You are on page 1of 86

UNIVERSITY OF SOUTHAMPTON SCHOOL OF MATHEMATICS

Does the Optimal Portfolio deliver the best Performance in the Stock Market?
A Project Report Submitted for the Award of B.Sc. in Mathematics with Economics

Martin Yau 21/5/2009

Supervisor: Dr. G. Kennedy

Summary
Portfolio theory describes how investors optimise their portfolios using mean and variance from a universe of stocks. We investigate the performance of optimal portfolios against those using investment strategies. We use MATLAB to construct the portfolios from the stocks listed on the London Stock Exchange excluding investment trusts. We briefly go through the history of Modern Portfolio Theory and other descendent theories. We explain the application of the theory under various conditions. We also explain the basic concepts of behavioural finance and the various formulas for evaluating the performance of portfolios. We conclude that the optimal portfolios perform worse than those using the investment strategies although additional investigation needs to be carried out due to the methods we used for the experiment.

2

Contents
1. 2. 3. Introduction .................................................................................................................... 9 Measures of Risk ........................................................................................................... 9 Modern Portfolio Theory............................................................................................... 10 3.1. Specifying the Opportunity Set .............................................................................. 10 Assumptions .................................................................................................. 10 Preliminaries .................................................................................................. 10

3.1.1. 3.1.2. 3.2. 3.3.

Building a portfolio ................................................................................................ 11 Examining Efficient Portfolios ................................................................................ 13 No short selling .............................................................................................. 13 With short selling ........................................................................................... 17 Including Riskless Assets and Riskless Borrowing and Lending .................... 18

3.3.1. 3.3.2. 3.3.3. 3.4. 3.5.

Optimal Portfolio ................................................................................................... 19 Finding the Optimal Portfolio ................................................................................. 20 Short selling with Riskless Borrowing and Lending ........................................ 20 Short selling without Riskless Borrowing and Lending ................................... 22 No Short selling with Riskless Borrowing and Lending ................................... 22 No Short selling and Riskless Borrowing and Lending ................................... 23 Including additional constraints ...................................................................... 23

3.5.1. 3.5.2. 3.5.3. 3.5.4. 3.5.5. 4. 5.

Behavioural Finance .................................................................................................... 24 Investment Strategies .................................................................................................. 25 5.1. 5.2. 5.3. Contrarian Investment Strategy............................................................................. 25 Value Investing ..................................................................................................... 27 Zulu Principle ........................................................................................................ 28

6.

Evaluating Investment Performance............................................................................. 29 6.1. 6.2. Rate of Return ...................................................................................................... 29 Risk Adjusted Returns........................................................................................... 29

7. 8. 9.

Plan of Experiment....................................................................................................... 31 Results of the Experiment ............................................................................................ 34 Conclusion and Recommendations for Improvement ................................................... 36

Bibliography ........................................................................................................................ 39 Appendices ......................................................................................................................... 43 A.1. Capital Asset Pricing Model (CAPM) ..................................................................... 43 Derivation of CAPM ....................................................................................... 43

A.1.1. A.2. A.3.

Quadratic Programming ........................................................................................ 48 An example of calculating the Optimal Portfolio .................................................... 50 3

A.4. A.5.

Transforming Non-Positive Semi-Definite Covariance Matrices ............................ 53 Experiment Instructions......................................................................................... 54 Contrarian Investment Strategy...................................................................... 54 Value Investing .............................................................................................. 60 Zulu Principle ................................................................................................. 62 Recording trades ........................................................................................... 63 Executing trades ............................................................................................ 65

A.5.1. A.5.2. A.5.3. A.5.4. A.5.5. A.6.

Experiment Data ................................................................................................... 67 Data of companies matching relevant criteria for Investment Strategies ........ 67 Portfolio Transaction Costs ............................................................................ 71 Portfolio Valuation at the end of Holding Period ............................................. 77 Market, Bid and Offer prices .......................................................................... 79 Stocks Returns from Portfolios ....................................................................... 81

A.6.1. A.6.2. A.6.3. A.6.4. A.6.5. A.7. A.8.

MATLAB codes ..................................................................................................... 83 Glossary of Terms................................................................................................. 85

4

List of Figures
Figure 1: Risk-return graph for two assets when ๐œŒ12 = 1. .................................................... 14 Figure 2: Risk-return graph for two assets when ๐œŒ12 = โˆ’1. ................................................. 15 Figure 3: Risk-return graph for two assets for various values of ๐œŒ12 . ................................... 16 Figure 4: Risk-return graph for ๐‘ assets each with various values of ๐‘…, ๐œŽ and ๐œŒ. ................ 16 Figure 5: Risk-return graph for two assets when ๐œŒ12 = 1 with short selling. ......................... 17 Figure 6: Risk-return graph for two assets when ๐œŒ12 = โˆ’1 with short selling. ...................... 17 Figure 7: Risk-return graph for ๐‘ assets each with various values of ๐‘…, ๐œŽ and ๐œŒ between -1 and 1. .................................................................................................................................. 18 Figure 8: Risk-return graph for riskless and risky assets with short selling. ......................... 19 Figure 9: Optimal portfolio, efficient frontier and CML. ......................................................... 20 Figure 10: Optimal portfolios A, B and C for riskless rates ๐‘…0 , ๐‘…0 and ๐‘…0 respectively which traces out the efficient frontier. .................................................................................. 22 Figure 11: Visual Tools webpage on Digital Look website. 'Heat Maps' link is highlighted by the underline. ...................................................................................................................... 54 Figure 12: Heat maps webpage on the Digital Look website. .............................................. 54 Figure 13: Average P/E values of sectors defined by LSE on the Digital Look website. ..... 55 Figure 14: Screener webpage with โ€˜full fundamental screenerโ€™ link highlighted on the Digital Look website. ...................................................................................................................... 55 Figure 15: Full Screener webpage. ..................................................................................... 56 Figure 16: Example result of application of steps 1 to 10 for Contrarian Investment Strategy. ........................................................................................................................................... 56 Figure 17: Heat map showing 12 companies with positive P/E values. The values are arranged in decreasing order so the last two are in the two lowest quintiles. ....................... 57 Figure 18: Summary page of a company. ............................................................................ 57 Figure 19: Final results showing accounts for a company going back up to five years. ....... 58 Figure 20: 'UK Shares' webpage showing indices. 'FTSE 100' is highlighted by the underline. ........................................................................................................................................... 58 Figure 21: Example result of application of steps 1 to 5 for Value Investing. ....................... 60 Figure 22: Example of application of steps 1 to 5 for the Zulu Principle. .............................. 62 Figure 23: 'Portfolio Selection' window with 'Portfolio Valuation' webpage in the background. ........................................................................................................................................... 63 Figure 24: 'Trade Security' webpage to record transactions. ............................................... 63 Figure 25: 'Portfolio summary' webpage displaying types of portfolios with their values. ..... 65 Figure 26: โ€˜Share Trading Portfolioโ€™ webpage showing stocks held in the portfolio with 'PLACE A TRADE' link above. ............................................................................................ 65 Figure 27: Illustration of step 4. ........................................................................................... 66 Figure 28: 'Stock Trading' webpage illustrating an example of a trade to buy ยฃ10,000 in Royal Bank of Scotland Group plc....................................................................................... 66 Figure 29: MATLAB code for Statdata_and_find_eigenvalues.m......................................... 83 Figure 30: MATLAB code for Gettranscovmatrix.m. ............................................................ 84 Figure 31: MATLAB code for Optimal_Portfolio.m. .............................................................. 84
(1) (2) (3)

5

List of Tables
Table 1: Timetable of experiment showing each of the optimal portfolios going to be compared with each of the portolios constructed using the investment strategies ............... 31 Table 2: Optimal Stock proportions for Optimal Portfolio 1 with relevant data...................... 34 Table 3: Optimal Stock proportions for Optimal Portfolio 2 with relevant data...................... 34 Table 4: Optimal Stock proportions for Optimal Portfolio 3 with relevant data...................... 34 Table 5: Investment Portfolio weightings ............................................................................. 34 Table 6: LIBOR used to calculate the optimal portfolios and evaluate the performance of portfolios for respective holding periods .............................................................................. 35 Table 7: Overall portfolio values based on all stocks held in the portfolios over the whole holding period ..................................................................................................................... 35 Table 8: Portfolio and market returns using various measures defined in the report............ 35 Table 9: Companies with the relevant ratios which satisfy at least all compulsory critera for Contrarian Investment Strategy........................................................................................... 67 Table 10: Sector average values for the Contrarian ratios ................................................... 67 Table 11: Companies with the relevant ratios which satisfy at least all compulsory criteria for Value Investing ................................................................................................................... 68 Table 12: Criteria satisfied by companies for respective sub-strategies for Value Investing 69 Table 13: Companies with the relevant ratios which satisfy at least all compulsory criteria for the Zulu Principle ................................................................................................................ 69 Table 14: Information which satisfies the Zulu Principle ...................................................... 70 Table 15: Parts of the Zulu Principle satisfied by companies ............................................... 70 Table 16: Transaction costs for the Contrarian Portfolio ...................................................... 71 Table 17: Transaction costs for the Value Investing Portfolio .............................................. 72 Table 18: Transaction costs for the Zulu Principle Portfolio ................................................. 73 Table 19: Transaction costs for the Optimal Portfolio 1 ....................................................... 74 Table 20: Transaction costs for the Optimal Portfolio 2 ....................................................... 75 Table 21: Transaction costs for the Optimal Portfolio 3 ....................................................... 76 Table 22: Portfolio valuation of Contrarian Portfolio at the end of the holding period before selling ................................................................................................................................. 77 Table 23: Portfolio valuation of Value Investing Portfolio at the end of the holding period before selling ...................................................................................................................... 77 Table 24: Portfolio valuation of Zulu Principle Portfolio at the end of the holding period before selling ................................................................................................................................. 77 Table 25: Portfolio valuation of Optimal Portfolio 1 at the end of the holding period before selling ................................................................................................................................. 78 Table 26: Portfolio valuation of Optimal Portfolio 2 at the end of the holding period before selling ................................................................................................................................. 78 Table 27: Portfolio valuation of Optimal Portfolio 3 at the end of the holding period before selling ................................................................................................................................. 78 Table 28: Market, Bid and Offer prices at the time of sale of stocks for the Contrarian Portfolio............................................................................................................................... 79 Table 29: Market, Bid and Offer prices at the time of sale of stocks for the Value Investing Portfolio............................................................................................................................... 79 Table 30: Market, Bid and Offer prices at the time of sale of stocks for the Zulu Principle Portfolio............................................................................................................................... 79 Table 31: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 1 ........................................................................................................................................... 80 6

Table 32: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 2 ........................................................................................................................................... 80 Table 33: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 3 ........................................................................................................................................... 80 Table 34: Stock Returns after transaction costs for the Contrarian Portfolio ........................ 81 Table 35: Stock Returns after transaction costs for the Value Investing Portfolio ................ 81 Table 36: Stock Returns after transaction costs for the Zulu Principle Portfolio ................... 81 Table 37: Stock Returns after transaction costs for Optimal Portfolio 1 ............................... 82 Table 38: Stock Returns after transaction costs for Optimal Portfolio 2 ............................... 82 Table 39: Stock Returns after transaction costs for Optimal Portfolio 3 ............................... 82

7

List of Files on the CD
Excel 1997-2003 files: LSE Data.xls LSE LISTDATE HISTORIC 2009.xls Optimal Portfolio Weights.xls Stock and Sector Ratios.xls MATLAB project files: Optimal Portfolio Project.mat Optimal Portfolio Project 2.mat Optimal Portfolio Project 3.mat MATLAB code files: Gettranscovmatrix.m Statdata_and_find_eigenvalues.m Optimal_Portfolio.m

8

1. Introduction
Portfolio theory is about rational investors using mean and variance to optimise their portfolios from a universe of stocks, which consist from two to a very large number of assets. This involves calculating the returns data from the time series of stock prices. Then this is in turn used to calculate the expected returns vector and covariance matrix. These are used to calculate the efficient frontier. The riskless rate of return is used calculate the optimal portfolio, which is determined at the point which the efficient frontier is tangent to the line of expected returns and standard deviation (or risk) with the intercept of the riskless rate. This report is aimed at final year mathematics students who have no knowledge of this field. The glossary of terms is in the Appendix A.8. The theory was first proposed in (Markowitz, 1952) and it was subsequently studied extensively by (Blume, 1970), (Davis & Norman, 1990), (Pardolos, Sandstrรถm, & Zopounidis, 1994), (Black & Litterman, 1992) and others under various cases. These cases include continuous time, transaction costs, portfolio optimisation from a universe of domestic and international stock and others. The theory leads to the Capital Asset Pricing Model in single and multiperiods, Behavioural Portfolio Theory, Post-Modern Portfolio Theory and Stochastic Portfolio Theory. From (Shefrin & Statman, 2000), Behavioural Portfolio Theory is about investors optimise their portfolios using expected wealth, desire for security and potential, aspiration levels, and probabilities of achieving aspiration levels. Post-Modern Portfolio Theory is just an extension of Modern Portfolio Theory since the downside semi-variance (this is explained in the next section) is used instead of variance to calculate the efficient frontier and optimal portfolio (see (Post-modern portfolio theory comes of age, 1993)). Stochastic Portfolio Theory basically takes into account the random movements of asset prices (see (Fernholz, 2002) for more details). The details of these are beyond the scope of this report. In this report we will compare the performance of the optimal portfolios against those using the investment strategies. The portfolios will be constructed from the universe of stocks listed on the London Stock Exchange (LSE) excluding investment trusts using the classical Portfolio Theory. In the next section we define and explain the different measures of risk. In section 3 we explain the concept of Modern Portfolio Theory and how it is applied to find optimal portfolios. In section 4 we explain the basic concepts of behavioural finance and the reasons why investor are irrational. This is important because one of the investment strategies was defined to help investors to be more rational in decision making. In section 5 we define the investment strategies and explain the history of each briefly. In section 6 we describe the various measures of portfolio performance evaluation. In section 7 we outline the assumptions and plans of the experiment. In section 8 we present the results and explain them in detail. In the final section we present the key findings from the results and suggest potential improvements to the experiment to future researchers.

