• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

Re-engineering financial planning for institutional investors

ProQuest Dissertations and Theses, 2009
Dissertation
Author: Woo Chang Kim
Abstract:
In this dissertation, we discuss several topics that arise throughout the financial planning process for institutional investors employing the asset allocation approaches. First, we evaluate characteristics of the current equity style segmentation rule as the basis to define core asset classes. We find that the traditional style definitions lack several key aspects such as stable definitions, critical diversification capabilities, low turn-over, and potential performance enhancement under the current practical settings. In contrast, the empirical test results indicate that industries possess better characteristics, and thus, are proper building blocks for portfolio construction. Second, we investigate the effect of the duration enhancing overlay (DEO) strategies to the defined benefit pension plans, many of which remain underfunded since the 2001-2002 recessionary periods. We show that the DEO strategies can improve performance and reduce risks by adding duration to the portfolio. Versions of the strategies are evaluated via historical data as well as forward-looking economic projection system. Third, we evaluate several versions of the Markowitz portfolio model with respect to patterns in equity markets. We form a long-only portfolio of momentum strategies via industry-level assets; the strategy beats many others over numerous markets and time periods and provides a good benchmark for competing optimization models. Simple Markowitz models are quite effective, as long as the proper historical time period is chosen for the stochastic projections. Investment performance of optimal asset allocation models can be improved by considering the momentum effects in the parameter estimation procedures. Last, we discuss the role of momentum in the performance of active equity managers. Empirical evidence shows that the excess return patterns of long-only industry-level momentum strategies are highly correlated with the active funds in the growth and the core domains, especially since the publication of the momentum effects in 1993. The best managers possess stronger similarities as compared to the worst performing managers, who have low correlation with momentum. Investment performance of momentum strategies at the industry level is competitive, lying between top 10% and 25% of funds on each period. We speculate on the causes and persistence of these patterns, relative to optimal asset allocation.

Contents

Abstract........................................................................................................iii Acknowledgements.....................................................................................v Contents.......................................................................................................vii List of Figures..............................................................................................xi List of Tables...............................................................................................xii 1

Introduction.........................................................................................1 1.1 Identifying proper security classification rules within the equity domain… 2 1.2 Enhancing performance by adding customized investment vehicles……… 4 1.3 Incorporating market anomalies to single-period portfolio models………..6 1.4 Identifying characteristics of winner funds…………………………...…… 7 1.5 Dissertation outline………………………………………………………… 8 2 Evaluating style investment – Does a fund market defined along equity

styles add value?..................................................................................11 2.1 Data………………………………………………………………………… 16 2.2 Equity segmentation schemes for empirical tests…………………………..17 2.2.1 Current classification rules – style and industry…………………....18 2.2.2 Customized indices………………………………………………...21 2.2.3 Issues regarding survivorship bias within customized indices….….28 2.3 Results from empirical tests………………………………………………..28 vii

2.3.1 Classification definitions………………………….………………..28 2.3.2 Diversification potentials………………………….………………..32 2.3.3 Membership consistency………………………….………………..39 2.3.4 Misclassification errors………………………….……………….… 44 2.4 Chapter summary…………………………………………………………...46 3 Duration-enhancing overlay strategies for defined-benefit pension plans

.……………………………………………………………………….48 3.1 Two approaches in core portfolio management for DB pension plans…….49 3.2 Duration enhancing overlay strategies……………………………………..51 3.3 Empirical Results with Historical Data…………………………………….54 3.4 Empirical results with a forward-looking scenario generator (CAP:Link)...60 3.4.1 Data Summary and Problem Settings…………………….………...62 3.4.2 Applying DEO Strategies within an ALM Model……………….… 66 3.5 Chapter summary…………………………………………………………..70 4 Linking momentum strategies with single-period portfolio models… 71 4.1 Models: Markowitz model and its variants………………………………...73 4.1.1 Robust estimator: Black-Litterman………….……………………..75 4.1.2 Re-sampled portfolio: Grauber-Hakansson……………….………..77 4.1.3 Robust optimization: SOCP for ellipsoidal relaxation of μ………...79 4.1.4 Markowitz model and general experiment settings….….………….81 4.2 Test results……………………….………………………………………… 82 4.2.1 Model comparisons: which model is better?.....................................83 4.2.2 Comparisons along different look-back periods: blending in momentum effects………………………………………………… 85 viii

