Pflug [12]. We note that expected shortfall and conditional VaR or tail conditional expectations are identical “extreme’’ risk measures only for continuous distributions, that is, To avoid confusion, in this paper, we use the term “risk measure’’ instead of terms like expected shortfall, CVaR, or tail conditional expectation.
Bassett et al. [3] showed that a portfolio with minimized risk can be constructed via the quantile regression (QR) methods of Koenker and Bassett [4]. QR is based on the fact that a quantile can be characterized as the minimizer of some expected asymmetric absolute loss function, namely, where , is called the check function (see [5]), and is the indicator function defined by if : if . To construct the optimal (i.e., risk minimized) portfolio, the following lemma is needed.
Lemma 2.1 (Theorem 2 of [3]). Let be a realvalued random variable with , then
Then, denotes a portfolio consisting of different assets with allocation weights (subject to ), and the optimization problem under study is, for some prespecified expected return , The sample or empirical analogue of this problem can be expressed as where denotes the th sample value of asset , , with some sufficiently large. The minimizer of (2.7), namely, and , provides the optimal weights yielding the minimal risk. The large sample properties of , especially its consistency, can be implied from the standard arguments and assumptions in the QR context (see [5]).
Let and be the mean vector and the covariance matrix of which are given by respectively. Here, the th element of is where Let . Then the correlation matrix of becomes , and the th elemant of is given by for . The above correlation coefficient can take values close to 1 when is close to 0 with and . Hence, the correlation of the estimated portfolio weights is possibly highly correlated among assets whose sample means differ from , while these problems are ignorable in an asymptotic inference problem if we take .
Thus far, we have seen that the riskminimizing portfolio can be obtained by (2.9), which was the result of Bassett et al. [3]. In what follows, we show that semiparametrically efficient inference of the optimal weights is feasible. The quantity estimated by (2.9) can be regarded as a QR coefficient , defined by where denotes a conditional quantile function, that is, . Note that here the QR model (2.14) has a random coefficient regression (RCR) interpretation of the form with componentwise monotone increasing function and random variables that are uniformly distributed over , that is, (see [5]). Here, a choice such that with the distribution function of some independent and identically distributed (i.i.d.) tuple yields Hence, recalling that the first component of is 1, it follows that, for any fixed , the QR coefficient can be characterized as the parameter of a model such as where the density of is subject to that is, . Let us describe this model as , with , where denotes the distribution of an observation . This model (2.16) is a fixed submodel of (2.14) and is the parametric submodel through which we will achieve semiparametric efficiency.
The model (2.16) is a quantilerestricted linear regression model. But here we have no knowledge about the true density , other than that it belongs to , which allows us to identify . So, we arbitrarily choose from and call it the “reference density" and correspondingly define a “reference model" where the density of is subject to . The goal of the next section is to construct an asymptotically efficient version of based on some feasible , that is, attaining the semiparametric lower bound at correctly specified density that nevertheless remains consistent under a misspecified density ().
3. Semiparametrically Efficient Estimation
The procedure that we will apply here to achieve semiparametric efficiency is based on the invariance principle, as introduced by Hallin and Werker [13]. To this end, first we should have locally asymptotic normality (LAN; see, e.g., van der Vaart [14]) for a parametric submodel , namely, where all the stochastic convergences are taken under . Here, the random vector is called the central sequence, and the positive definite matrix is the information matrix. To ensure the LAN condition for model (2.18), the following assumption is required.
Assumption 3.1. The reference density has finite Fisher information for location:
Assumption 3.2. The regression vectors satisfy, under , for some vector and positive definite , where and are defined by (2.10).
Then, by Theorem 2.1 and Example 4.1 of Drost et al. [15], model (2.18) satisfies the uniform LAN condition for any of the form , with central sequence and information matrix where denotes the residual (i.e., ). Consequently, we have the contiguity , and of course as well. Recall that the contiguity means that for any sequence , if , then also. The reason why we have specified uniform LAN, rather than LAN at single , is the onestep improvement, which will be discussed later.