2. Measures of Risk
The variance of return is defined as
+โˆž โˆ’โˆž ๐œ‡

โˆ’ ๐‘ฅ 2 ๐‘“ ๐‘ฅ ๐‘‘๐‘ฅ

where ๐œ‡ is the mean return at the end of the chosen period and ๐‘“ ๐‘ฅ is the probability of density function of return. The advantages of using this as a measure of risk are it is easy to use and this leads to neat optimal portfolio solutions. The disadvantage is that the distribution of returns can be asymmetric so using the variance is inappropriate in that case. The downside semi-variance of return is 9 ๐œ‡

โˆ’โˆž ๐œ‡

โˆ’ ๐‘ฅ 2 ๐‘“ ๐‘ฅ ๐‘‘๐‘ฅ.

This is less easy to use and it takes no account of variability above the mean. This measure is useful in the cases where the distribution of returns asymmetric. Hence, this is used in Post-Modern Portfolio Theory.

3. Modern Portfolio Theory
The entire section is based on (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 44-246) and (Levy & Sarnat, 1984, pp. 235-266). Modern portfolio theory is about how investors can construct a portfolio that gives the maximum return for a given level of risk, or a level of return for minimum risk. This is also known as Mean-Variance theory. It is divided into two parts; ๏‚ท First, the specification of the opportunity set, which is the universe of portfolio choices available to the investor. The higher the number of assets available to the investor, the larger the opportunity set. Second, determination of the portfolio out of all feasible portfolios in the opportunity set.

๏‚ท

3.1.

Specifying the Opportunity Set

3.1.1. Assumptions We assume the following; 1) Investors build portfolios based on the expected return and variance of that return of the assets over a single time period. 2) Investors prefer a portfolio with a higher return to one with a lower return at a given level of risk. 3) Investors are risk averse, where they prefer a portfolio with lower risk to one with higher risk for a given level of return. 3.1.2. Preliminaries An efficient portfolio is one that has the highest possible expected return with the lowest possible risk. Therefore the investor cannot find a better one with higher expected return and lower risk than the former. A portfolio is inefficient if the investor can find another one with the same expected return with lower risk, or same level of risk with higher expected returns. Other portfolios can be ignored when the set of efficient portfolios has been identified. We assume that asset ๐‘– has a return ๐‘…๐‘– , which is known in various time periods ๐‘ก๐‘ฃ where ๐‘ฃ = 1, โ€ฆ , ๐‘‰. Then the expected return on ๐‘…๐‘– is defined as 1 ๐ธ ๐‘…๐‘– = ๐‘…๐‘– = ๐‘‰ ๐‘‰ ๐‘…๐‘–

๐‘ก๐‘ฃ . ๐‘ฃ
=1

The length of time period can be in days, months, or years. In this report, we assume that returns have equal probabilities. The variance of ๐‘…๐‘– is defined as

10 ๐œŽ๐‘–

2

= ๐ธ ๐‘…๐‘– ๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘–

2

1 = ๐‘‰ ๐‘‰ ๐‘…๐‘–

๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘– 2 . ๐‘ฃ
=1

The variance measures the distance of returns of assets from the mean return. In financial terms, variance is risk so the higher the variance means higher risk and vice versa. Standard deviation of return is also risk due to the fact that it is the square root of variance. The covariance ๐œŽ๐‘–๐‘— between assets ๐‘– and ๐‘— is defined as ๐œŽ๐‘–๐‘— = ๐ธ ๐‘…๐‘– ๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘– ๐‘…๐‘— ๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘— 1 = ๐‘‰ ๐‘‰ ๐‘…๐‘–

๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘– ๐‘…๐‘— ๐‘ก๐‘ฃ โˆ’ ๐‘…๐‘— . ๐‘ฃ
=1

(1)

The covariance measures the correlation between two assets. We can represent equation (1) as ๐œŽ11 ๐œŽ = ๐œŽ21 ๐œŽ31 ๐œŽ12 ๐œŽ22 ๐œŽ32 ๐œŽ13 ๐œŽ23 . ๐œŽ33 (2)

From equation (2) we see that the covariance matrix is symmetrical since ๐œŽ๐‘–๐‘— = ๐œŽ๐‘—๐‘– and ๐œŽ๐‘–๐‘– = ๐œŽ๐‘–2 . The values of the non-diagonal elements can be positive, zero or negative and the maximum value is specified by ๐œŽ๐‘–2 ๐œŽ๐‘—2 = ๐œŽ๐‘– ๐œŽ๐‘— . If ๐œŽ๐‘–๐‘— is near the above value, then the prices of the two assets move together, or they are closely correlated with each other. If ๐œŽ๐‘–๐‘— = 0, then the price movements of the assets are independent and they have no correlation. If ๐œŽ๐‘–๐‘— = โˆ’ ๐œŽ๐‘– ๐œŽ๐‘— , then prices of the assets move in opposite directions. However, it is often easier to use correlation ๐œŒ๐‘–๐‘— between the two assets and this is defined by ๐œŒ๐‘–๐‘— = ๐œŽ๐‘–๐‘— = ๐œŽ๐‘– ๐œŽ๐‘— ๐œŽ๐‘–๐‘— ๐œŽ๐‘–๐‘– ๐œŽ๐‘—๐‘— .

The values of ๐œŒ๐‘–๐‘— lie between -1 and 1. In the case of ๐‘ assets, the arguments are very similar. Now we can assume that the portfolio has ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๐‘ risky assets ๐‘†๐‘– , where ๐‘– = 1, โ€ฆ ๐‘. An expected return for each risky asset. A variance of return regarding each asset. A covariance between each pair of risky assets. In this report we assume that investors hold no riskless assets.

3.2.

Building a portfolio

We will explain how a portfolio ฮ  is constructed by considering two assets ๐‘†1 and ๐‘†2 . We will assume that ๐œ†1 + ๐œ†2 = 1 where ๐œ†1 and ๐œ†2 are proportions of each asset in a portfolio. We note that the portfolio return is the weighted average of return of individual assets. The weight applied to each individual return is the fraction of the portfolio invested in that asset. 11

Therefore, the expected return of a portfolio ๐‘…ฮ  is a weighted average of the expected returns of individual assets so we note that ๐‘…ฮ  = E ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 = ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 . (3)

2 Using the same weighted average argument, the variance of a portfolio ๐œŽฮ  is also the weighted average of the variance of the return on the portfolio from the mean return on the portfolio. We have 2 ๐œŽฮ  = ๐ธ ๐‘…ฮ  โˆ’ ๐‘…ฮ  2 = ๐ธ ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 โˆ’ ๐œ†1 ๐‘…1 โˆ’ ๐œ†2 ๐‘…2 = ๐ธ ๐œ†1 ๐‘…1 โˆ’ ๐‘…1 + ๐œ†2 ๐‘…2 โˆ’ ๐‘…2

2 2 2

2 = ๐ธ ๐œ†1 ๐‘…1 โˆ’ ๐‘…1 2 + 2๐œ†1 ๐œ†2 ๐‘…1 โˆ’ ๐‘…1 ๐‘…2 โˆ’ ๐‘…2 + ๐œ†2 ๐‘…2 โˆ’ ๐‘…2 2 2 2 2 = ๐œ†1 ๐œŽ1 + 2๐œ†1 ๐œ†2 ๐œŽ12 + ๐œ†2 ๐œŽ2 2 2 2 2 = ๐œ†1 ๐œŽ1 + 2๐œ†1 ๐œ†2 ๐œŒ12 ๐œŽ1 ๐œŽ2 + ๐œ†2 ๐œŽ2 . 2

(4)

We notice from equation (4) that if ๐œŒ12 > 0 then the portfolio variance is larger than the weighted sum of the individual variances therefore risk is higher. Conversely, if ๐œŒ12 < 0 then the risk is lower. In the case of ๐‘ assets, the above argument can be applied so suppose that a portfolio is given by ๐‘
๐‘

ฮ = ๐‘–
=1 ๐œ†๐‘–

๐‘†๐‘– ,

where ๐‘–
=1 ๐œ†๐‘–

= 1.

The expected return is ๐‘
๐‘ ๐‘…

ฮ  = ๐‘–
=1 ๐œ†๐‘–

E ๐‘…๐‘– = ๐‘–
=1 ๐œ†๐‘–

๐‘…๐‘–

and the variance is ๐‘
2 ๐œŽฮ  ๐‘ ๐‘

= ๐‘–
=1 ๐œ†

2 ๐œŽ๐‘–2 ๐‘–

+ ๐‘–
=1 ๐‘— =1 ๐‘— โ‰ ๐‘– ๐œ†๐‘–

๐œ†๐‘— ๐œŒ๐‘–๐‘— ๐œŽ๐‘– ๐œŽ๐‘— .

(5)

Regarding equation (6), if the covariances of all assets are zero, that is the prices of the assets move independently, then ๐œŒ๐‘–๐‘— = 0 and we end up with ๐‘
2 ๐œŽฮ 

= ๐‘–
=1 ๐œ†

2 ๐œŽ๐‘–2 . ๐‘–

Suppose that each asset is held in equal proportions, that is ๐œ†๐‘– = 1 ๐‘ for all ๐‘–, then
2 ๐œŽฮ 

1 1 = ๐‘ ๐‘ ๐‘ ๐œŽ๐‘–

2 . ๐‘–
=1

2 As ๐‘ tends to infinity ๐œŽฮ  tends to zero, implying that in theory holding more uncorrelated assets reduces risk and this disappears as the number of holdings increases to a very large number. However, assets are almost never uncorrelated in reality. Diversification may

12

reduce the portfolio risk to a level which is less than the risk of an individual asset. We again assume that ๐œ†๐‘– = 1 ๐‘ for all ๐‘– with ๐œŒ๐‘–๐‘— โ‰  0, then the portfolio variance will be the same as equation (4). Again assume ๐œ†๐‘– = 1 ๐‘ , so we have
2 ๐œŽฮ 

1 1 = ๐‘ ๐‘ ๐‘ ๐‘ ๐‘ ๐œŽ๐‘–

2 ๐‘–
=1

+ ๐‘–
=1 ๐‘— =1 ๐‘— โ‰ ๐‘–

1 ๐œŒ ๐œŽ ๐œŽ ๐‘ 2 ๐‘–๐‘— ๐‘– ๐‘—

and factoring out ๐‘ โˆ’ 1 ๐‘ from the second summation we arrive at
2 ๐œŽฮ 

1 1 = ๐‘ ๐‘ ๐‘ ๐œŽ๐‘–

2 ๐‘–
=1 ๐‘

โˆ’ 1 1 + ๐‘ ๐‘ ๐‘ โˆ’ 1 ๐‘ ๐‘ ๐œŒ๐‘–๐‘—

๐œŽ๐‘– ๐œŽ๐‘— = ๐‘–
=1 ๐‘— =1 ๐‘— โ‰ ๐‘–

1 2 ๐‘ โˆ’ 1 ๐œŽ + ๐œŽ๐‘–๐‘— ๐‘ ๐‘– ๐‘

The first and second terms in brackets are the average variance and average covariance respectively. As the number of assets tends to infinity then the risk tends to the average covariance therefore the individual risk of assets can be diversified away but the risk caused by covariances cannot be diversified away. An example is shown in (Elton, Gruber, Brown, & Goetzmann, 2007, p. 59) in Table 4.8 Effect of Diversification.

3.3.