4.3 Chapter summary……………….…………………………………………..90 5 Active equity managers in the U.S.: do the best follow momentum strategies?.............................................................................................91 5.1 Data and industry-level momentum portfolios………………………..…… 92 5.2 Empirical analysis…………………………………………………………..95 5.2.1 Comparing average funds to momentum portfolios…….………….96 5.2.2 Are the best more likely to follow momentum strategies?................99 5.2.3 Can individual funds benefit from following momentum rules?......102 5.3 Chapter summary…………………………………………………………...108 6 Conclusions and future research directions….……………………... 109 6.1 Designing optimal security classification rules and re-evaluating styles as stock risk factors……………………………………………………………109 6.2 Extending DEO strategies to other investment domains…………………...110 6.3 Incorporating momentum effects with ALM models………………………111 A Tests on diversification potential for different style indices (S&P-Citi,

DJ-Wilshire)…………………………………………………………113 A.1 Average correlations of S&P-Citi and DJ-Wilshire style indices………….114 A.2 Graphical illustration of trend in average correlation………………………115 B Test results with 30-yr government bond index as a proxy for liabilities ……………………………………………………………………….116 B.1 Performance of key strategies over 1982 to 2007 with 30-yr bond index as liability……………………………………………………………………...116 B.2 Performance of 60-40 strategies after applying DEO with 30-yr bond index as liability……………………………………………………………………...117 ix

B.3 Performance of 70-30 strategies after applying DEO with 30-yr bond index as liability……………………………………………………………………...118 C Additional results on empirical tests with ALM models…………… 119 C.1 Additional ALM model (1)…………………………………………………120 C.2 Additional ALM model (2)…………………………………………………122 C.3 Additional ALM model (3)…………………………………………………124 C.4 Additional ALM model (4)…………………………………………………126 C.5 Additional ALM model (5)…………………………………………………128 C.6 Additional ALM model (6)…………………………………………………130 D List of Datastream Industry Classification Benchmark (ICB)………132 References………………………………………………………………...133

x

List of Figures

1.1 Dissertation outline………………………………………………………………10 2.1 Size classification on domestic stock universe by Russell…………………….19 2.2 Average correlations of various segmentation schemes……………………….35 2.3 Efficient frontiers from actual style and industry indices under perfect information ……………………………………………………………………………………37 2.4 Misclassification errors of customized indices in ST4-C during 1989 to 2008 ………………………………………………………………………………..…..45 3.1 Comparisons on portfolio performance at different levels of DEO.......…………59 3.2 CAP Link – cascade of stochastic processes………………………….…………61 3.3 Efficient frontiers of ALM models…………………………………...………….67 4.1 Tree representation of the Grauber-Hakansson model………………...………...77 4.2 Performance of Markowitz (M), Black Litterman (BL), Grauber-Hakansson (H), and robust optimization (R) Models from Jan.1980 to Dec.2007………..………84 4.3 Performance across different look-back periods……………………...………….86 4.4 Performance Comparisons to Selected Benchmarks………………...…………..88 5.1 Comparisons of performance of momentum portfolio and representative funds …………………………………………………………………………………..108 a