By following Hallin and Werker [13], a semiparametrically efficient procedure can be obtained by projecting on some field to which the generating group for becomes maximal invariant (see, e.g., Schmetterer [16]). For the quantilerestricted regression model (2.16), such a field is studied by Hallin et al. [6] and found to be generated by signs and ranks of the residuals. Here, let us denote the sign of a residual as , the rank of a residual as , and the field generated by them as Then, “good’’ inference should be based on where is i.i.d. uniform on under and hence approximated by with . In short, we are first rewriting the residual as with realizations of a uniform random variable, and then approximating those as given .
Using this rankbased central sequence, we can construct the onestep estimator (see, e.g., Bickel [17]; Bickel et al. [18]) as follows.
Definition 3.3. For any sequence of estimators , the discretized estimator is defined to be the nearest vertex of .
Definition 3.4. Let be the discretized version of defined at (2.9). We define the (rankbased) onestep estimator of based on reference density as where and are consistent estimates of respectively.
Consistent estimates and can be obtained in the manner of Hallin et al. [19], which is done without the kernel estimation of , though here we omit the details.
Lemma 3.5 (Section 4.1 of [6]). Under with , Therefore, the onestep estimator defined by (3.8) for is semiparametrically efficient at .
In our original notation, the above statement can be rewritten as, for some fixed, .
Recall that the standard QR estimator, defined at (2.9), is asymptotically normal (see Koenker [5]): where Denote the true portfolio weight with respect to risk probability by , where , and its standard quantile regression and our onestep estimators by and , respectively. Denote the block matrix of the covariance matrix of standard quantile and onestep estimators by where submatrices and are symmetric matrices for the covariance of portfolio weights . Then we obtain the variances of the riskminimizing portfolio constructed by the standard quantile, and the onestep estimators are stated in the following proposition. Since direct evaluation gives the following statement, we skip its proof.
Proposition 3.6. The asymptotic conditional variances of an riskminimizing portfolio using the standard quantile regression and onestep estimators given at are, respectively, where .
For any positive definite matrices and , we say if is nonnegative definite. To compare the efficiency of the standard quantile regression estimator and the onestep estimator, we need to show that . To see this, as in Section 3 of Koenker and Zhao [20], let us consider Note that is a nonnegative definite matrix. If is a positive definite, then there exists orthogonal matrix , such that so is nonnegative definite. Hence, is nonnegative definite if is nonsingular. This result assures that the onestep estimator is asymptotically more efficient than the standard quantile regression estimator. From this result, it is easy to see that Also, by taking expectation on both sides, the same inequality holds for unconditional variances.
4. Numerical Studies
In this section, we examine the finite sample properties of the proposed onestep estimator described in Section 3 for the cases where and 0.5. Our simulations are performed with two data generating processes to focus on the underlying true density and how the choice of the reference density might affect the finite sample performances.
The first datagenerating process (DGP1) is the same as that investigated by Bassett et al. [3]. For DGP1, we consider the construction of an minimizing portfolio from four independently distributed assets, that is, asset 1 is normally distributed with mean 0.05 and standard deviation 0.02. Asset 2 is a reversed density with location and scale chosen so that its mean and variance are identical to those of asset 1. Asset 3 is normally distributed with mean 0.09 and standard deviation 0.05. Finally, asset 4 has a density with identical mean and standard deviation to asset 3. DGP2 is a fourdimensional normal distribution with mean vectors the same as those of DGP1, and covariance matrix with and for is . Here, we set , which indicates that the asset returns have correlation 0.5. Notice that both DGP1 and DGP2 have the same mean and variance structures. The underlying true conditional densities of for DGP1 and DGP2 are a mixture of the normal and reversed distribution and the normal distributions, respectively. A simulation of the estimator, for sample, size , and 1000 consists of 1000 replications. We choose prespecified expected return at 0.07.