Examining Efficient Portfolios

In this section we analyse the combinations of two assets with and without short selling. 3.3.1. No short selling Case 1: ๐œŒ12 = 1. Since the assets are perfectly correlated and from equations (3) and (4), we have ๐‘…ฮ  = ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 and
2 2 2 2 ๐œŽฮ  = ๐œ†1 ๐œŽ1 + 2๐œ†1 ๐œ†2 ๐œŽ1 ๐œŽ2 + ๐œ†2 ๐œŽ2 , 2

(7)

or ๐œŽฮ  = ๐œ†1 ๐œŽ1 + ๐œ†2 ๐œŽ2 . (8)

In order to examine the effect of varying the proportions of the assets, we let ๐œ†1 = ๐ด and ๐œ†2 = 1 โˆ’ ๐ด, where ๐ด โˆˆ 0,1 . Then equations (7) and (8) becomes ๐‘…ฮ  = ๐ด๐‘…1 + 1 โˆ’ ๐ด ๐‘…2 and ๐œŽฮ  = ๐ด๐œŽ1 + 1 โˆ’ ๐ด ๐œŽ2 . Starting off with equation (10), we solve for ๐ด to get ๐ด = ๐œŽฮ  โˆ’ ๐œŽ2 ๐œŽ1 โˆ’ ๐œŽ2 (10) (9)

13

and substituting this into equation (9) yields ๐‘…ฮ  = ๐œŽฮ  โˆ’ ๐œŽ2 ๐œŽฮ  โˆ’ ๐œŽ2 ๐‘…1 + 1 โˆ’ ๐‘… ๐œŽ1 โˆ’ ๐œŽ2 ๐œŽ1 โˆ’ ๐œŽ2 2 ๐‘…1 โˆ’ ๐‘…2 ๐‘…1 โˆ’ ๐‘…2 = ๐‘…2 โˆ’ ๐œŽ2 + ๐œŽ ๐œŽ1 โˆ’ ๐œŽ2 ๐œŽ1 โˆ’ ๐œŽ2 ฮ 

which is the equation of a straight line with gradient ๐‘…1 โˆ’ ๐‘…2 ๐œŽ1 โˆ’ ๐œŽ2 in the risk-return graph shown in Figure 1. From the graph, diversification cannot reduce risk of the portfolio and that proportionally higher risk leads to proportionally higher return. ๐‘…ฮ  ๐‘…2 ๐ด
= 0, ๐œ†1 = 0, ๐œ†2 = 1 ๐‘…

1 ๐œŽ1 ๐ด

= 1, ๐œ†1 = 1, ๐œ†2 = 0 ๐œŽ

2 ๐œŽ

ฮ 

Figure 1: Risk-return graph for two assets when ๐†๐Ÿ๐Ÿ = ๐Ÿ.

Case 2: ๐œŒ12 = โˆ’1. Since the assets are perfectly negatively correlated and from equations (3) and (4), we have ๐‘…ฮ  = ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 and
2 2 2 2 ๐œŽฮ  = ๐œ†1 ๐œŽ1 โˆ’ 2๐œ†1 ๐œ†2 ๐œŽ1 ๐œŽ2 + ๐œ†2 ๐œŽ2 , 2

or ๐œŽฮ  = ๐œ†1 ๐œŽ1 โˆ’ ๐œ†2 ๐œŽ2 . (11)

Again, in order to examine the effect of varying the proportions of the assets, we again let ๐œ†1 = ๐ด and ๐œ†2 = 1 โˆ’ ๐ด, where ๐ด โˆˆ 0,1 . Then we have ๐‘…ฮ  = ๐ด๐‘…1 + 1 โˆ’ ๐ด ๐‘…2 and ๐œŽฮ  = ๐ด๐œŽ1 โˆ’ 1 โˆ’ ๐ด ๐œŽ2 . We notice that when ๐œ†1 = ๐œŽ2 , ๐œŽ1 + ๐œŽ2 (13) (12)

14

risk is non-existent so ๐œŽฮ  = 0 therefore a risk free portfolio can be constructed by diversifying and choosing ๐œ†1 . From the above the return is ๐œŽ1 ๐‘…2 + ๐œŽ2 ๐‘…1 ๐œŽ1 + ๐œŽ2 . Solving for ๐ด in equation (13) to get ๐ด = ๐œŽฮ  โˆ’ ๐œŽ2 ๐œŽ1 + ๐œŽ2

and substituting this into equation (12) and manipulating the subsequent equation we arrive at ๐œŽ1 ๐‘…2 + ๐œŽ2 ๐‘…1 ๐‘…2 โˆ’ ๐‘…1 + ๐œŽ ๐œŽ1 + ๐œŽ2 ๐œŽ1 + ๐œŽ2 ฮ  ๐œŽ1 ๐‘…2 + ๐œŽ2 ๐‘…1 ๐‘…1 โˆ’ ๐‘…2 + ๐œŽ ๐œŽ1 + ๐œŽ2 ๐œŽ1 + ๐œŽ2 ฮ  ๐œŽ2 ๐œŽ1 + ๐œŽ2 , ๐œŽ2 if ๐ด > ๐œŽ1 + ๐œŽ2 if ๐ด < ๐‘…

ฮ  =

which the risk-return curve is two straight lines (see Figure 2). ๐‘…ฮ  ๐‘…2 ๐ด
= 0, ๐œ†1 = 0, ๐œ†2 = 1 Efficient frontier ๐‘…

1 ๐œŽ1 ๐ด

= 1, ๐œ†1 = 1, ๐œ†2 = 0 ๐œŽ

2 ๐œŽ

ฮ 

Figure 2: Risk-return graph for two assets when ๐œŒ12 = โˆ’1.

Case 3: ๐œŒ12 = 0 Since the assets are uncorrelated and their prices move independently, we have ๐‘…ฮ  = ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 and
2 2 2 2 ๐œŽฮ  = ๐œ†1 ๐œŽ1 + ๐œ†2 ๐œŽ2 . 2

As before, we have ๐‘…ฮ  = ๐ด๐‘…1 + 1 โˆ’ ๐ด ๐‘…2 and
2 2 2 ๐œŽฮ  = ๐ด2 ๐œŽ1 + 1 โˆ’ ๐ด 2 ๐œŽ2 .

(14)

Differentiating equation (14) yields
2 ๐œ•๐œŽฮ  2 2 = 2๐ด๐œŽ1 โˆ’ 2 1 โˆ’ ๐ด ๐œŽ2 = 0 ๐œ•๐ด

15

2 and solving for ๐ด yields the minimum value for ๐œŽฮ  when ๐ด

= ๐œŽ

2 , ๐œŽ1 + ๐œŽ2

which ๐ด is between 0 and 1 for all values of ๐œŽ1 and ๐œŽ2 . This implies that there is a combination of assets that has a lower risk than the risk of each asset on its own therefore this is preferred by investors who are risk averse. The risk-return curve is curved (see Figure 3). ๐‘…ฮ  ๐‘…2 ๐œŒ
12 = โˆ’1 ๐ด = 0, ๐œ†1 = 0, ๐œ†2 = 1 Efficient frontier ๐œŒ12 = โˆ’2 ๐œŒ12 = 0 ๐œŒ12 = 2 ๐ด = 1, ๐œ†1 = 1, ๐œ†2 = 0 ๐œŒ12 = 1
1 1 ๐‘…

1 ๐œŽ1 ๐œŽ

2 ๐œŽ

ฮ 

Figure 3: Risk-return graph for two assets for various values of ๐†๐Ÿ๐Ÿ.

General case The existence of minimum ๐œŽฮ  value depends on the values of ๐œŽ1 , ๐œŽ2 and ๐œŒ12 . The minimum value of ๐œŽฮ  always exist when ๐œŒ12 โ‰ค 0 and only exits for ๐œŒ12 โ‰ฅ 0 when ๐œŒ12 < ๐œŽ2 ๐œŽ1 . The riskreturn curve always lie between ๐œŒ12 = โˆ’1 and ๐œŒ12 = 1. Also the curve will always be convex and that concave curves cannot be possible. This is because the risk from a combination of assets cannot be equal to or greater than the risk of individual assets. The only case that occurs is perfect correlation. In reality, perfect correlation does not exist. ๐‘ต assets In the case of ๐‘ assets, the arguments in all of the above cases are the same than the case of two assets. Since lower correlation between assets will result in higher payoff from diversification, we will get a region of possible portfolios as well as the risk-return curve on the risk-return graph (see Figure 4). ๐‘…ฮ  ๐‘…N ๐ด
= 0, ๐œ†1 = 0, ๐œ†2 = 1 ๐‘…

1 ฯƒ1 ๐ด

= 1, ๐œ†1 = 1, ๐œ†2 = 0 ๐œŽ๐‘ ๐œŽ

ฮ 

Figure 4: Risk-return graph for ๐‘ต assets each with various values of ๐‘น, ๐ˆ and ๐†.

ฯƒ 16 1

3.3.2. With short selling Case 1: ๐œŒ12 = 1. The argument is the same as in the case without short selling but the risk-return curve extends from both sides indefinitely (see Figure 5). These extensions correspond to ๐ด < 0, where the investor short sells asset 1 and invests the proceeds into asset 2, and ๐ด > 1, where the investor short sells asset 2 and invests the proceeds into asset 1. ๐‘…ฮ  ๐œ†
1 < 0, ๐œ†2 > 1 ๐‘…

2 ๐ด

= 0 ๐‘…

1 ๐œ†
1 > 0, ๐œ†2 < 1 ๐ด

= 1 ๐œŽ

1 ๐œŽ

2 ๐œŽ

ฮ 

Figure 5: Risk-return graph for two assets when ๐†๐Ÿ๐Ÿ = ๐Ÿ with short selling.

Case 2: ๐œŒ12 = โˆ’1. Again, the argument is identical to the one without short selling but with two straight lines extended (see Figure 6). Risk-free portfolio can still be achieved through perfect hedging of assets, though it is difficult to do so in reality. ๐‘…ฮ  ๐‘…2 ๐ด
= 0 Efficient frontier ๐œ†

1 < 0, ๐œ†2 > 1 ๐‘…

1 ๐œŽ1 ๐ด

= 1 ๐œŽ

2 ๐œŽ

ฮ  ๐œ†
1 > 0, ๐œ†2 < 1

Figure 6: Risk-return graph for two assets when ๐†๐Ÿ๐Ÿ = โˆ’๐Ÿ with short selling.

Case 3: โˆ’1 โ‰ค ๐œŒ12 โ‰ค 1 Again, the case is identical to the one without short selling, except that curves are extended in both directions.

17 ๐‘…

ฮ  ๐‘…N ๐ด
= 0, ๐œ†1 < 0, ๐œ†2 > 1 ๐‘…

1 ฯƒ1 ๐ด

= 1 ๐œ†

1 > 0, ๐œ†2 < 1 ๐œŽ๐‘ ๐œŽ

ฮ 

Figure 7: Risk-return graph for ๐‘ต assets each with various values of ๐‘น, ๐ˆ and ๐† between -1 and 1. ๐‘ต

assets with short selling

ฯƒ1

The differences between this case and the one without short selling is that the region of possible portfolios is in a convex shape, as well as the risk-return curve, and there are portfolios with returns are arbitrarily large, due to the fact that selling of assets with low returns and purchasing of assets with high returns with the proceeds takes place. Therefore, the efficient frontier has no upper bound. 3.3.3. Including Riskless Assets and Riskless Borrowing and Lending Suppose that a portfolio has one riskless asset ๐‘†0 with return ๐‘…0 and one risky asset ๐‘†1 . This asset has no variance and correlation with any of the risky assets so that ๐‘…0 = ๐‘…0 and ๐œŽ0 = ๐œŒ0๐‘– = 0. Suppose that a portfolio has ๐œ†0 invested in the riskless asset and the rest ๐œ†1 = 1 โˆ’ ๐œ†0 in the risky asset. Then the portfolio is ฮ  = ๐œ†0 ๐‘†0 + ๐œ†1 ๐‘†1 and we have ๐‘…ฮ  = ๐œ†0 ๐‘…0 + ๐œ†1 ๐‘…1 = 1 โˆ’ ๐œ†1 ๐‘…0 + ๐œ†1 ๐‘…1 and
2 2 2 ๐œŽฮ  = ๐œ†1 ๐œŽ1 ,

(15)

or ๐œŽฮ  = ๐œ†1 ๐œŽ1 . Solving equation (16) in terms of ๐œ†1 yields ๐œ†1 = and substituting this into equation (15) yields ๐œŽฮ  ๐œŽ1 (16)

18 ๐‘…

ฮ  = ๐œŽ

ฮ  ๐œŽฮ  ๐‘…1 โˆ’ ๐‘…0 ๐‘…0 + ๐‘…1 = ๐‘…0 + ๐œŽฮ  ๐œŽ1 ๐œŽ1 ๐œŽ1 ๐œŽฮ  ๐œŽฮ  ๐‘…1 โˆ’ ๐‘…0 1+ ๐‘…0 โˆ’ ๐‘…1 = ๐‘…0 โˆ’ ๐œŽฮ  ๐œŽ1 ๐œŽ1 ๐œŽ1 1โˆ’

if ๐œ†1 > 0 , if ๐œ†1 < 0

which the risk-return curve is a pair of straight lines (see Figure 8). When ๐œ†0 < 0 and ๐œ†1 > 1, the investor is borrowing from the bank at ๐‘…0 and investing the loan in more risky asset. Conversely, when ๐œ†0 > 1 and ๐œ†1 < 0 , short selling of risky asset takes place and the proceeds are deposited in the bank account to earn interest at ๐‘…0 . ๐‘…ฮ  ๐‘…1 ๐œ†
0 = 1, ๐œ†1 = 0
โ€ฒ ๐‘…0 ๐œ†

0 < 0, ๐œ†1 > 1
โ€ฒ ๐‘†1 ๐‘†

1 ๐œ†0 = 0, ๐œ†1 = 1 ๐‘…

0 ๐œŽ

1 ๐œŽ

ฮ  ๐œ†
0 > 1, ๐œ†1 < 0

Figure 8: Risk-return graph for riskless and risky assets with short selling.