xi

List of Tables

2.1 Correlations on daily returns of Russell style indices and corresponding customized indices……………………………………………………………….24 2.2 Correlations on daily returns of ICB indices and corresponding customized indices……………………………………………………………………………24 2.3 Equity segmentation schemes for empirical tests………………………………..26 2.4 Number of stocks assigned to different growth-value universe by S&P and Russell……………………………………………………………………………31 2.5 Turnovers of Russell style indices in ST4-A during 2003 to 2007………………41 2.6 Turnovers of customized indices in ST4-C, ST9-C-NOL, and ICB-C during 1989 to 2008…………………………………………………………………………...43 3.1 Performance of the key strategies………………………………………………..55 3.2 Performance of key strategies after applying DEO……………………………...57 3.3 Summary statistics of data generated by CAP:LINK……………………………63 3.4 List of performance-risk measures for ALM model……………………………..65 3.5 Performance-risk measures of ALM models…………………………………….68 4.1 Summary for asset allocation models……………………………………………82 5.1 Style classification, corresponding benchmark indices and number of funds…...93 5.2 Performance of momentum strategies from 1987 to 2006……………………….95 5.3 Excess return correlations of active fund and momentum strategies…………….98 xii

xiii

5.4 Correlation analysis for ranked active fund groups…………………………….101 5.5 Comparisons of investment performance among funds with different correlation levels to momentum portfolios…………………………………………………104 5.6 Comparisons of investment performance and momentum similarity…………106 s

1  

Chapter 1

Introduction

A hallmark of current investment management practice involves asset allocation decisions among a set of broadly separated asset categories. The importance of sound asset allocation decisions has been studied extensively and, in fact, is a part of the gospel of investment management. For instance, Hensel et al. (1991) show that asset allocation determines more than 90% of a typical pension fund’s return among several investment decisions such as security selection and market timing. Consequently, many institutional investors such as pension plans and university endowments employ asset allocation procedures for their financial planning. While asset allocation addresses decision making at the broad asset level, a portfolio is the collection of the single securities. Therefore, in order to construct their actual portfolios, investors employing asset allocation should follow a sequential decision making process. The first stage is to identify a feasible set of asset classes. Naturally, common practice defines the asset classes by grouping similar single securities together. A target mix is set for the portfolio at the second stage. Herein, various methods from portfolio theory can be applied based on their investment objectives. Third, the actual portfolio is

2   constructed by assigning capital to various types of passive and/or active funds, corresponding to the broad asset classes defined at the first step. It is clear that following the standardized procedure allows investors to make their investment decisions for financial planning more efficiently. Importantly, there exist critical issues embedded at each stage which should be carefully addressed in order to obtain good investment performance. In this dissertation, we identify the key problems for the institution’s decision making process and study possible improvements of the current practice.

1.1. Identifying proper security classification rules within the equity domain The main task at the first stage of financial planning is to construct a set of asset classes from the single securities. While the concept of grouping similar securities is straight- forward, the actual execution could be tricky for the definition of the similarities is not unique. For instance, stocks can be segmented by their return characteristics which are often referred to as equity styles, while they can also be grouped based on the industries that the companies belong to. One question arises from this reasoning – what is the best way to define the broad asset classes from the set of available stocks to improve investors’ performance? In other words, what is the best equity classification rule for the institutional investors to maximize the benefits of asset allocation? In fact, this question has a subtle, but profound implication on the active fund market structure. Investors employing asset allocation manage their portfolios by controlling their positions on broad asset classes while

3   delegating micro management within each asset class to the hired fund managers. Consequently, the equity fund market structure is closely aligned with the equity classification rule. Therefore, by answering the question, we can also see if the current fund market structure is adequately defined. For the empirical tests, we focus on four key features. First, the definition for the stock classification should be universal. Ambiguous definition may lead different index providers to compose the same category with significantly different constituents, causing too much grey region. Second, segmentation should provide acceptable diversification potentials. Portfolio performance is heavily dependent upon the diversification effect, and it can be better achieved when assets possess low correlations. Third, each break-out should possess a reasonable level of persistence on membership. In addition to high transaction costs, high turnover causes poor performance of active funds for it acts as a constraint for fund managers’ investment decisions. Last, classified stocks should stay in the designated segment between reconstruction dates. Stocks are classified partly based on forecast variables such as earnings growth, and the component lists are updated occasionally in practice. Thus, misclassification due to forecasting errors as well as changes in the key variables is unavoidable. However, it can discount the very meaning of the asset allocation, if the value is too high. We conduct empirical tests to compare the equity style classification, the dominantly employed equity classification rule, to various segmentation rules including the industry classification. Surprisingly and unfortunately, the test results suggest that the style segments lack the key features. First, the definitions on the equity styles contain ambiguity, so it is inevitable to encounter various versions of style classification