For each scenario, we computed standard quantile regression estimates with corresponding portfolio weights , and our onestep estimates are defined by (3.8) for various choices of the reference density and actual density in the minimizing portfolio allocation problem.
To make the problem a pure location model, we set the variance of the estimated residual to have one, that is, , where . The true density can be estimated by the kernel estimator for DGP1, where is a kernel function, and is a bandwidth. The first derivative is estimated by As for the DGP2, the actual density becomes normal because the portfolio is constructed by normally distributed returns. We use the normal distribution (), the asymmetric Laplace distribution (AL), the logistic distribution (LGT), and the asymmetric power distribution (APD) with for the reference density .
The density function of the asymmetric power distribution introduced by Komunjer [7] is given by where , and When , the APD pdf is symmetric around zero. In this case, the APD density reduces to the standard generalized power distribution (GPD) [21, pages 194195]. Special cases of the GPD include uniform (), Gaussian , and Laplace () distributions. When , the APD pdf is asymmetric. Special cases include asymmetric Laplace (), the twopiece normal () distributions.
For a given sample size, we compute simulated mean and standard deviation of and and the relative efficiency for .
Table 1 gives the results of the relative efficiencies for DGP 1. When , we see that the efficiency gains of onestep estimators with asymmetric Laplace reference density are large compared with other reference densities with , while these efficiency gains are less when sample size is . When , relative efficiency of assets 3 and 4 with asymmetric Laplace reference density is minimum, while for assets 1 and 2, relative efficiency with normal reference density is minimum. This is because of the covariance structure of defined by (2.10). As can be seen in Section 2, if and , the th element of the correlation matrix defined by (2.13) has a value close to unity. In this case, the asymptotic variance of the usual quantile regression estimator becomes large, which leads to unsatisfactorily large variances in assets 3 and 4. However, the asymptotic variance of our onestep estimator does not have such problems.

Table 2 gives the results of the relative efficiencies for DGP2. In line with efficiency at a correctly specified reference density , we see that the relative efficiency is minimal for all assets and sample sizes with and 0.5. Even though we misspecify the reference density , there exists some sort of efficiency gain except for assets 1 and 2 of the asymmetric Laplace reference density with and . Efficiency gains for the normal reference density and logistic reference density are almost the same because the underlying true density is a symmetric normal distribution, and the asymmetric power reference density with outperforms the asymmetric Laplace reference density.

Figure 1 plots the kernel densities for the estimated portfolio weights for DGP2 with and . We see that the standard quantile regression estimators have long tails on both sides for all assets, whereas onestep estimators have a narrower interval and higher peak at the true weight. This confirms that the onestep estimators are more semiparametrically efficient than the standard ones.
(a)
(b)
(c)
(d)
5. Empirical Application
We apply our methodology to weekly log returns of the 96 stocks of the TOPIX large 100 index. The samples run from January 5, 2007, to December 2, 2011, for a total of 257 observations. The stock prices are adjusted to take into account events such as stock splits on individual securities. Preliminary tests reveal that most log return series have high values of kurtosis and negative values of skewness in general, which indicates that the log returns are nonGaussian.
We computed the optimal portfolio allocations for , and 0.5. We set and , which is the third quartile of the average logreturn distribution. For the firstround estimates, we used the standard quantile regression estimator, and for the onestep estimates, we chose a normal distribution as a reference density. Since we do not have enough information about the shape of the portfolio distributions for the various choices of , the actual density is estimated by the kernel method.
Figure 2 plots the cumulative distribution functions of the riskminimizing portfolios obtained by the standard quantile regression estimates and onestep estimates for , and 0.5. Summary statistics for the distributions of the different portfolios are reported in Table 3.
 
Note: the corresponding summary statistics of the TOPIX log returns for minimum, maximum, mean, and standard deviation are −0.2202, 0.0924, −0.0032, and 0.0328, respectively. Also, quantiles for the TOPIX log returns for , and are −0.0981, −0.0536, −0.0158, −0.0068, and 0.0006, respectively. 