Examining Figure 8 carefully, if the investor can borrow but cannot lend at the riskless rate, then one would invest all of the funds in the risky asset to earn ๐‘…1 . Additionally, borrowing may take place to invest in more of the risky asset to earn higher return than ๐‘…1 . If one had some funds invested in the riskless asset then one would sell that asset and invest the proceeds in the risky asset. If one cannot borrow but can lend at the riskless rate, then one would invest all funds in the riskless asset to earn interest at that rate. Short selling of the risky asset does not take place because the rational investor would not take on higher risk to earn lower return than the riskless rate. If the lending and borrowing rates are different then the efficient frontier would โ€ฒ change. If the lending rate is below the borrowing rate ๐‘…0 then the efficient frontier would โ€ฒ start from ๐‘…0 going through ๐‘†1 and ๐‘†1 onto the line with intercept at the borrowing rate. In the case of one riskless asset and ๐‘ risky assets the above arguments are essentially the same.

3.4.

Optimal Portfolio

In this section we explain the basic ideas concerning optimal portfolios. If all investors have the same expectations, then they face the same efficient frontier of risky assets. Also if they borrow and lend at the sane risk-free interest rate, then the efficient frontier is just a straight line with intercept ๐‘…0 in ๐‘…0 , ๐œŽ space. Therefore there is a unique point that lies within the opportunity set of ๐‘ risky assets and the efficient frontier comprising a portfolio of ๐‘ risky assets and one riskless asset. This point also maximises the gradient of the line of possible risky and riskless portfolios. Hence that a rational investor chooses to invest in a combination of riskless asset and portfolio of risky assets at a point on the efficient frontier tangent to the line of maximum gradient, which that line is called the Capital Market Line (CML). This is because other mixtures of riskless and portfolio of risky assets lie on the efficient frontier which would result in a lower returns than the point tangent to the CML. Also there are no combinations of 19

riskless and risky assets with higher returns proportional to risk is possible since the CML would lie above the efficient frontier. From an assumption of CAPM (see Appendix A.1), all investors are rational, if one invests at a point of the CML, others should do the same so that they will achieve the same rate of return proportional to risk. If all investors invest on the CML, then all of them should have the same portfolio comprising the same proportions of risky assets. The risky part of the portfolio is called the Optimum Portfolio, which is point ๐‘€, and gradient of the CML is called the Market Price of Risk because of the above reason. The Market Price of Risk measures the amount of profit over the riskless rate for accepting a certain risk level (see Figure 9). The point the investor chooses on the CML depends on the amount of risk they would tolerate. Looking at the following cases, if the investor wants to; ๏‚ท ๏‚ท ๏‚ท Invest only in risky assets, then no proportion of their wealth is invested in the riskless asset, that is ๐œ†0 = 0. Increase their return and risk they will increase their holdings in the risky asset by borrowing from the bank. Reduce risk and return by depositing some of their wealth in the bank, that is ๐œ†0 > 0. ๐‘…ฮ  ๐‘…P ๐‘€ ๐‘…0
Efficient Frontier Optimum Portfolio

Capital Market Line

Inefficient portfolios ๐œŽ

P ๐œŽ

ฮ 

Figure 9: Optimal portfolio, efficient frontier and CML.

Adapting the expected return and variance expressions from previous section, the proportions of the portfolio add up to one, that is ๐œ†0 + ๐œ†๐‘– = 1, and the CML is defined by ๐‘…ฮ  = ๐‘…0 + ๐‘…๐‘ƒ โˆ’ ๐‘…0 ๐œŽฮ  , ๐œŽ๐‘ƒ

the risk-return curve is two straight lines. Hence, the portfolio can be optimised by maximising the gradient of the CML, in other words finding the portfolio with the greatest ratio of excess return to risk, which we will show in the next section.

3.5.

Finding the Optimal Portfolio

3.5.1. Short selling with Riskless Borrowing and Lending In this case we maximise the objective function ๐œƒ = subject to the constraint ๐‘…๐‘ƒ โˆ’ ๐‘…0 ๐œŽ๐‘ƒ

20 ๐‘ ๐œ†๐‘–

= 1. ๐‘–
=1

There are numerous methods for solving the above problem such as Lagrangian multipliers and substitution of constraint into the objective function and maximise it without constraints. In this case we will use the latter method. First we write ๐‘…0 as ๐‘
๐‘ ๐‘…

0 = 1๐‘…0 = ๐‘–
=1 ๐œ†๐‘–

๐‘…0 = ๐‘–
=1 ๐œ†๐‘–

๐‘…0 .

(17)

Then we substitute equation (17) and stating the expected return and standard deviation of return in the general form to get ๐œƒ = ๐‘
๐‘–=1 ๐œ†๐‘– ๐‘ 2 2 ๐‘–=1 ๐œ†๐‘– ๐œŽ๐‘– ๐‘…๐‘–

โˆ’ ๐‘…0 ๐‘
๐‘–=1 ๐‘ ๐‘— =1 ๐‘— โ‰ ๐‘–

+ ๐œ†๐‘–

๐œ†๐‘— ๐œŽ๐‘–๐‘—

1 2

.

(18)

Using the chain and product rules, we differentiate equation (18) with respect to each variable ๐œ†๐‘– and set it to zero to get a system of simultaneous equations: ๐œ•๐œƒ =0 ๐‘‘๐œ†1 ๐œ•๐œƒ =0 ๐‘‘๐œ†2 โ‹ฎ ๐œ•๐œƒ = 0, ๐‘‘๐œ†๐‘ where ๐œ•๐œƒ = โˆ’ ๐›พ๐œ†1 ๐œŽ1๐‘– + ๐›พ๐œ†2 ๐œŽ2๐‘– + โ‹ฏ ๐›พ๐œ†๐‘– ๐œŽ๐‘–2 + โ‹ฏ + ๐›พ๐œ†๐‘โˆ’1 ๐œŽ๐‘โˆ’1๐‘– + ๐›พ๐œ†๐‘ ๐œŽ๐‘๐‘– + ๐‘…๐‘– โˆ’ ๐‘…0 = 0, ๐‘‘๐œ†๐‘– where ๐›พ is a constant. Then we use a trick which involves modifying the derivative. Let ๐‘๐‘– = ๐›พ๐œ†๐‘– . Since we know that ๐œ†๐‘– is a proportion of each asset then the ๐‘๐‘– is proportional to this proportion. In order to solve for ๐œ†๐‘– after solving for ๐‘๐‘– , we divide each ๐‘๐‘– by the sum of ๐‘๐‘– . We substitute ๐‘๐‘– for ๐›ฝ๐›พ and moving the covariance terms to the right-hand side to obtain ๐‘…๐‘– + ๐‘…0 = ๐‘1 ๐œŽ1๐‘– + ๐‘2 ๐œŽ2๐‘– + โ‹ฏ ๐‘๐‘– ๐œŽ๐‘–2 + โ‹ฏ + ๐‘๐‘โˆ’1 ๐œŽ๐‘โˆ’1๐‘– + ๐‘๐‘ ๐œŽ๐‘๐‘– . There is one equation similar to the above for each ๐‘– so we solve the following system of simultaneous equations: ๐‘…1 + ๐‘…0 = ๐‘1 ๐œŽ11 + ๐‘2 ๐œŽ12 โ€ฆ + ๐‘๐‘ ๐œŽ1๐‘ ๐‘…2 + ๐‘…0 = ๐‘1 ๐œŽ12 + ๐‘2 ๐œŽ22 โ€ฆ + ๐‘๐‘ ๐œŽ2๐‘ โ‹ฎ ๐‘…๐‘ + ๐‘…0 = ๐‘1 ๐œŽ1๐‘ + ๐‘2 ๐œŽ2๐‘ โ€ฆ + ๐‘๐‘ ๐œŽ๐‘๐‘ .

(19)

We can determine optimum investment amount in each asset by solving equation (19). The solution is the optimum proportions of investment in asset ๐‘– is ๐œ†๐‘– , where

21 ๐œ†๐‘–

= An example is shown in Appendix A.3. ๐‘๐‘–

. ๐‘ ๐‘–=1 ๐‘๐‘–

3.5.2. Short selling without Riskless Borrowing and Lending In this case, we assume that there are different lending and borrowing rates. The method presented in the previous section can be adapted for this case. We first arbitrarily choose a riskless rate ๐‘…01 and then find the optimal portfolio which corresponds to that rate. Then we choose another riskless rate ๐‘…02 and again find the optimal portfolio which corresponds to the second rate. We repeat this process again and again until the full efficient frontier is determined (see Figure 10). Portfolio A would be selected if the riskless rate is ๐‘…01 . Portfolio B would be chosen if riskless rate is ๐‘…02 and C is selected if the rate is ๐‘…03 . Note that the efficient frontier can be determined if more riskless rates are taken into account. ๐‘…ฮ  ๐ถ ๐‘…03 ๐‘…02 ๐‘…01 ๐œŽฮ 
Figure 10: Optimal portfolios A, B and C for riskless rates ๐‘น๐ŸŽ๐Ÿ , ๐‘น๐ŸŽ๐Ÿ and ๐‘น๐ŸŽ๐Ÿ‘ respectively which traces out the efficient frontier. ๐ด ๐ต

3.5.3. No Short selling with Riskless Borrowing and Lending This problem is very similar to the case of riskless lending and borrowing with short selling. Again, at the point on the CML tangent to the efficient frontier we maximise the gradient of the CML. However, assets cannot be held in negative quantities. Therefore we maximise the objective function ๐œƒ = subject to the constraints ๐‘ ๐‘…๐‘ƒ

โˆ’ ๐‘…0 ๐œŽ๐‘ƒ ๐œ†๐‘–

= 1 ๐‘–
=1

and ๐œ†๐‘– โ‰ฅ 0 โˆ€๐‘–.

22

Unfortunately, we cannot solve the problem using the method described in section 3.5.1 because it is a mathematical programming problem, due to the inequality constraint on ๐œ†๐‘– . This problem is called a quadratic programming problem because the objective function contains ๐œ†2 and ๐œ†๐‘– ๐œ†๐‘— terms, hence it is non-linear. Therefore quadratic programming is used ๐‘– to solve the problems of this type, which will explain in Appendix A.2. 3.5.4. No Short selling and Riskless Borrowing and Lending This problem is similar to the one in the previous section. In this case, we minimise the portfolio risk subject to the constraints from the previous section and the return being some level. Mathematically, we minimise ๐‘
2 ๐œŽฮ  ๐‘ ๐‘

= ๐‘–
=1 ๐œ†

2 ๐œŽ๐‘–2 ๐‘–

+ ๐‘–
=1 ๐‘— =1 ๐‘— โ‰ ๐‘– ๐œ†๐‘–

๐œ†๐‘— ๐œŽ๐‘–๐‘—

subject to the constraints ๐‘ ๐œ†๐‘–

= 1 ๐‘–
=1 ๐‘ ๐œ†๐‘–

๐‘…๐‘– = ๐‘…ฮ  ๐‘–
=1

and ๐œ†๐‘– โ‰ฅ 0 โˆ€๐‘–. From section 3.5.3, we assume that the riskless rate exists and varying this determines the efficient frontier. Since the above problem is quadratic programming problem, we cannot solve the problem using the method described in section 3.5.1. 3.5.5. Including additional constraints Any requirement that can be formulated as a linear function of investment weight can be imposed on the solution. For example, choosing optimal portfolios given that the dividend yield on these is at some minimum, for example 4%. Let ๐‘‘๐‘– be dividend yield on asset ๐‘– and ๐ท be target dividend yield. Then this requirement can be imposed as a constraint ๐‘ ๐œ†๐‘–

๐‘‘๐‘– = ๐ท ๐‘–
=1

on any of the problems described in sections 3.5.1 to 3.5.4. In general, if inequality constraints are imposed on the problem then it has to be solved through the use of quadratic programming. Otherwise, the method of a system of simultaneous equations is used.

23

4. Behavioural Finance
This section is based on (Montier, 2002, pp. 1-15), (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 485-492) and (Thaler, 1985). We included this section in this report because the Contrarian investment strategy is designed to minimise the investorโ€™s tendency to make choices based on facts instead of their perceptions and preferences. In other words, the strategy helps the investors behave more rationally. Behavioural Finance is the study of psychology the investor and their subsequent effects on financial markets. This field explores the reasons private and professional investors, traders and other participants make predictable and systematic errors. In other words, the following alters investorsโ€™ perception regarding reality, decision making more bias towards certain choices and preferences: ๏‚ท Over-optimism is where investors tend to exaggerate their own abilities, for example, regarding selecting the right stocks to earn high returns on their portfolio. Investors are often deceived by randomness. Illusion of control is where investors believe that they are in control of a situation more often than they are. Self-attribution bias is reference to situations where good outcomes are attributed to good skill whilst bad outcomes are attributed to lack of luck. Hence, illusion of control and self-attribution bias tends to lead investors to be over-optimistic. There are other psychological biases which also lead to over-optimism. Over-confidence is where investors tend to believe that their judgements of a situation are correct more often than they are. This and over-optimism lead to investors to overestimate their knowledge, underestimate risk and exaggerate their ability to control the situation. Cognitive dissonance is the mental conflict that investors experience when they are presented with evidence which confirms that their beliefs or assumptions are incorrect. This leads to self-denial and they jump through mental hurdles to reduce or avoid mental inconsistencies. Confirmation bias is the investorโ€™s need to discover information that agrees with their view. Information which conflict with their view is ignored whilst ones that support their view is over weighted. Conservatism bias is a tendency to cling tenaciously to a view or forecast. When this happens, most investors find it very difficult to change their view. When change occurs it takes place very slowly, which causes under-reaction to events. Anchoring is a tendency to rely heavily on a single piece of information when making decisions. Representativeness heuristic is a rule of thumb (heuristic) that refers a tendency to evaluate the likelihood of something is with reference to how closely it resembles to rather than using probabilities. For example, an investor notices that a company has many periods of good earnings and, from the law of small numbers, leads to their belief that the company is one with high earnings growth, and hence likely to continue to deliver high earnings growth in the future. If the company has some average earnings in some periods then its earnings would appear average in their view even if there are many periods of high earnings. Availability bias is a heuristic which the decision makers assess the frequency of class or probability of an event based on the ease of examples can be brought to mind. Ceteris paribus, common events are easier to come up with than rare events. For example, an investor is faced with a choice of thousands of stocks. Their choice is affected by the press coverage of stocks, such as record earnings of a company and tipping. Therefore, one or more stocks featured will be selected. Another example is that events which occurred recently are more easily remembered than the events which occurred in the distant past.