4   methodologies. Second, the style segments are highly correlated, leading to possibly poor diversification. Third, the turnovers are high, reaching up to more than 50%. Last, the misclassification errors are excessive. The implication is obvious – investors can improve their performance by simply employing the industry classification rule, instead of the style segmentation scheme. Consequently, the current fund market structure may need to be revised.

1.2. Enhancing performance by adding customized investment vehicles Investors often find it difficult to achieve their goals by constructing portfolios only from traditional asset classes. Therefore, another important task at the first stage is to include customized investment vehicles such as futures and options in the set of the feasible asset classes. A good example comes from the defined-benefit (DB) pension plan management domain. Since retirees are guaranteed to receive a pre-determined amount of money in the future from the pension plan, the liability can be understood as a form of a fixed income security. Thus, a natural approach is to hedge interest rate risks via immunization. However, this conservative approach provides relatively low returns, which could give rise to the expected contributions to the sponsoring company. Another popular approach as opposed to the duration matching strategy is to invest more on high performance assets. Of course, its risk exposure is higher, although the expected contributions are lower. Unfortunately, many large corporate and public pension trusts remain underfunded since the 2001-2002 recessionary periods. These plans are challenged by global demo-

5   graphic trends and the recent slowing economic conditions as well. Under these circumstances, companies are experiencing many difficulties in their pension fund management. In order to address this issue, we investigate the duration enhancing overlay (DEO) strategy. It is a zero investment strategy constructed by taking long future positions in long-dated treasury bonds, while taking opposite positions in shorter term treasury securities. As a standalone asset, it is a risky security as exemplified in the Orange County bankruptcy incident in 1994. However, we find that

investors possessing a set of long-term liabilities and future contributions such as DB pension plans, there is much to gain by implementing the DEO strategy. Intuitively, since the DEO strategy adds duration to the core portfolio, the performance pattern of core assets would become more similar to that of liabilities as the interest rate changes over time. This gives investors more room to invest in high performance securities with low duration while controlling exposure to the interest risk at a reasonable level. In a long run, the approach improves investment performance, causing healthier plans and lower contributions. We conduct forward looking tests in conjunction with the asset-liability management framework as well as empirical tests with historical data. In almost all cases, portfolios with DEO possess lower contribution risks and higher expected portfolio performance, as long as the level of DEO is not excessively employed.

6   1.3. Incorporating market anomalies to single-period portfolio models In order to set the target mix at the second stage, investors could employ various portfolio models. Among them, the most popular model within the institutional investment domain is arguably the single-period mean-variance model proposed by Markowitz (1952). However, there is a major issue regarding the practical implementation of the Markowitz model. Since the optimal portfolio from the mean-variance approach is chosen among the extreme points of the feasible region, small changes in the estimated parameters of the market distribution can cause radically different optimal points. Thus, relatively small errors in the parameter estimation can potentially cause a steep decrease of investment performance. Since such a high sensitivity is undesirable for practical applications, several versions of robust models such as robust estimator, portfolio re-sampling, and robust optimization techniques have been proposed. Unfortunately, empirical tests suggest that investment performance of such models has not been outstanding in many cases. What could be the possible remedies? In order to answer this question, we investigate performance patterns of four single period models (Markowitz, Black- Litterman, Grauber-Hakansson, and SOCP relaxation on μ). One of the significant results from the empirical tests is that, for every model, the out-of-sample performance is better when the input parameters are estimated from the recent historical data (up to 12 months) than the longer time horizon (3 to 5 years). Also, the efficient frontiers preserve the normal shape (upward concave slope) when the look-back period is shorter, while they become downward sloping when it is longer than a year.