(a)
(b)
(c)
(d)
Figure 2 and Table 3 clearly show that the optimal riskminimizing portfolio manages to reduce the occurrence of events in the left tail when is small for both standard QR estimates and onestep estimates. The standard deviation of the onestep estimates of an minimizing portfolio is smaller than that of the standard QR estimates. We can also observe that the range of a constructed portfolio with onestep estimates is much smaller than that of standard QR estimates, due to the semiparametric efficiency properties of our onestep estimators. When becomes large, the difference in the standard deviation of the constructed portfolio between standard QR estimates and onestep estimates tends to become large. Hence, efficiency gains are large for , which is the mean absolute deviation portfolio. Another interesting finding is that the standard QRconstructed portfolios have highdensity peaks at the required quantiles for all values of , whereas the portfolio constructed by onestep estimates has a quite moderate density reduction at the required quantiles.
Consistent with economic intuition, higher risk aversion is associated with a shorter left tail. In the case where , maximum loss is limited to less than −0.02. This result is particularly striking given that the sample includes the stock market crash of October 2008 due to the US subprime mortgage crisis and the bankruptcy of Lehman Brothers, which resulted in a weekly loss of more than −0.220 for TOPIX. The sample also includes the stock market crash of March 2011 due to the catastrophic earthquake and tsunami that hit Japan, which resulted in a weekly loss of −0.104.
Figure 3 presents empirical efficient frontiers corresponding to the standard quantile regressionbased portfolios and onestep estimates of a portfolio with and 0.5. Figure 3 clearly illustrates that the standard quantile regressionbased portfolio is completely inefficient, far from the onestep frontier.
6. Summary and Conclusions
This paper considered a semiparametrically efficient estimation of an riskminimizing portfolio. A onestep estimator based on residual signs and ranks was proposed, and simulations were performed to compare the finite sample relative efficiencies for the standard quantile regression estimators and the onestep one. These simulations confirmed our theoretical findings. An empirical application to construct a portfolio using 96 Japanese stocks was investigated and confirms that the onestep riskminimizing portfolio has smaller variance that is obtained by the standard quantile regression estimator.
Further research topics include (1) construction of portfolios without shortsale constraints and (2) extending the results to the covariates of time series with heteroskedastic returns. For the former, we impose nonnegativity of the weights by using a penalty function containing a term that diverges to infinity as any of the weights becomes negative (see [22]). For the latter, we refer to Hallin et al. [6] and Taniai [23].
Acknowledgments
This paper was supported by Norinchukin Bank and the Nochu Information System Endowed Chair of Financial Engineering in the Department of Management Science, Tokyo University of Science. The authors thank Professors Masanobu Taniguchi and Marc Hallin and an anonymous referee for their helpful comments.
References
 P. Artzner, F. Delbaen, J.M. Eber, and D. Heath, “Coherent measures of risk,” Mathematical Finance, vol. 9, no. 3, pp. 203–228, 1999. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 H. Föllmer and A. Schied, “Convex measures of risk and trading constraints,” Finance and Stochastics, vol. 6, no. 4, pp. 429–447, 2002. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 G. W. J. Bassett, R. Koenker, and G. Kordas, “Pessimistic portfolio allocation and choquet expected utility,” Journal of Financial Econometrics, vol. 2, no. 4, pp. 477–492, 2004. View at: Google Scholar
 R. Koenker and G. Bassett, Jr., “Regression quantiles,” Econometrica, vol. 46, no. 1, pp. 33–50, 1978. View at: Google Scholar  Zentralblatt MATH