๏‚ท

๏‚ท

๏‚ท ๏‚ท ๏‚ท ๏‚ท

๏‚ท

24

๏‚ท ๏‚ท

๏‚ท

Ambiguity aversion is where the investor has preference for the known over unknown, for example, they choose a stock they have knowledge of instead of another they have no knowledge of. Narrow framing (mental accounting) is the failure of the investor to consider all parts of the portfolio, where each part is already allocated, separate and non-transferrable. For example, an investor allocates their money into two separate halves for different investment purposes. The first and second parts are for investing in stocks in FTSE 100 and FTSE AIM indices. The investor already has invested one stock in each index and decides to diversify to reduce the portfolio risk but only considers FTSE AIM stocks. Therefore, their portfolio is not sufficiently diversified and the portfolio risk is higher than it should be. Herd behaviour where investors follow market trends. For example, investors bought Internet stocks in the late 1990s, due to the forecasted high earnings growth. The stock prices collapsed in the early 2000s when that did not materialise and most made huge losses. One or more of the above can cause herd behaviour.

There are other factors which affect the investorโ€™s psychology other than those mentioned. However, these are beyond the scope and relevance to this report. The above concepts will provide enough information in order to build and run the portfolio using the Contrarian investment strategy successfully without making the above errors.

5. Investment Strategies
5.1. Contrarian Investment Strategy

This section is based on (Dreman, 1982) and (Dreman, 1998). From (Investopedia ULC, 2009), Contrarian Investment Strategy is one that involves investing in an approach which goes against prevailing market trends by buying stocks that are perform poorly, that most investors do not want to hold and sell, and sell them when they perform well, when other investors buy in large quantities. The first strategy was outlined in (Dreman, 1980) in an attempt to prevent investors from making the same predictable mistakes as outlined in the previous section. This strategy involved purchasing stocks with low price to earnings (P/E) ratios and going against the market trend. The strategy was refined in (Dreman, 1982). In (Dreman, 1998), the strategy was expanded considerably to purchasing stock with low price to book value (P/BV), low price to cash flow (P/CF) and high dividend yield ratios, as well as P/E below their market values (or averages). There are other variants which are devised by others. The strategies have been researched by many, such as (Chan K. C., 1988) and (Gregory, Harris, & Michou, 2001). They generally indicate that stocks with low P/E, P/BV, P/CF and dividend yield ratios consistently on average earn above-market returns, capital and income, than their high value counterparts. In this report, we will use the strategy based on (Dreman, 1998, pp. 160-173 and 406-410). Mandatory: 1. Lowest possible P/E, P/BV and P/CF for a stock, while using the P/E as a core ratio. These ratios have to be below market value for the industry or index. The companies have to lie within the two lowest quintiles (lowest 40% of companies based on the ratios mentioned). 2. Dividend yield should be highest possible, which has to be above market yield. Companies have to lie within two highest quintiles. 3. Dividend yield which is above average which the company can sustain and increase. Dividend cover of 2 or more is desirable. The value must be more than 1.5. 25

4. Current assets should be twice the amount of current liabilities is desirable, that is current ratio of 2:1. This value varies from industry to industry. Minimum ratio is 1:1. Optional: 1. Many favourable financial ratios as possible; high return on capital employed (ROCE) and return on assets (ROA), low price to sales (P/S) and PEG less than1. For ROCE, values over 20% are attractive in general. For the latter two, values greater than 15 are unsuitable. 2. Earnings per share (EPS) should be higher than FTSE 100 in the previous year. 3. Earnings should always lean on the conservative side. Selling: Sell a stock when itโ€™s P/E, P/BV, P/CF or dividend yield approaches that of the overall market, regardless of how favourable prospects may appear. Replace it with another contrarian stock. We will abide by the following points in order to avoid the main pitfalls of behavioural finance: 1. Favoured stocks underperform the market, while out-of-favour companies outperform the market, but the reappraisal often happens slowly, even glacially. 2. Buy only contrarian stocks because of their superior performance characteristics. 3. Buy the least expensive stocks within an industry, as determined by the, four contrarian strategies, regardless of how high or low the general price of the industry group. 4. Look beyond obvious similarities between a current investment situation and one that appears equivalent in the past. Consider other important factors that may result in a markedly different outcome. 5. Don't be influenced by the short-term record of a money manager, broker, analyst, or advisor, no matter how impressive; do not accept cursory economic or investment news without significant substantiation. 6. Don't rely solely on the short term. Take into account the long term the prior probabilities of profit or loss. 7. It is far safer to project a continuation of the psychological reactions of investors than it is to project the visibility of the companies themselves. (Very essential) 8. Political and financial crises lead investors to sell stocks. This is precisely the wrong reaction. Buy during a panic, don't sell. (Very essential) 9. In a crisis, carefully analyse the reasons put forward to support lower stock prices โ€“ more often than not they will disintegrate under scrutiny. 10. a. Diversify extensively. No matter how cheap a group of stocks looks, you never know for sure that you aren't getting a clinker. b. In a crisis, these criteria get dramatically better as prices plummet (that is P/E, P/BV and P/CF falls and dividend yields rises), markedly improving your chances of a big score. 11. A given in markets is that perceptions change rapidly. Optional: Small-cap investing: 1. Buy companies that are strong financially (normally no more than 60% debt in the capital structure for a manufacturing firm). 2. Buy companies with increasing and well-protected dividends that also provide an abovemarket yield. 3. Pick companies with above-average earnings growth rates. 26

4. Diversify widely, particularly in small companies because these issues have far less liquidity. A good portfolio should contain about twice as many stocks as an equivalent large-cap one.

5.2.

Value Investing

This section is based on (Graham, Zweig, & Buffett, The Intelligent Investor: The Definitive Book on Value Investing, 2003). Value investing involves investing in companies that are worth less than their intrinsic value or underpriced, which is usually measured by P/BV. The strategy was first published in (Graham & Dodd, 1934) and was refined in subsequent editions ending with the final form of the strategy in (Graham, Dodd, & Cottle, 1963). The strategy was simplified in (Graham, 1973). There has been studied in various sources, such as (Capaul, Rowley, & Sharpe, 1993) and (Chan & Lakonishok, 2004). These usually indicate that P/E and P/BV ratios have higher returns than their high value counterparts, hence, support that the strategy on average produces above market returns. We will use the strategy based on (Graham, Zweig, & Buffett, 2003). In general: 1. The size of the company must be adequate, that is, in terms in turnover which must be at least ยฃ30m. 2. Sufficiently strong financial condition. Current ratio must be at least equal to 2. 3. Continued dividends for the previous 5 years. 4. No earnings loss for the previous 5 years. 5. 4 year EPS growth must be at least 1/3. 6. P/BV must be at most 1.5. 7. Average P/E must be at most 15 for the previous 3 years. For defensive strategy: 1. As in general, the size of the company must be adequate. However, turnover which must be at least ยฃ70m for manufacturers and at least ยฃ30m for utilities. 2. Net current assets must be at least twice the amount of current liabilities (Current ratio of at least equal to 2). For manufacturers, long term debt (those due in more than 1 year) should be less than net current assets. For utilities, debt should less than twice net asset value. 3. Continued dividends for the previous 5 years. 4. No earnings loss for the previous 5 years. 5. 4 year EPS growth must be at least 1/3. 6. P/BV must be at most 1.5. 7. Average P/E must be at most 15 for the previous 3 years. For enterprising strategy, as in general but with the following: 1. Current assets at least 1.5 times current liabilities (Current ratio of 1.5). For manufacturers, debt must be at most 110% of net current assets. Replaces criteria 2. 2. No earnings loss for the previous 5 years. 3. Some dividends paid. Replaces criteria 3. 4. Market capitalisation less than 120% of net tangible assets. Selling: Only sell when the situation changes to the extent that the company no longer satisfies the above criteria. 27

Regarding requirement 1 for both general and defensive strategies, (Graham, Zweig, & Buffett, 2003) does not define this requirement clearly in terms of pounds sterling (pound or ยฃ). Therefore, a reasonable value is used based on the exchange rate between the US dollar and the pound in the year the book was published.

5.3.

Zulu Principle

This section is based heavily on (Slater, 1992) and (Slater, 1998). The Zulu Principle is about dedicating a disproportionate amount of effort to become an expert within a narrow subject though learning. From (Slater, 1992), this name is applied to the investment strategy because the Jim Slaterโ€™s wife read an Readerโ€™s Digest article on the Zulu people and subsequently she had more knowledge than him within a few minutes. He thought that if she carried on studying on the same subject and lived in a Zulu village from a fixed period of time she would become a leading expert. The investor can be very successful by applying the same method to the stock market. One can become an expert in blue chips, growth companies, turnarounds and/or cyclical companies. There are sources, such as (Bary, 1997) and (O'Shaughnessy, 2005), prove that low P/E, P/BV, P/S, price to earnings growth (PEG) and price to research (P/R) ratios, high values of relative strength and EPS produce higher returns than their opposite counterparts. Therefore, the strategy produces above market returns on average. The strategy we will be using is extracted from (Slater, 1992) and (Slater, 1998): Mandatory: 1. PEG less than 1 for large companies and less than 0.75 for smaller companies. Preferably PEG below 0.66. PEG works best if companies have EPS growth rate between 15%-30 and P/E between 12 and 20. 2. P/E less than 20. 3. Strong cash flow and in particular cash flow per share greater than EPS for last reported year and average for the last five years. 4. Low gearing, preferably below 50%. Positive cash balances are better. 5. High relative strength (RS) in the last 12 months with high relative strength in previous one month or three months. Alternatively, relative strength index (RSI) of at least 83 1/3%. 6. Strong competitive advantage, for example, strong brand names, patents, copyrights, legal monopolies, dominance in an industry and established position in a niche market. 7. No active selling by a group of directors. Highly desirable (optional): 1. Increasing EPS, preferably a company can clone its activities, for example Ikea opening a new store using a successful business formula. 2. A number of directors buying new shares. 3. Market capitalisation between ยฃ30m-ยฃ250m. 4. Dividend yield. Bonus (optional): 1. Low P/S. 2. Something new, for example, new events in the industry, new acquisitions, new product/service, new management or director. 3. Low P/R. This value must be at most 10. 4. Reasonable asset position. That is low P/BV at most 1/3.

28

We note that very few companies meet all of the above mandatory and highly desirable requirements. A company which lacks one of these requirements then compromise a little. Regarding mandatory requirement 5, no number for relative strength is stated in (Slater, 1998) and relative strength is not available on the Digital Look website. Therefore we use the relative strength index (RSI) which is defined as ๐‘…๐‘†๐ผ = 100 โˆ’ 100 . 1 + ๐‘…๐‘†

The range for the above is 0 โ‰ค ๐‘…๐‘†๐ผ โ‰ค 100. Higher values indicate that better performance of the stock. An arbitrary value of ๐‘…๐‘† = 5 is used instead to calculate the value for RSI.

6. Evaluating Investment Performance
This section is based on (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 632-658), (Kevin, 2007, pp. 197-211) and (Maginn, Tuttle, McLeavey, & Pinto, 2007, pp. 720-723). In this section we will use various measures to measure investment performance of the portfolios, constructed using Portfolio Theory and investment strategies outlined in this report.

6.1.

Rate of Return

The rate of return ๐‘Ÿฮ  in percent over one period is the difference of the net asset value (NAV), dividend payments and capital gains ๐ถ๐‘ก . Then the return of the portfolio is ๐‘Ÿฮ  = ๐‘๐ด๐‘‰๐‘ก โˆ’ ๐‘๐ด๐‘‰๐‘กโˆ’1 + ๐›ฟ๐‘ก + ๐ถ๐‘ก ๐‘๐ด๐‘‰๐‘กโˆ’1

where ๐‘๐ด๐‘‰๐‘กโˆ’1 and ๐‘๐ด๐‘‰๐‘ก are NAVs at the start and end of the holding period respectively, ๐›ฟ๐‘ก and ๐ถ๐‘ก are dividend payments and capital gains during the holding period. Portfolios may have different returns because they have different compositions and proportions of assets. Hence, the returns have to be adjusted before comparisons can be made. In this report the NAV at the end of the holding period will be the sale value of the shares held at the time after transaction costs.

6.2.

Risk Adjusted Returns

Sharpe Ratio The Shape ratio (SR) was developed in (Sharpe, Mutual Fund Performance, 1966). This is the ratio of the difference of portfolio return and riskless return to risk. The expression is defined as ๐‘†๐‘… = ๐‘…ฮ  โˆ’ ๐‘…0 ๐œŽฮ 

where ๐‘…ฮ  is the realised return on the portfolio after transaction costs. Higher values indicate that the portfolio is performing better than the ones with lower values.