7   Interestingly, we find that these performance patterns share similar features with the momentum effect. Empirical studies suggest that winner stocks for the past 3 to 12 months outperform loser stocks for the following 3 to 12 months and show worse performance after 3 to 5 years. Therefore, when shorter holding periods are employed, the models put more weights on the recent winners, leading to a successful blend of the momentum effects with optimal asset allocation models. In contrast, the portfolios from longer look-back period bet against the momentum effects, which would potentially cause inferior investment performance. It can be speculated that investors can incorporate single-period models with the momentum effect to exploit the asset return predictability simply by estimating the input parameters with relatively recent data.

1.4. Identifying characteristics of winner funds At the last stage of financial planning, institutional investors assign their capital to the money managers according to the target mix set at the second stage. Jensen (1968), Cumby and Glen (1990), and Eun, et al. (1991) have illustrated that the average mutual fund does not outperform the benchmark portfolio if transaction costs and fees are taken into account. Therefore, in order to achieve their financial objectives, investors are required to identify fund managers with strong performance, namely, winners. One of the natural approaches for this task is to predict a fund manager’s performance by investigating her past record. However, previous studies indicate that it might not be easy to do so. For instance, Jensen (1968) shows that there is no strong evidence of the performance persistence. Carhart (1997) and Goetzmann et al. (1998) also report results supporting Jensen’s finding.

8   Thus, we employ an alternative approach. Instead of identifying which funds are winners, we try to identity investment strategies that have similar return patterns with the winner funds. If such strategies can be obtained, investors could improve performance by hiring money managers who are employing the strategies, or simply by forming portfolios according to the strategies to replicate winner’s performance. Empirical evidence from a database free of survivorship bias suggests that the long- only industry-level momentum strategies could be a good candidate. Within the growth and the value universes, the excess returns of the momentum strategies are highly correlated with the active funds in the growth and the core domains, especially since publication of the momentum effects in 1993. Also, the best managers possess stronger similarities as compared to the worst performing managers, who have low correlation with momentum. Furthermore, investment performance of momentum strategies at the industry level is competitive, lying between top 10% and 25% of funds on each period.

1.5. Dissertation outline In this dissertation, we discuss the topics above as illustrated in Figure 1.1. In chapter 2, we first evaluate the current equity segmentation rules and search for a better alternative. The data for the empirical analysis is described in 2.1. Section 2.2 summarizes the current equity segmentation rules (2.2.1) as well as the candidates for potential improvements (2.2.2). We also discuss the issues regarding the survivorship bias in 2.2.3. Section 2.3 documents the empirical test results for both historical and customized equity market segments. Section 2.4 summarizes the chapter.

9   Chapter 3 documents an example of the customized investment vehicles that can improve the defined-benefit pension plan management. Two dominant approaches in the pension plan management domain are described in 3.1, and a specialized overlay strategy for DB pension plans is explained in 3.2. Section 3.3 provides the empirical test results with historical data, and section 3.4 shows the results from a forward-looking model. The chapter summary follows in 3.5. Chapter 4 discusses a technique to link the stock market anomaly to the single-period portfolio models. In section 4.1, we briefly discuss four asset allocation models employed in the chapter – traditional Markowitz, Black-Litterman, Grauber-Hakansson, and robust optimization models. The empirical test results are shown in 4.2. In 4.2.1, we compare the investment performance of the single period models. Section 4.2.2 illustrates how the momentum effects can be exploited in the portfolio models. Section 4.3 summarizes the chapter. Chapter 5 discusses the relationship between the momentum strategies and the best active funds. Section 5.1 depicts the long-only industry-level momentum strategies employed in the chapter. Empirical tests are shown in 5.2. Section 5.2.1 compares the average funds with the momentum portfolio. Section 5.2.2 illustrates that the best funds are more likely to follow the momentum strategies, and section 5.2.3 suggests that individual funds can benefit from momentum effects. The chapter is summarized in 5.3. Chapter 6 concludes the dissertation with future research topics. Note that chapters 2 to 5 are based on Kim and Mulvey (2009), Mulvey, et al. (2009a), Mulvey, et al. (2009b), and Mulvey and Kim (2008a), respectively.