 R. Koenker, Quantile Regression, vol. 38 of Econometric Society Monographs, Cambridge University Press, Cambridge, UK, 2005. View at: Publisher Site
 M. Hallin, C. Vermandele, and B. J. M. Werker, “Semiparametrically efficient inference based on signs and ranks for medianrestricted models,” Journal of the Royal Statistical Society B, vol. 70, no. 2, pp. 389–412, 2008. View at: Publisher Site  Google Scholar
 I. Komunjer, “Asymmetric power distribution: theory and applications to risk measurement,” Journal of Applied Econometrics, vol. 22, no. 5, pp. 891–921, 2007. View at: Publisher Site  Google Scholar
 Q. Zhao, “Asymptotically efficient median regression in the presence of heteroskedasticity of unknown form,” Econometric Theory, vol. 17, no. 4, pp. 765–784, 2001. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 Y.J. Whang, “Smoothed empirical likelihood methods for quantile regression models,” Econometric Theory, vol. 22, no. 2, pp. 173–205, 2006. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 I. Komunjer and Q. Vuong, “Semiparametric efficiency bound in timeseries models for conditional quantiles,” Econometric Theory, vol. 26, no. 2, pp. 383–405, 2010. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 S. Uryasev, “Conditional valueatrisk: optimization algorithms and applications,” in Proceedings of the IEEE/IAFE/INFORMS Conference on Computational Intelligence for Financial Engineering (CIFEr '00), pp. 49–57, 2000. View at: Google Scholar
 G. C. Pflug, “Some remarks on the valueatrisk and the conditional valueatrisk,” in Probabilistic Constrained Optimization, vol. 49 of Nonconvex Optimization and Its Applications, pp. 272–281, Kluwer Academic Publishers, Dodrecht, The Netherlands, 2000. View at: Google Scholar  Zentralblatt MATH
 M. Hallin and B. J. M. Werker, “Semiparametric efficiency, distributionfreeness and invariance,” Bernoulli, vol. 9, no. 1, pp. 137–165, 2003. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 A. W. van der Vaart, Asymptotic Statistics, vol. 3 of Cambridge Series in Statistical and Probabilistic Mathematics, Cambridge University Press, Cambridge, UK, 1998.
 F. C. Drost, C. A. J. Klaassen, and B. J. M. Werker, “Adaptive estimation in timeseries models,” The Annals of Statistics, vol. 25, no. 2, pp. 786–817, 1997. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 L. Schmetterer, Introduction to Mathematical Statistics, Springer, Berlin, Germany, 1974.
 P. J. Bickel, “On adaptive estimation,” The Annals of Statistics, vol. 10, no. 3, pp. 647–671, 1982. View at: Publisher Site  Google Scholar  Zentralblatt MATH
 P. J. Bickel, C. A. J. Klaassen, Y. Ritov, and J. A. Wellner, Efficient and Adaptive Estimation for Semiparametric Models, Johns Hopkins Series in the Mathematical Sciences, Johns Hopkins University Press, Baltimore, Md, USA, 1993.
 M. Hallin, H. Oja, and D. Paindaveine, “Semiparametrically efficient rankbased inference for shape. II. Optimal $R$estimation of shape,” The Annals of Statistics, vol. 34, no. 6, pp. 2757–2789, 2006. View at: Publisher Site  Google Scholar
 R. Koenker and Q. Zhao, “Conditional quantile estimation and inference for ARCH models,” Econometric Theory, vol. 12, no. 5, pp. 793–813, 1996. View at: Publisher Site  Google Scholar
 N. L. Johnson, S. Kotz, and N. Balakrishnan, Continuous Univariate Distributions. Vol. 1, Wiley Series in Probability and Mathematical Statistics: Applied Probability and Statistics, John Wiley & Sons, New York, NY, USA, 2nd edition, 1994.
 S. Leorato, F. Peracchi, and A. V. Tanase, “Asymptotically efficient estimation of the conditional expected shortfall,” Computational Statistics & Data Analysis, vol. 56, no. 4, pp. 768–784, 2012. View at: Google Scholar
 H. Taniai, Inference for the quantiles of ARCH processes, Ph.D. thesis, Université libre de Bruxelles, Brussels, Belgium, 2009.
Copyright
Copyright © 2012 Hiroyuki Taniai and Takayuki Shiohama. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.