14.4 Markowitz Portfolio Model m5 Goo Emer 6 Subassembly 2 5 : Fabrication Assembly ‘ *Teol Bin . 7 x Location Station XY. Fabrication 14 Paint 12 Subassembly1 25 2 Subassembly2 3 5 Assembly 4 4 We may measure the distance from a station to the tool bin located at (X, ¥) by using Euclidean (straight-line) distance. For example, the distance from fabrication located at the coordinates (1, 4) tothe tool bin located at the coordinates (X, ¥) is given by aT “The unconstrained optimization problem is as follows: Min (JQx— 1) +4) + Jey + 2) + x25) FO + =a) + -5f + ay + -y) Note that we do not require that the variables X or Y be nonnegative. The optimal solution found by Excel Solver is X = 2.230, ¥ = 3.349, The solution is shown in Figure 14,10, Location models ate used extensively for determining the optimal locations for The exercises atthe end ofthis chapter prowdo practice in creating sever Sfrent forms of cation everything from drilling holes in computer circuit boards to locating distribution models centers and retail stores in supply chains. A variety of location models can be created by using different objective functions or by adding additional constraints on distances taveled, 14.4 Markowitz Portfolio Model Harry Markowitz received the 1990 Nobel Prize for his ground-breaking work in portfolio ‘optimization. The Markowitz mean-variance portfolio mode! is a classic application of nonlinear programming, In this section, we present the Markowitz mean-variance portfolio model. Money management firms throughout the world use numerous variations of this basie modelChapter 14 Nonlinear Optimization Models ‘A key trade-off in financial planning is that between risk and return, For a chance to ‘cam greater returns, the investor must also aecept greater risk. In most portfolio optimiza- tion models, the return used isthe expected (or average) return of the possible outcomes, and the risk is some measure of variability in these possible outcomes. To illustrate the “Markowitz portfolio model, let us consider the case of Hauck Investment Services. Hauck Investment Services designs annuities, IRAs, 401(K) plans, and other investment Vehicles for investors with a variety of risk tolerances. Hauck would like to develop a portfolio model that can be used to determine an optimal portfolio involving a mix of six ‘mutual funds. Table 14.2 shows the annual return (%e) for five 1-year periods for the six ‘mutual funds. Year | represents a year in which all mutual funds yield good returns. Year 2 is also a good year for most of the mutual funds. But year 3 is a bad year for the small-cap vvalue fund, year 4 isa bad year for the intermediate-term bond fund, and year Sis a bad ‘year for four ofthe six mutual funds. Is not possible to predict the exact returns for any of the funds over the next 12 ‘months, but the portfolio managers at Hauck Financial Services think that the returns for the five years shown in Table 14,2 are scenarios that can be used to represent the possibi ties for the next year. For the purpose of building portfolios for their clients, Hauck’s port folio managers will choose a mix of these six mutual funds and assurne that one of the five possible scenarios will describe the return over the next 12 months, "The portfolio construction problem is to determine how much of the portfolio to invest in each investment alternative. To determine the proportion of the portfolio that will be invested in each of the mutual funds we use the following decision variables: FS = proportion of portfolio invested in the foreign stock mutual fund 1B = proportion of portfolio invested in the imermediate-term bond fund LG = proportion of portfolio invested in the large-cap growth fund LV = proportion of portfolio invested in the large-cap value fund SG = proportion of portfolio invested in the small-cap growth fund SV. = proportion of portfolio invested in the small-cap value fund Because the sum of these proportions must equal one, we need the following constraint: FS + IB +LG+LV +SG +8V ‘The other constraints are concerned withthe return that the portfolio will earn under ‘each of the planning scenarios in Table 14.2. “The portfolio retum over the next 12 months depends on which of the possible scenarios (years I through 5) in Table 14.2 occurs. Let R, denote the portfolio return if the scenario represented by year 1 occurs, R; denote the portfolio return ifthe scenario represented by TABLE 14.2 P ery ‘Annual Return (%) Mutual Fund Year1 Year2 -Year3.«Year4 Year 5 Foreign Stock 1006 13121347 4542-2193 Intermediate-Term Bond 17.64 3.25 7st 133 7.36 Large-Cap Growth 3241 187133284146 23.26 Large-Cap Value 32.36 2061 1293 708 837 SmallCap Growth 3342 19.40 385 5848-902 Small-Cap Value 2056 28.32, 6.70 43173Other common risk Value at se VaR) and concitonal value at mak (CVaR}. One of the hhomewerk problems at tha end of the chapter Invoduees the eoncopt of 14.4 Markowitz Portfolio Model 7 year 2 occurs, and so on. The portfolio returns forthe five planning years (scenarios) are as follows: Scenario I return R Scenario 2 return O.06FS + 17.6418 + 32.41LG + 32.36LV + 33.4456 + 24.56SV. Ry = 1B.12FS + 3.251B + I8.71LG + 20.61LV + 19.408G + 25.325 Scenario 3 return: 347FS + 7SUB + 33.28LG + 12.93LV + 3.858G ~ 6.70SV R Scenario 4 return: Ry = 4542S ~ 1.3318 + 41.46LG + 7.06LV + 58.6856 + 5.438V Scenario 5 return R= ~21.93F8 + 7.3618 — 23.26LG ~ 5.37LV — 9.0286 + 17318V If p, is the probability of scenario s, among m possible scenarios, then the expected retum for the portfolio is R, where R= Spk, 46) If we assume that the five planning scenarios in the Hauck Financial Services model are equally likely to occur, then R= ur = %YR Measuring risk is a bit more difficult, Entire books are devoted to the topic of risk mea- surement, The measure of risk most often associated with the Markowitz portfolio model is the variance of the portfolio's return. If the expected return is defined by equation (14.6), the variance of the portfolio’s return is, var = Sp.