29

Treynor Ratio The Treynor ratio (TR) was devised in (Treynor, 1965). This is the ratio of the difference of portfolio return and riskless return to volatility (or systematic risk) measured by beta. The equation is defined as ๐‘‡๐‘… = ๐‘…ฮ  โˆ’ ๐‘…0 ๐›ฝฮ 

where ๐›ฝฮ  is the portfolio beta, which is covariance of asset ๐‘– with the market portfolio divided by the variance of the market portfolio. This measure the correlation compared to the whole market. Similarly for the Treynor ratio, higher values indicate that the portfolio is performing better than the ones with lower values. The values calculated will be different for both ratios because the Sharpe ratio uses standard deviation, while the Treynor ratio uses the systematic risk measured. If the portfolio is well diversified then the latter would be appropriate to use. Otherwise, the former is appropriate to use. Jensen Ratio The Jensen ratio (JR) was developed in (Jensen, 1968). This measures the difference between the actual return on the portfolio and its expected return at a given level of risk. This indicates the portfolio managerโ€™s skill level. The Security Market Line (SML) is ๐‘…ฮ  = ๐‘…0 + ๐‘…๐‘€ โˆ’ ๐‘…0 ๐œŽ๐‘–๐‘€ = ๐‘…0 + ๐‘…๐‘€ โˆ’ ๐‘…0 ๐›ฝฮ  2 ๐œŽ๐‘€

where ๐‘…๐‘€ is the market index expected return. The formula for JR is ๐›ผฮ  = ๐‘…ฮ  โˆ’ ๐‘…ฮ  where ๐›ผฮ  and ๐‘…ฮ  are the differential and actual returns respectively. If ๐›ผฮ  > 0 , then it indicates that the portfolio performance is better than the market and the excess return is contributed to good management skills. If ๐›ผฮ  = 0, then it indicates neutral performance since the portfolio performance is identical to an unmanaged portfolio with a buy and hold strategy or the market. If ๐›ผฮ  < 0, then the portfolio performance is worse than that of the market or a randomly selected portfolio with the same level of risk. Fama's Net Selectivity The Fama's net selectivity (FNS) was first defined in (Fama, 1972). This measure provides a breakdown of the portfolio's performance into different components. The total return on a portfolio is defined as Total return = Risk free return + Excess return. Excess return is defined as Excess return = Risk premium + Return from stock selection where risk premium is reward for taking on risk. Since no portfolio would be fully diversified in reality, it would have systematic risk and small amount of diversifiable risk. Therefore we have

30

Risk premium = Return from systematic risk + Return from diversifiable risk. Thus the portfolio's total return can be decomposed into four components. We have Total return = Risk free return + Return from systematic risk + Return from diversifiable risk + Return from pure selectivity. We can represent the above mathematically as ๐‘…ฮ  = ๐‘…0 + ๐‘…๐‘Ž + ๐‘…๐‘ + ๐‘…๐‘ where ๐‘…๐‘Ž is the return from market risk, which is defined as ๐‘…๐‘Ž = ๐‘…๐‘€ โˆ’ ๐‘…0 ๐›ฝฮ  , ๐‘…๐‘ is the return from diversifiable risk, which is ๐‘…๐‘ = ๐‘…๐‘€ โˆ’ ๐‘…0 and ๐‘…๐‘ is the return from pure selectivity, which is ๐‘…๐‘ = ๐‘…ฮ  โˆ’ ๐‘…0 + ๐‘…๐‘Ž + ๐‘…๐‘ . We can simplify equation (20) to get ๐น๐‘๐‘† = ๐‘…ฮ  โˆ’ ๐‘…0 + ๐œŽฮ  ๐‘… โˆ’ ๐‘…0 . ๐œŽ๐‘€ ๐‘€ ๐œŽฮ  โˆ’ ๐›ฝฮ  , ๐œŽ๐‘€ (20)

This basically measures the actual return of the portfolio and the return mandated by its total risk. If ๐น๐‘๐‘† > 0 then the portfolio has achieved the return equal or above its expected return justified by its total risk due to good stock selection. If ๐น๐‘๐‘† < 0 then the portfolio has not achieved its expected return justified by its total risk due to poor stock selection.

7. Plan of Experiment
We will go through the details of the experiment. We will construct three optimal portfolios and a portfolio using each of the strategies outlined in this report, starting with the first optimal portfolio and portfolio using the Contrarian Investment strategy. Then we run them for two weeks and compare the performance of both to find out which produces the highest return during that period. We repeat the process with the second optimal portfolio and Value Investing portfolio. Then we repeat again with the third optimal portfolio and the Zulu Principle portfolio. The timetable of the experiment is shown below.
Table 1: Timetable of experiment showing each of the optimal portfolios going to be compared with each of the portolios constructed using the investment strategies Period 1 2 3 Start Date 9/3/2009 23/3/2009 6/4/2009 End Date 20/3/2009 3/4/2009 17/4/2009 Optimal Portfolio 1 2 3 Type of strategy used for portfolio Contrarian Investment Value Investing Zulu Principle

We will use the following assumptions in order to make the experiment realistic as possible: 31

1) There are transaction costs when buying and selling stocks. a. When buying stocks on the LSE, the following costs will be incurred; i. Broker commission of ยฃ10. ii. Stamp duty of 0.5% on the value of the stock excluding commission iii. Panel of Takeovers and Mergers (PTM) levy of ยฃ1 per transaction if the value of stock is over ยฃ10,000. The levy is ยฃ1.50 per transaction if the value is greater than ยฃ15,000. These values are from (The Panel on Takeovers and Mergers, 2009). b. When selling stocks, only ai and aiii apply. 2) There are capital gains tax and income tax on dividends, though the latter is very unlikely during each two week period. 3) Bid/offer spread. This is the difference between the buying (offer) and selling (bid) price. 4) Only integer quantities of stock can be bought or sold. 5) No short selling because private investors cannot short sell stocks whereas institutional investors can. 6) Borrowing and lending rates are the same. The interest rate we will use is the two week London Interbank Offered Rate (LIBOR). 7) The FTSE All-Share index is the market for performance comparison purposes because it represents approximately 98% of the total Main Market capitalisation, which also represents approximately 98% of the LSE capitalisation. Therefore assume that it represents the value of the Main Market. We also assume that the Main Market represents LSE. Regarding the following assumptions: ๏‚ท Number 4, in real life any positive quantity of stocks can be bought and sold through some share dealing accounts, such as Halifax Sharebuilder (Halifax Share Dealing Ltd, 2009). For example, an investor buys 100.5 RBS shares. The stockbroker buys 101 shares on the LSE. The former receives the required number of shares while the latter keeps the remaining fraction. ๏‚ท Number 5, if private investors are allowed to short sell stocks the market would be more volatile. ๏‚ท Number 6, investors in real life do not borrow and lend using the 2-week LIBOR rate. Banks do not allow investors to borrow funds for investment purposes. Investors can lend at the annual interest rate the bank offers on the savings account. Rates differ across different savings accounts and banks. ๏‚ท Number 7, we will use LSE capitalisations values as of 31/03/2009. We will use the following in order to conduct the experiment: ๏‚ท Digital Lookยฎ website (which the web address is www.digitallook.com) to: o Look up: ๏‚ง Portfolio and FTSE All-Share values at the end of the holding period in order to evaluate their performance. ๏‚ง Relevant ratios and information at the end of each trading day during the holding period to determine which stock no longer matches the criteria and sell them. o Keep track of the portfolios. o Select suitable stocks, defined by the criteria from the strategies, though the stock screener. o Execute transactions. o Construct and run portfolios. Microsoft Excelยฎ to: o Record transactions. 32

๏‚ท

๏‚ท

๏‚ท ๏‚ท

Calculate: ๏‚ง Portfolio betas, standard deviation and expected returns. ๏‚ง Overall holding period portfolio betas, standard deviation and expected returns (Details are shown below). ๏‚ง Value of portfolio after selling all stocks at the end of the period. o Record: ๏‚ง Portfolio values at the end of the holding period in order to evaluate their performance. ๏‚ง Relevant ratios and information at the end of each trading day during the holding period to determine which stock no longer matches the criteria and sell them. o Evaluate portfolio performance using the ratios outlined in the previous section. Thomson Datastream to download: o Past weekly stock prices for the past year of all stocks listed on the LSE in order to construct optimal portfolios. o Past weekly share prices for the past year of stocks held in the portfolios to evaluate their performance. o Stock betas. BullBearings website (which the web address is www.bullbearings.co.uk) to: o Execute stock transactions. o Run the portfolios. MATLABยฎ to o Calculate the efficient frontier of 20 portfolios and optimal portfolio from the universe of 1914 stocks, which is all of the companies, excluding investment trusts, listed on the LSE, using portopt and portalloc commands respectively. o If necessary, transform non-positive semi-definite covariance matrix into positive semi-definite one using the eigenvalue method of (Rousseeuw & Molenberghs, 1993, pp. 971-973) (see Appendix A.4).

o

The instructions regarding the use of the Digital Look and BullBearings websites are in Appendix A.5. The MATLAB codes for transforming the non-positive semi-definite covariance matrix into positive semi-definite one and calculating the efficient frontier and optimal portfolio are in Appendix A.7 We can calculate portfolio expected returns, standard deviation and betas for the whole holding period ๐ป by using the following: ๐ป๐‘…ฮ  = 1 1 1 ๐‘… + ๐‘… โ‹ฏ + ๐‘…ฮ ๐‘› ๐‘•1 ฮ 1 ๐‘•2 ฮ 2 ๐‘•๐‘› 1 1 1 ๐ป๐œŽฮ  = ๐œŽฮ 1 + ๐œŽฮ 2 โ‹ฏ + ๐œŽฮ ๐‘› ๐‘•1 ๐‘•2 ๐‘•๐‘›

and ๐ป๐›ฝฮ  = 1 1 1 ๐›ฝฮ 1 + ๐›ฝฮ 2 โ‹ฏ + ๐›ฝฮ ๐‘› ๐‘•1 ๐‘•2 ๐‘•๐‘›

where ๐‘•๐‘– is the number of days in a sub-period, where ๐‘– = 1, โ€ฆ , ๐‘›. Note that ๐‘•๐‘– are not necessarily equal to each other. We will use these to evaluate the performance of the investment portfolios because these take into account of all the stocks held during the holding period and this give an accurate figure for the portfolio. Regarding the universe of LSE stocks, we exclude investment trusts because some have assets consisting of assets in one or more sectors or indices on the exchange, some have assets which are unlisted and other have assets that are listed solely in other stock 33

exchanges. Including investment trusts in the universe will result in the optimal portfolio not having optimal allocations of companies listed on the LSE.

8. Results of the Experiment
Table 2: Optimal Stock proportions for Optimal Portfolio 1 with relevant data Stock Name Abcam ADL Advent Capital (Holdings) AEC Education Regenesis Group Ticker ABC AD. ADV AEC RGN Optimal Weighting (%) 0.26 75.63 0.03 0.15 23.94 ๐‘น๐šท (%) 0.87 0.96 14.27 0.93 1.80 ๐ˆ๐Ÿ (%) ๐šท 0.16 3.49 123.50 0.24 12.75 ๐ˆ๐šท (%) 4.06 18.69 111.13 4.92 35.71

Table 3: Optimal Stock proportions for Optimal Portfolio 2 with relevant data Stock Name Abcam Access Intelligence ADL Advanced Computer Software Advanced Medical Solutions Group Advent Capital (Holdings) Ticker ABC ACC AD. ASW AMS ADV Optimal Weighting (%) 0.52 1.13 91.36 1.61 3.52 1.86 ๐‘น๐šท (%) 1.09 0.49 0.96 0.58 0.36 14.27 ๐ˆ๐Ÿ (%) ๐šท 0.18 2.23 3.49 0.46 0.21 123.50 ๐ˆ๐šท (%) 4.24 14.92 18.69 6.76 4.63 111.13

Table 4: Optimal Stock proportions for Optimal Portfolio 3 with relevant data Stock Name ADL Advent Capital (Holdings) Albemarle and Bond Holdings Advanced Computer Software Alexander David Securities GRP AEC Education Admiral Group Altona Energy Ticker AD. ADV ABM ASW ADS AEC ADM ANR Optimal Weighting (%) 88.74 2.56 2.51 2.23 1.71 1.63 0.41 0.22 ๐‘น๐šท (%) 1.15 14.50 0.16 0.60 0.53 0.93 0.35 1.26 ๐ˆ๐Ÿ (%) ๐šท 3.47 123.47 0.30 0.46 2.85 0.24 0.39 2.78 ๐ˆ๐šท (%) 18.63 111.12 5.47 6.76 16.88 4.92 6.25 16.68

Table 5: Investment Portfolio weightings Portfolio Contrarian Value Investing Zulu Principle Number of Stocks invested 10 8 6 Initial Weighting (%) 10 12.5 16.66

We used MATLAB to calculate efficient frontier of 20 portfolios and the optimal portfolio for each of the three holding periods using the three sets of returns data and LIBORs (see Table 6). Regarding these sets of returns data, the history of returns goes back one year from the last trading day before the start of each holding period. The universes of stocks used to calculate the optimal portfolios are 294, 352 and 430 for the first, second and third portfolios respectively because the computer used struggles to calculate the solutions from the original universe of 1914 stocks. We explain this in more detail in the next section. Past index