10   Financial planning process for institutions Key issues

Discussion

Can current security classification rule be improved? → Section 1.1 and Chapter 2 Stage 1: identifying feasible set of asset classes → → Which non-tradition asset class can improve DB pensions? Section 1.2 and Chapter 3

Stage 2: setting target mix → How can momentum effects be exploited in single-period portfolio models → Section 1.3 and Chapter 4

Stage 3: assigning capital to money managers → Which investment strategy can replicate winner active funds? → Section 1.4 and Chapter 5

Figure 1.1: Dissertation outline

11  

Chapter 2

Evaluating style investment – Does a fund market defined along equity styles add value?

Institutional investors employing asset allocation manage their portfolios by controlling their positions on broad asset classes while delegating micro management within each asset class to the hired fund managers. Since funds are generally restricted to stay within their universes, a prevailing determinant for an institutional investor’s performance is the characteristics of the benchmark indices which define the fund market structure. For instance, when the benchmark indices are highly correlated, corresponding funds are also likely to be highly correlated potentially causing poor diversification benefits. Similarly, unstable membership leads to high turnover, causing high fees. From this, one question arises – what is the best way to define the fund universe to improve institutional investors’ performance? In other words, what is the best security segmentation scheme? We focus on the structure of the U.S. equity fund market in this chapter. The dominant security segmentation scheme in the U.S. equity market is based on the style rule. It classifies equities by market capitalizations and growth-value perspectives.

12   Accordingly, a stock with a large market-cap and a high growth score is categorized as a large-growth stock. The significance of equity styles arises from its extensive application. Mutual funds, the main suppliers of “style products”, are defined and scored along the style universe. Also, the style investment approach, which employs diversification across style segments, has become a norm in the institutional domain such as pension plans and university endowments. As of 2007, almost half of private pension plan assets worth six trillion dollars are managed under the jurisdiction of the style management 1 . Accordingly, many institutional investors make their investment decisions along the style universe defined by popular index providers such as SP-Citi, Russell, or DJ-Wilshire. Style investment and its related topics have been extensively studied. Early on, Graham (1949) and Graham and Dodd (1951) argue that investors can achieve superb performance by picking underpriced stocks through careful analysis on variables such as price-to-earning, price-to-book, and dividend yields, which is eventually defined as the value investing, a special form of the style investment which is still widely employed in practice. Theoretical backgrounds for systematic classification on equity styles have been flourished in two closely related fields in finance since 1970’s. The first one has originated from the cross-sectional factor models for expected stock returns. Sharpe (1964), Lintner (1965), and Black (1972) propose an asset pricing model (SLB), suggesting the efficiency of the market portfolio. The main implication of the model is that market betas are sufficient to explain the cross-sectional expected returns. However, empirical analyses find contradictory evidence. Banz (1981) and Rosenberg et al. (1985)

1 Both defined benefit and defined contribution pension plans are included. When retire- ment plans in public sectors are included, the scope of the style investment becomes even greater, covering approximately half of 10 trillion dollars. Board of Governors of the Federal Reserve System, 2008, Flow of Funds Accounts of the United States, Flows and Outstandings, First Quarter 2008, p75-76.

13   show that the market capitalization provides a significant explanatory factor for cross- sectional expected returns, which is often referred to as the size effect. They find that small stocks tend to have higher expected returns than the predicted values by SLB, while large stocks have lower returns. Another pattern involves variables related to book values. Stattman (1980) finds that the expected return of a stock with a high book value of common equity is higher than one with a low value. Chan et al. (1991) illustrate similar empirical results with the book-to-market equity and cashflow-to-price. Also, Fama and French (1992) show that size and book-to-market combine to explain the stock return characteristics. In addition, Basu (1983) shows that earning-to-price ratio can be also included as the explanatory factors. Accordingly, stocks with low (high) book-to-market equity, earnings-to-price, and cashflow-to-price are classified as growth (value) stocks with further segmentation by size in many related studies 2 . Another contributor to the explosion of style investing is the active equity manage- ment domain. Sharpe (1977, 1981) shows that active equity managers can be classified based on their investment styles, and they are closely aligned with the equity styles from the factor models. In response to his findings, a handful of style evaluation models for active funds have been proposed. Sharpe (1988, 1992) illustrates a return-based approach to determine a fund’s style. Brown and Goetzmann (1997) and Carhart (1997) classify fund styles based on the sensitivity of returns to style factors. In contrast, Grinblatt and Titman (1989) and Daniel et al. (1997) adopt an approach based on the characteristics of the fund holdings. Chan et al. (2002) shows that two classification methods based on factor loadings and holding characteristics generally yield similar outcomes on styles.

