Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC24

4 - Adjusted Portfolio Selection Model Reflecting the End of the Year Effect of Global Economic Growth Jihye Yang, PhD Student, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul, 03722, Korea, Republic of, Hongseon Kim, Seongmoon Kim This study suggests improved investment strategy based on adjusted portfolio selection model reflecting The End of the Year Effect, a strong negative relation between end of the year global economic growth and average stock returns(Stig V.2016). We can adjust the proportion of capital invested in risk assets and risk- free assets on the rebalancing date according to end of the year global economic growth. Specifically, if end of the year global economic growth is high, we decrease the proportion of capital invested in risk assets and increase risk-free assets in order to avoid risk. Thus, we can expect superior performance using the proposed investment strategy reflecting The End of the Year Effect. 5 - Optimal Control of an Individual Savings Account in a Defined Contribution Pension System Luis Chavez-Bedoya, Professor, Universidad Esan, Nestor Bermudez 191, Chorrillos, Lima, LIMA 9, Peru, Ranu Castaneda In this article, we examine the optimal investment strategy of a particular utility- maximizer participant in a defined contribution pension fund under the system of individual accounts; but, we assume deterministic functions for the fees paid to the pension fund administrator and for the contribution rate of the participant. Finally, we perform a thorough comparison between the two different types of fees (on balance and on flow) and we apply our results to the Peruvian Private Pension System in order to determine equivalent fees on balance and flow. 6 - A Dynamic Mean Variance Analysis with Application to Robo Advising Min Dai, National University of Singapore, 10 Lower Kent Ridge Road, Dept of Math, Singapore, 119076, Singapore, Hanqing Jin, Steven Kou, Yuhong Xu In asset allocation for robo-advising, it is desirable to elicit investors’ risk profile via several simple online questions and to provide advice consistent with conventional investment wisdom (e.g. rich people should invest more money in risky assets, and for long-term investment people should not short sell major stock indices whose returns are higher than the risk-free rate). We propose a dynamic portfolio choice model with the mean-variance criterion for portfolio log-returns that meets the two challenges. The model yields analytical and time- consistent optimal portfolio policies. n WC23 North Bldg 131A Optimal Stopping and Stochastic Control for Financial and Engineering Applications Sponsored: Finance Sponsored Session Chair: Philip Ernst, Rice University, Houston, TX, 8, United States 1 - When is it Best to Follow the Leader We study a classical continuous-time optimal scanning problem with N boxes. An object is hidden in one box according to a given prior distribution. The goal is to determine which box contains the object. When we search a box, we get a signal evolving as a Brownian motion with a constant drift if the box contains the object, or as a Brownian motion with drift zero otherwise. This problem was first studied by Posner and Rumsey (1966), who conjectured that the optimal policy is to always observe the box with the largest posterior. However, we prove that this is not true in general and give counterexamples for some specific prior distributions. It remains unclear whether this policy is optimal for the uniform prior. 2 - The Value of Foresight Philip Ernst, 6100 Main St, Houston, TX, 77005, United States, L.C.G. Rogers, Quan Zhou Suppose you have one unit of stock, currently worth 1, which you must sell before time $T$. The Optional Sampling Theorem tells us that whatever stopping time we choose to sell, the expected discounted value we get when we sell will be 1. Suppose however that we are able to see $a$ units of time into the future, and base our stopping rule on that; we should be able to do better than expected value 1. But how much better can we do? And how would we exploit the additional information? The optimal solution to this problem will never be found, but in this paper we establish remarkably close bounds on the value of the problem, and we derive a fairly simple exercise rule that manages to extract most of the value of foresight. Quan Zhou, Rice University, 6100 Main St. Duncan Hall 2076, Houston, TX, 77005, United States, Philip Ernst, L. C. G. Rogers

3 - Optimal Allocation of Retirement Portfolios Stan Uryasev, University of Florida-Gainesville, AOD, 303 Weilhall, Gainesville, FL, 32611, United States, Giorgi Pertaia, Morton Lane, Matthew Murphy Without a pension the function of the savings is to provide post-employment income to a retiree. At the same time, most retirees will want to leave an estate to their heirs. Guaranteed income can be acquired by investing in an annuity. However, that decision necessarily takes funds away from investment alternatives that might grow the estate. The decision is made even more complicated because one does not know how long one will live. This paper presents a stochastic programming model of an optimal portfolio for the retiree. 4 - Non-linear Risk Parity Portfolios Nathan Lassance, Universit Catholique de Louvain, Louvain-la-Neuve, Belgium, Victor DeMiguel, Frédéric Vins Factor risk parity portfolios are defined such that the portfolio risk is projected homogeneously on the basis of uncorrelated principal components. This basis is however arbitrary: it is one basis of uncorrelated factors out of infinitely many others, each yielding different optimal solutions. Instead, looking for a basis of independent factors is more meaningful for non-Gaussian assets, and removes this arbitrariness. This basis is found by independent component analysis (ICA), a non-linear higher-order extension of PCA. ICA also provides a simplified setting to deal with higher-order risk parity. The proposed portfolios exhibit solid performances in terms of Sharpe ratio and downside risk. n WC24 North Bldg 131B Practice- Programming and Applications I Contributed Session Chair: Frank Muldoon, Applied Materials, Piedmont, SC, 29673, United States 1 - Revised Adaptive Linear Programming Algorithm Lin Guo, University of Oklahoma, Norman, OK, 73071, United States We revise Adaptive Linear Programming (ALP) to have fewer heuristics in determining the value of critical parameters, using the insight that we get through post-solution analysis. ALP is a second-order derivative linearization algorithm. The algorithm has limitations: 1) the value of critical parameters are determined with heuristics; 2) we use the critical parameters to control the approximation accuracy as well as convergence efficiency, but little knowledge of their tradeoffs has been used to make rules of the parameters setting and updating. To fill in these gaps, we revise the ALP algorithm to self-update the critical parameters using the knowledge from the post-solution analysis. 2 - Managing Navigation Channel Traffic and Anchorage Area Utilization of a Container Port Shuai Jia, The Hong Kong Polytechnic University, Kowloon, Hong Kong, Chung-Lun Li, Zhou Xu Navigation channels are fairways for vessels to travel in and out of a container terminal basin. The capacity and availability of a navigation channel is restricted by the traffic lanes and tides. When the navigation channels run out of capacity, the anchorage areas can serve as a buffer. This paper aims to simultaneously optimize the navigation channel traffic and anchorage area utilization of a container port. We provide a mixed integer program of the problem, analyze its complexity, and propose a Lagrangian relaxation heuristic for solving the problem. Computational performance of the heuristic is evaluated on problem instances generated based on the operational data of a port in Shanghai. 3 - On Sensitivity Analysis of Linear Integer Program: The Case of Stochastic Programming and Non-linearity in Parameters CY (Chor-yiu) Sin, National Tsing Hua University, Kuang-Fu Road, Hsinchu, 30013, Taiwan A stochastic programming with P states can well be formulated as a deterministic programming with P-1 probabilities. In this and the case with non-linearity in parameters, when one parameter changes the others may also change. Consequently the existing sensitivity analysis in the linear integer program may not be applicable, as it confines the attention to one-parameter change. This paper considers a special type of Lagrangian dual function which renders strong duality. Using this Lagrangian dual function, we first generalizes the result in Shapiro (1977) to multi-parameter cases. Further, we show with synthetic data that our analyses save a lot of computer time in large-scale optimization.

477