34

values, stock prices and returns, additional details of optimal portfolio allocations and MATLAB files are on the CD provided. Table 2, Table 3 and Table 4 show the optimal proportions for optimal portfolios 1, 2 and 3 respectively. For each optimal portfolio, the data of expected returns, variances and betas are shown for each stock. We calculated the efficient frontiers for each universe and they were unsatisfactory so we decided not to include them in the report. We constructed the portfolios for each investment strategy using the criteria defined in section 4 (see Appendix A.6.1). We calculated overall portfolio expected return, standard deviation and beta for each portfolio (see Table 7). We also calculated the expected return and standard deviation for FTSE All-Share, going back one year from the last trading day before the start of each holding period, because no dividends were paid during each of the holding periods. We will use these to evaluate the performance of these portfolios.
Table 6: LIBOR used to calculate the optimal portfolios and evaluate the performance of portfolios for respective holding periods Period 1 2 3 2-week LIBOR (%) 0.82125 0.77 0.7275

Source: British Bankers Association Table 7: Overall portfolio values based on all stocks held in the portfolios over the whole holding period Portfolio Contrarian Value Zulu Principle Optimal 1 Optimal 2 Optimal 3 ๐‘ฏ๐‘น๐šท (%) -0.18 -0.73 -0.19 1.17 1.18 1.44% ๐‘ฏ๐ˆ๐šท (%) 1.24 10.12 4.89 22.73 19.60 20.09% ๐‘ฏ๐œท๐šท 1.15 0.88 0.48 0.34 0.40 0.42

Table 8: Portfolio and market returns using various measures defined in the report Period Portfolio Contrarian 1 Optimal 1 FTSE All-Share Value Investing 2 Optimal 2 FTSE All-Share Zulu Principle 3 Optimal 3 FTSE All-Share Return measure ๐‘น๐šท (%) 6.58 -32.10 8.44 -8.92 -20.85 2.76 -10.55 -25.07 3.06 ๐’“๐šท (%) 13.15 -64.20 0.17 -17.84 -41.71 0.06 -21.09 -50.15 0.06 SR 4.64 -1.44 1.51 -0.96 -1.1 0.39 -2.31 -1.28 0.45 TR 5.01 -96.83 7.61 -11.01 -54.05 1.99 -23.49 -61.42 2.33 -10.68 -25.27 -13.49 -34.88 -8.48 -21.07 -13.59 -29.18 JR 7.66 -35.51 FNS (%) 3.89 -67.14

Note that the values are calculated using the portfolio values after incurring transaction costs during the sale of all stocks for each portfolio.

Table 8 illustrates that overall the Contrarian Portfolio performed better than Optimal Portfolio 1 and the market. This is because, partly on an economic level, there were hopes that at the London G20 (group of 20 countries with developed and emerging economies) summit leaders would agree on a global plan to prevent the global economy from going into a depression and revive it. This made the investors regain some of the confidence lost 35

during the Credit Crunch to invest in the stock market. For the Contrarian Portfolio, Inchape returned a profit of 77.11% due to the fact that, from (London Stock Exchange Plc, 2009), the company announced that the rights issue of 9 shares for every one held at a price of 6 pence would take place. This resulted in a massive demand for the stock so the price increased dramatically since investors wanted to gain massive profit from the sale of the newly issued shares (an example of arbitrage). Kazakhmys returned a profit of 38.35% probably due to the fact the copper, zinc and gold prices were rising during the holding period. A few other companies made small profits and other made small losses. Regarding the latter, this was due to transaction costs and spread between the buying and selling prices (see Table 16 and Table 28). For the Optimal 1, the loss is due to the transaction costs and large spread between buying and selling prices (see Table 20 and Table 31). This is because no significant events took place during the holding period therefore prices of the stocks hardly changed during the holding period therefore Overall, the Value Investing Portfolio performed better than Optimal Portfolio 2 but worse than the market. This was due to transaction costs and some stocks have large bid/offer spreads (see Table 17 and Table 29). The overall portfolio loss is reduced by the fact that the share price of Jersey Electricity (A shares) increased during the holding period probably due to the company performed to its expectations (see (London Stock Exchange Plc, 2009)). Also the share price of Delta increased due to fact that the company performed better than the market expected. From (Delta Plc, 2009), turnover increased by 24%, operating profit increased by 109%, cash flow increased by 82%, EPS raised by 22% and dividend has gone up by 30%. This resulted in increased demand for Delta shares. Thus, it delivered 6.25% profit. Regarding Optimal Portfolio 2, the loss was due to transaction costs and large bid/offer spreads since the shares hardly changed during the holding period (see Table 20 and Table 32). The reasons are the same as in the case of Optimal 1. Overall, the Zulu Principle Portfolio performed better than Optimal Portfolio 3 but worse than the market. This was also due to the transaction costs and most stocks have large bid/offer spreads (see Table 18 and Table 30) since the shares prices generally have not changed very much during the holding period. Even the strong first half performance from Cash Converters (see (London Stock Exchange Plc, 2009)), strong performance from Visonic (see (London Stock Exchange Plc, 2009)) and a director buying shares in Zytronic (see (London Stock Exchange Plc, 2009)) failed to raise the confidence of investors to invest in these stocks. For Optimal Portfolio 3, the loss was due to transaction costs and large bid/offer spreads since the shares hardly changed during the holding period (see Table 21 and Table 32). The reasons are the same as in the case of first two optimal portfolios. Even quite strong performances of Advanced Medical Solutions Group (see (London Stock Exchange Plc, 2009)) failed to increase the demand for the stock. Regarding the above results, the returns would have been higher if the stocks have not been sold (see Appendix A.6.3). Additional details regarding the actual returns of stock after transaction costs in each portfolio are shown in Appendix A.6.5. During the holding period, we had to sell some stocks because they no longer match the investment criteria defined in the report. Details in the Stock and Sector Ratios.xls file.

9. Conclusion and Recommendations for Improvement
We used MATLAB to calculate efficient frontier and the optimal portfolio for each of the three holding periods using relevant data. For each holding period we used the past year's stock prices, going back one year from the last trading day before the start of the holding period, to calculate the data of returns. The resulting data had missing observations due to the fact that some stocks were not listed during these periods. Therefore we use nanmean and nancov commands to calculate the expected return vector and covariance matrix from observed data. If we used the mean and cov commands, the portopt command will not calculate the efficient frontier since not all elements in the covariance matrix are numbers. We had to 36

transform a non-positive semi-definite covariance matrix into a positive semi-definite one using the eigenvalue method of (Rousseeuw & Molenberghs, 1993). Then we use the expected returns vector, transformed covariance matrix and 2-week LIBOR to calculate the efficient frontier and optimal portfolio, using the portopt and portalloc commands. MATLAB did produce quite strange looking efficient frontiers because we did not use longer stock price history, that is, too few observations for each stock. We note that when we used the transformed covariance matrix to calculate the efficient frontier and optimal portfolio, MATLAB displays the messages stating that the covariance matrix is non-symmetric and non-positive definite. This is because computer have finite amount of memory to store numbers and MATLAB only stores number up to 15 digits. Also the algorithms in MATLAB tend to round the solutions to the nearest number resulting in inaccuracies. When we investigated the matrix manually, we noted that they were symmetric and positive semidefinite. We argue the following so that we can calculate the efficient frontiers and optimal portfolios in a reasonable amount of time: ๏‚ท ๏‚ท Investors will only choose stocks with positive expected returns because MATLAB takes extremely long time to calculate efficient frontier and the optimal portfolio from a universe of 1914 stocks. This reduced the universe of stocks to considerably. Using simple regression analysis of expected returns and variance for each returns data we excluded stocks with variances greater than 300% because including these would adversely affect the solutions, that is, this would lead to optimal portfolios consisting of just one stock which is unsatisfactory. Note that we did not include the regression results in this report because we only used this to decide the number of stocks to exclude.

These substantially reduced the universe of stocks. Hence, the number of stocks used to calculate the optimal portfolios are 294, 352 and 430 for the first, second and third portfolios respectively. Using the new respective universe of stocks, MATLAB only takes about 5 minutes to calculate the efficient frontier and an optimal portfolio. Based on the evidence in the previous section, we conclude that the portfolios using the investment strategies performed better than the optimal portfolios. This is because, in general, there was good economic news throughout the holding periods, especially regarding the London G20 summit. . Regarding individual companies, significant events took place which, in general, indicated that they were performing better than the market expected. There was a case of arbitrage regarding Inchape which contributed to the massive profit for the Contrarian Portfolio during the holding period. This lead to increased investor confidence in general which resulted in increased share prices generally. When Inchape announced that the 9 to 1 rights issue at the price of 6 pence would take place, large numbers of investors bought the shares in the hope of earning a large amount of risk free profit, indicating herd behaviour. Regarding Kazakhmys, the share price increased probably due to the fact the media noted that copper, zinc and gold prices has risen during the holding period leading to increased demand for the stock. Regarding Royal Dutch Shell, the media also mentioned that oil prices started to rise from their lowest point of $40 per barrel, leading increased demand for the stock. When Delta reported strong financial results demand increased for the stock. These suggest there is some evidence of anchoring and may be over-optimism. For other stocks of companies held in investment portfolios, the share prices have not changed very much, even though the ratios of these indicate that they are in good financial health, because there was no information which indicated that these would perform better than market expectations. This suggests that the investors were over-pessimistic. Overall, these suggest that there is some qualitative evidence that some aspects of behavioural finance had an effect on share prices of stocks. The optimal portfolios performed very badly for the following reasons; firstly, we only used past share prices going back one year to calculate the returns, expected returns and variances of stocks. This resulted in biased optimal allocations of stock for each optimal 37

portfolio. These in turn probably made the portfolio evaluation results biased since these suggest that the portfolios using investment strategies in general will perform better than optimal portfolios. Secondly, the stocks held in the optimal portfolios had large bid/offer spread. Therefore, the losses are due to this and the transaction costs. Overall, two week holding period is too short so that the results in this report are inconclusive. This illustrates that the evidence presented in this report is insufficient to prove that the portfolios using investment strategies consistently perform better than the optimal portfolios. Investors do not borrow or lend using 2 week LIBOR. Banks have policies that investors cannot borrow money for investment in the stock market. However, investors can lend at the rate of a savings account they deposit their money in. Under this case and if riskless lending rate is the same, we would end up with the same optimal portfolios. The rates differ across different savings and investment accounts and different banks and building societies. In some cases it is possible for investors to borrow at 0% interest rate (if they are being dishonest at the time application of an overdraft). These are university students who have student overdrafts on their student bank accounts and people who have bank accounts which offer interest free overdraft up to a certain amount (which is usually very small amount, say ยฃ200). The amount they can borrow is determined by their credit history and usage behaviour of their accounts. Similarly to the previous case, we would have the same optimal portfolios. We suggest improvements to the experiment to potential researchers: 1) Use 5 year history of stock prices to calculate the returns data. Ideally, the history of stock prices should be weekly because more observations usually mean more accurate results. 2) Use longer holding period so that the results are more conclusive. 3) If dividends are paid during holding periods, use the total return index to calculate ๐‘…ฮ  , ๐‘…ฮ  and ๐œŽฮ  of that index because this takes into account of price movements of stock and dividends paid. Also, the calculated values are more appropriate for portfolio performance evaluation. 4) Use weighted total return values, consisting of FTSE All-Share and FTSE AIM according to the market capitalisations of each index. Based on the market capitalisation values on the LSE on 31/03/2009, the Main Market represents approximately 98% of the value on the stock exchange. FTSE AIM represents approximately 2% of the value. Using these values, we can calculate the weighted total return values as follows: Weighted Value = (0.98 ร— FTSE All-Share index value) + (0.02 ร— FTSE AIM index value) Then we can use the above to calculate history of weighted total index values, which in turn we use to calculate the history of weighted returns. Then we use this to calculate ๐‘…ฮ  , ๐‘…ฮ  and ๐œŽฮ  of that index. 5) Use average instant access savings rate. Regarding all the above points except point 4, the actual returns of the optimal portfolios would probably be closer to their expected returns. Regarding point 4, we believe this would help future researchers to evaluate the performance of portfolios more accurately. Hence, these recommendations should make the results more reliable and conclusive.

38

Bibliography
Bary, A. (1997, 08 25). New Value in Old Saw. (The Wall Street Journal Digital Network, Dow Jones & Company Inc.) Retrieved 04 10, 2009, from Barrons: http://online.barrons.com/article/SB872305155168532500.html Black, F., & Litterman, R. (1992). Global portfolio optimization. Financial Analysts Journal , 48 (5), 28-43. Blume, M. E. (1970). Portfolio Theory: A Step Toward Its Practical Application. The Journal of Business , 43 (2), 152-173. Brandimarte, P. (2002). Numerical methods in finance: A MATLAB-based introduction (1st ed.). New York: Wiley-Interscience. Capaul, C., Rowley, I., & Sharpe, W. F. (1993). International Value and Growth Stock Returns. Financial Analysts Journal , 49 (1), 27-36. Chan, K. C. (1988). On the Contrarian Investment Strategy. The Journal of Business , 61 (2), 147-163. Chan, L. K., & Lakonishok, J. (2004). Value and Growth Investing: Review and Update. Financial Analysts Journal , 60 (1), 71-86. Davis, M. H., & Norman, A. R. (1990). Portfolio Selection with Transaction Costs. Mathematics of Operations Research , 15 (4), 676-713. Delta Plc. (2009). Delta Annual Report and Accounts. Dreman, D. (1998). Contrarian Investment Strategies: The Next Generation. New York: Simon & Schuster. Dreman, D. (1980). Contrarian Investment Strategy: The Psychology of Stock Market Success. New York: Random House. Dreman, D. (1982). The New Contrarian Investment Strategy. New York: Random House. Elton, E. J., Gruber, M. J., Brown, S. J., & Goetzmann, W. N. (2007). Modern Portfolio Theory and Investment Analysis (7th ed.). John Wiley & Sons. Fama, E. F. (1972). Components of Investment Performance. The Journal of Finance , 27 (3), 551-567. Fernholz, E. R. (2002). Stochastic Portfolio Theory. New York: Springer. Graham, B. (1973). The Intelligent Investor (4th Revised ed.). New York: Harper & Row. Graham, B., & Dodd, D. (1934). Security Analysis: Principles and Techniques (1st ed.). New York: McGraw-Hill. Graham, B., Dodd, D., & Cottle, C. S. (1963). Security Analysis: Principles and Techniques (4th ed.). New York: McGraw-Hill.