(R, -R) aa) Forthe Hauck Financial Services example the ive planning scenarios are equally kel, thus var = S(R, RY ‘The portfolio variance isthe average ofthe sum of the squares of the deviation from the mean value under each scenario, The larger this number, the more widely dispersed the scenario return are about the average value. IF the portfolio variance were equal 10 72r0, then every scenario return R, would be equal, and there would be no risk. ‘Two basic ways to formulate the Markowitz, model are (1) to minimize the variance of the portfolio subject to a constraint on the expected return ofthe portfolio and (2) to maximize the expected return ofthe portfolio subject ta constraint on variance. Consider the frst ease. Assume that Hauck clients would like to construct a portfolio from the six ‘mutual funds listed in Table 14.2 that will minimize ther risk as measured by the portolio variance. However, the clients also require the expected portfolio return to be atleast 10%. In our notation, the objective function is Min %>.(R: -R) The constraint on expected portfolio return is R = 10, The complete Markowitz model involves 12 variables and 8 constrains (excluding the nonnegatvity consiains) Min ¢>.(Rs — R) (148)ne Chapter 14 Nonlinear Optimizat a 1OO6RS + 17648 + 324116 + 32.361 4 334450 + 24565V = R149) IBI2FS + 432818 + 187ULG + 2061LV + 194086 + 28:28 = R, (10) IBATRS + 751B + 382816 + 1293LV +8556 ~ 6708V =, (14.1 ASA2FS ~ 1338+ ALAGLG + TO6LV + SBOBSG + S435 =R, 142) ~21.93FS + 7.361B — 23.26LG ~ 5.37LV — 9.02SG + 17.31SV = R, (14.13) FSV VLGVIV 18G-4SV=1 (44) uER=R (4s) Re asso) FS,IB,LG,LV,SG,SV =0 (14,17) ‘The objective for the Markowitz model isto minimize portfolio variance Equations (14.9) through (14.13) define the return for each scenario, Equation (14.14) requires all of the money to be invested in the mutual funds; this constraint is often called the unity constraint. Equation (14.15) defines R, which is the expected return of the portfolio, Equation (14,16) requires the portfolio return to be at least 10%. Finally, equation (14.17) requires a nonnegative investment in each Hauck mutual fund. Note that R, Rs, Rs, Re, and Re, as well as R, are not required to be nonnegative. It is possible that the return in a given scenario or the expected return of the portfolio is negative. ‘The solution for this model using a required retuen of atleast 10% appears in Figure 14.11, The minimum value for the portfolio variance is 27.136, This solution implies FIGURE 14.11 ae ee 3k s cose pat ® r14.4 Markowitz Portfolio Model m9 thatthe clients will get an expected retum of 10% (R= 10) and minimize their sk as mea sured by portfolio variance by investing approximately 16% ofthe portfolio in the foreign stock fund (FS = 0.158), 53% in the intermediate bond fund (/B = 0.525), 4% in the large- cap growth fund (LG = 0,042), and 27% in the small-cap value fund (SV = 0.274). The Solver Parameters dialog box i also shown in Figure 14.11, Note that we have selected GRG Nonlinear asthe method and we have nor selected Make Unconstrained Variables Non-Negative. Instead we have entered as an explicit constraint set that B17 through B22 must be= 0 The Markowitz portolio model provides a convenient way for an investor to trade off risk versus return. In practice, this model is typically solved iteratively for different val tes of return, Figure 14.12 sa graph ofthe minimum portfolio variances versus required expected retums as required expected retum is varied from 8 to 12% in increments of 1. In finance, this graph is called the efficient frontier. Each point onthe ecient fron- ticr is the minimum possible risk (measured by portfolio variance) for the given return. By Joking atthe graph ofthe efficient frontier, investors can sclct the mean-variance combi- nation with which they ae most comforbl HauckMarcowis OT ai en ne Notice tha n givan in Figure 14.11 has more than 50% ofthe portfolio invested the intermaciate-tarm bond fund. It may be urwise to allow ore asst to contrio- the: Upper and lower bourds on the amount ofan asset type in the portfolio can be easly ute 50 heavily to Inthissection, pontoio variance war ured to measure risk as\tie defined, counts davis above and below the mean, Mostinvestors are happy with However variance, re both 1d. Hence, upper bounds are on placed percentage ofthe portolio invested in singe eset. Like he raturns above the meanbut wish to avoid eatuns bow the mean, Hence, numerous portolio models allow for exible wise itmight be undesirable to include an extremely sl ‘quantity of an ass in the portfolio, Th constraints that require nonzero amaunts of an atsettobe 4, Inpract sta minimum percentage of th 2. Inthe Hauck exarpl, 100% o invested in mutual funds. However, risk-averse investors ‘often prefarto have some oftheir money ina so-called rsk froo asset, such a5 US. Treasury Bills. Thus, many peralio risk measures. A problem atthe end of this chapter ius hare may be ates the use of alternative risk measures ce, bath brokers and mutual nd companies adjust lo ponfolos as new information becomes available, Hone oliowss constantly adjusting a porflio may lead to larg action costs. The case problem at she end cf this requires you to develop 2 mocificatian of the Markowitz pporoio selection prble 5 asia teeta nena 45 40 30 25 201 5 Portfolio Variance 9 10 1 12 Required Return (%)