2 See, for example, Fama and French (1993, 1995, and 1996), Lakonishok et al. (1994), and Teo and Woo (2004)

14   To cope with the expansion in the size of managed funds, institutional investors begin to delegate micro management decisions such as stock selection to active fund managers during 60’s and 70’s. As theories and practical tools for equity styles and active management styles emerge, investors soon realize that diversification can be better achieved by differentiating their capital allocations across styles, leading to extensive implementation of style investing (Swensen (2000)). Difficulties of portfolio management via equity styles have been pointed out. Daniel and Titman (1997) report that two style factors (size and book-to-market) do not yield high co-movement within each class, and conclude that they may not be risk factors. Ahn et al. (2006) point out high correlations among size and book-to-out portfolios and claim that they do not fully generate the opportunity set. Mulvey and Kim (2008b) also report similar results. Clearly, theories on equity styles have been developed to reveal hidden characteristics of stocks and active funds rather than to optimally design equity classi- fication rules for asset allocation, so there still seems to be room for improvement in the portfolio management context. In addition, Swensen (2005) accuses the high turnover of Russell style indices as one of “obvious sources of mutual fund failure” and index fund failure. He argues that inadequate reconstruction rules cost fund managers not only high transaction costs but also poor performance by allowing activities of index fund arbitrageurs. This chapter evaluates characteristics of style classification rules under the current practical settings to see if the defined structure of the fund market provides proper core assets for institutional investors’ portfolios. Among several investment decisions, the asset allocation policy is reported to be the most important factor in determining invest-

Full document contains 153 pages
Abstract: In this dissertation, we discuss several topics that arise throughout the financial planning process for institutional investors employing the asset allocation approaches. First, we evaluate characteristics of the current equity style segmentation rule as the basis to define core asset classes. We find that the traditional style definitions lack several key aspects such as stable definitions, critical diversification capabilities, low turn-over, and potential performance enhancement under the current practical settings. In contrast, the empirical test results indicate that industries possess better characteristics, and thus, are proper building blocks for portfolio construction. Second, we investigate the effect of the duration enhancing overlay (DEO) strategies to the defined benefit pension plans, many of which remain underfunded since the 2001-2002 recessionary periods. We show that the DEO strategies can improve performance and reduce risks by adding duration to the portfolio. Versions of the strategies are evaluated via historical data as well as forward-looking economic projection system. Third, we evaluate several versions of the Markowitz portfolio model with respect to patterns in equity markets. We form a long-only portfolio of momentum strategies via industry-level assets; the strategy beats many others over numerous markets and time periods and provides a good benchmark for competing optimization models. Simple Markowitz models are quite effective, as long as the proper historical time period is chosen for the stochastic projections. Investment performance of optimal asset allocation models can be improved by considering the momentum effects in the parameter estimation procedures. Last, we discuss the role of momentum in the performance of active equity managers. Empirical evidence shows that the excess return patterns of long-only industry-level momentum strategies are highly correlated with the active funds in the growth and the core domains, especially since the publication of the momentum effects in 1993. The best managers possess stronger similarities as compared to the worst performing managers, who have low correlation with momentum. Investment performance of momentum strategies at the industry level is competitive, lying between top 10% and 25% of funds on each period. We speculate on the causes and persistence of these patterns, relative to optimal asset allocation.