39

Graham, B., Zweig, J., & Buffett, W. E. (2003). The Intelligent Investor: The Definitive Book on Value Investing (Revised ed.). New York: Harper Business Essentials. Gregory, A., Harris, R. D., & Michou, M. (2001). An Analysis of Contrarian Investment Strategies in the UK. Journal of Business Finance & Accounting , 28 (9), 1192-1228. Halifax Share Dealing Ltd. (2009). Halifax ShareBuilder - Halifax Share Dealing. UK Shares and Stockbrokers. Retrieved 04 13, 2009, from Halifax: http://www.halifax.co.uk/sharedealing/products/halifax_sharebuilder.asp Howls, C. J., & Fitt, A. D. (2009). MATH 3022: Mathematical Finance. University of Southampton. Investopedia ULC. (2009). Contrarian. (Investopedia ULC) Retrieved 04 10, 2009, from Investopedia: http://www.investopedia.com/terms/c/contrarian.asp Jacobs, B. I., Levy, K. N., & Markowitz, H. M. (2005). Portfolio Optimization with Factors, Scenarios, and Realistic Short Positions. OPERATIONS RESEARCH , 53 (4), 586โ€“599. Jensen, M. C. (1968). The Performance of Mutual Funds in the Period 1945-1964. Journal of Finance , 23 (2), 389-415. Jucker, J. V., & de Faro, C. (1975). A Simple Algorithm for Stone's Version of the Portfolio Selection Problem. The Journal of Financial and Quantitative Analysis , 10 (5), 859-870. Kevin, S. (2007). Portfolio Management (2nd ed.). New Delhi: Prentice Hall of India. Levy, H., & Sarnat, M. (1984). Portfolio and Investment Selection: Theory and Practice. Englewood Cliffs, New Jersey: Prentice Hall. Lintner, J. (1965). The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets. The Review of Economics and Statistics , 47 (1), 1337. London Stock Exchange Plc. (2009, 03 17). Advanced Medical Solutions Group plc Preliminary Results for the Year Ended 31 December 2008. Retrieved 05 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=211 3532&source=RNS London Stock Exchange Plc. (2009, 02 23). Cash Converters International Ld ANNOUNCEMENT HALF YEAR RESULTS. Retrieved 02 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=209 6472&source=RNS London Stock Exchange Plc. (2009, 03 19). Inchcape PLC Announcement re: Rights Issue. Retrieved 05 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=211 5529&source=RNS

40

London Stock Exchange Plc. (2009, 01 23). Jersey Electricity Company Limited Interim Management Statement. Retrieved 05 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=207 5004&source=RNS London Stock Exchange Plc. (2009, 03 27). Visonic Limited Preliminary Results for the year ended 31 December 2008. Retrieved 05 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=212 2072&source=RNS London Stock Exchange Plc. (2009, 03 31). ZYTRONIC PLC DIRECTOR / PDMR SHAREHOLDING. Retrieved 05 20, 2009, from London Stock Exchange: http://www.londonstockexchange.com/LSECWS/IFSPages/MarketNewsPopup.aspx?id=212 4643&source=RNS Maginn, J. L., Tuttle, D. L., McLeavey, D. W., & Pinto, J. E. (2007). Managing Investment Portfolios: A Dynamic Process (3rd ed.). Hoboken, New Jersey: John Wiley & Sons. Markowitz, H. M. (1952). Portfolio Selection. The Journal of Finance , 7 (1), 77-91. Montier, J. (2002). Behavioural Finance: Insights into Irrational Minds and Markets. Chichester: John Wiley & Sons. Mossin, J. (1966). Equilibrium in a Capital Asset Market. Econometrica , 34 (4), 768-783. O'Shaughnessy, J. P. (2005). What Works on Wall Street: A Guide to the Best-Performing Investment Strategies of All Time (3rd ed.). New York: McGraw-Hill Professional. Pardolos, P. M., Sandstrรถm, M., & Zopounidis, C. (1994). On the Use of Optimization Models for Portfolio: A Review and Some Computational Results. Computational Economics , 7 (4), 227-244. Post-modern portfolio theory comes of age. (1993). Journal of Investing , 2 (3), 349-364. Rousseeuw, P. J., & Molenberghs, G. (1993). Transformation of non positive semidefinite correlation matrices. Communications in Statistics - Theory and Methods , 22 (4), 965 โ€“ 984. Sharpe, W. F. (1964). Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk. The Journal of Finance , 19 (3), 425-442. Sharpe, W. F. (1966). Mutual Fund Performance. The Journal of Business , 39 (1, Part 2), 119-138. Shefrin, H., & Statman, M. (2000). Behavioral Portfolio Theory. The Journal of Financial and Quantitative Analysis , 35 (2), 127-151. Slater, J. (1998). Beyond the Zulu Principle: Extraordinary Profits from Growth Shares (1st ed.). London: Orion Business. Slater, J. (1992). The Zulu Principle: Making Extraordinary Profits from Ordinary Shares (1st ed.). London: Orion.

41

Thaler, R. (1985). Mental Accounting and Consumer Choice. Marketing Science , 4 (3), 199214. The Panel on Takeovers and Mergers. (2009). Fees and Charges. Retrieved 04 13, 2009, from The Panel on Takeovers and Mergers: http://www.thetakeoverpanel.org.uk/new/fees_charges/fees_charges_main.htm Treynor, J. (1965). How to Rate Management of Investment Funds. Harvard Business Review , 43 (1), 63-75.

42

Appendices
A.1. Capital Asset Pricing Model (CAPM)
This section is based on (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 284-286).The CAPM was suggested independently by (Sharpe, 1964), (Lintner, 1965) and (Mossin, 1966). The theory introduces a number of additional assumptions regarding the market and the behaviour of other investors in order to construct a general equilibrium of asset prices in the entire market. The assumptions of CAPM are: 1) No transaction costs and taxes, such as stockbroker commissions, stamp duty, capital gains tax and so on. 2) Investors prefer a portfolio with a higher return to one with a lower return at a given level of risk. 3) Investors are risk averse, where they prefer a portfolio with lower risk to one with higher risk for a given level of return. 4) Assets are infinitely divisible. This means that investors can own any quantity of an asset. 5) Investors are price takers, meaning that they cannot affect the asset prices through buying and selling. 6) Investors make decisions based on Mean-Variance theory. 7) Short selling at any amount is allowed. 8) Investors can borrow and lend any amount at a given interest rate. 9) Investors are concerned with the expected return and variance in a single period, and define the relevant period in the same way. 10) Investors have identical expectations with respect to expected return, variance and correlation matrix. 11) All assets can be bought or sold on the market. In real life, all assumptions, expect assumption 7, do not hold. Under these assumptions, all investors face the same efficient frontier and the same risk free rate. Hence the efficient frontier is just a straight line. This line is called the CML. The optimal portfolio is at a point where the CML is tangent to the efficient frontier and that all investors have the same optimal allocations in their portfolios. Also all investors will hold a combination of risky and riskless assets. A.1.1. Derivation of CAPM In this section we will rigorously derive the CAPM. There are various ways of deriving the theory (see (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 293-294), (Sharpe, 1964), (Lintner, 1965) and (Mossin, 1966)). From (Howls & Fitt, 2009, pp. 239-240), let there be ๐‘• = 1, โ€ฆ , ๐ป investors in the market of ๐‘ risky assets and one riskless asset. The usual notations of expected returns and risk free rate of return will be used. Suppose that ๐œ”๐‘–๐‘• is the proportion of the total market in asset ๐‘– that investor ๐‘• has in the portfolio. Then ๐‘ ๐œ”๐‘–๐‘•

= ๐œ™๐‘–๐‘• ๐‘–
=0

where ๐œ™๐‘–๐‘• is the total in the economy that the investor has. From the above, there is a relationship between the investorโ€™s proportion of the total market of asset ๐‘–, ๐œ”๐‘–๐‘• , and the proportion of their wealth invested in the asset, ๐œ†๐‘–๐‘• . Thus

43 ๐œ”๐‘–๐‘•

= ๐œ†๐‘–๐‘• ๐œ™๐‘–๐‘• with ๐‘ ๐œ†๐‘–๐‘•

= 1. ๐‘–
=0

Then the portfolio expected return, comprising of risky and riskless assets, is ๐‘ ๐‘…

ฮ ๐‘• = ๐‘–
=0

1 ๐œ†๐‘–๐‘• ๐‘…๐‘– = ๐œ™๐‘–๐‘• ๐‘ ๐œ”๐‘–๐‘•

๐‘…๐‘– ๐‘–
=0

(21)

and the variance is ๐‘
2 ๐œŽฮ ๐‘• ๐‘

= ๐‘–
=1 ๐‘— =1

1 ๐œ†๐‘–๐‘• ๐œ†๐‘—๐‘• ๐œŽ๐‘–๐‘— = 2 ๐œ™๐‘• ๐‘ ๐‘ ๐œ”๐‘–๐‘•

๐œ”๐‘—๐‘• ๐œŽ๐‘–๐‘— . ๐‘–
=1 ๐‘— =1

From assumptions 2 and 3, each investor maximises their expected utility, where utility is the measurement of happiness,
2 ๐‘ขฮ ๐‘• = ๐‘ข๐‘• ๐‘…ฮ ๐‘• , ๐œŽฮ ๐‘•

subject to ๐‘ ๐œ”๐‘–๐‘•

= ๐œ™๐‘–๐‘• . ๐‘–
=0

Using Lagrange multipliers, the expected utility can be rewritten as ๐‘

max

2 ๐‘ข๐‘• ๐‘…ฮ ๐‘• , ๐œŽฮ ๐‘•

+ ๐œ‡๐‘• ๐‘–
=0 ๐œ”๐‘–๐‘•

โˆ’ ๐œ™๐‘–๐‘•

,

where ๐œ‡๐‘• is the Lagrange multiplier for each investor which maximises the constraint which, if satisfied, leaves the above function unchanged. The conditions for a critical point for the above function is vanishing of all the first partial derivatives with respect to ๐œ”๐‘–๐‘• . Using the chain rule, this results in
2 ๐œ•๐‘ข๐‘• ๐œ•๐‘…ฮ ๐‘• ๐œ•๐‘ข๐‘• ๐œ•๐œŽฮ ๐‘• . + 2 . + ๐œ‡๐‘• = 0 ๐œ•๐‘…ฮ ๐‘• ๐œ•๐œ”๐‘–๐‘• ๐œ•๐œŽฮ ๐‘• ๐œ•๐œ”๐‘–๐‘•

for โˆ€๐‘–.

Using the portfolio expected return and variance expressions on the above to become ๐œ•๐‘ข๐‘• ๐‘…๐‘– ๐œ•๐‘ข๐‘• . + 2 . ๐œ•๐‘…ฮ ๐‘• ๐œ™๐‘–๐‘• ๐œ•๐œŽฮ ๐‘• and ๐‘ ๐œ”๐‘—๐‘•

๐œŽ๐‘–๐‘—
2 ๐œ™๐‘•

+ ๐œ‡๐‘• = 0

for ๐‘– = 1, โ€ฆ , ๐‘ ๐‘—

=1

44 ๐œ•๐‘ข๐‘•

๐‘…0 ๐œ•๐‘ข๐‘• . + 2 . 0 + ๐œ‡๐‘• = 0 ๐œ•๐‘…ฮ ๐‘• ๐œ™๐‘–๐‘• ๐œ•๐œŽฮ ๐‘• which simplifies to ๐œ•๐‘ข๐‘• ๐œ•๐‘ข๐‘• . ๐‘…1 + 2 . ๐œ•๐‘…ฮ ๐‘• ๐œ•๐œŽฮ ๐‘• and ๐œ•๐‘ข๐‘• ๐‘…0 . + ๐œ‡๐‘• = 0 ๐œ•๐‘…ฮ ๐‘• ๐œ™๐‘• ๐‘

for ๐‘– = 0, ๐‘—

=1 ๐œ”๐‘—๐‘•

๐œŽ๐‘–๐‘— + ๐œ‡๐‘• = 0 ๐œ™๐‘•

for ๐‘– = 1, โ€ฆ , ๐‘

(22)

for ๐‘– = 0.

(23)

Rearranging equation (23) in terms of ๐œ‡๐‘• and substituting into equation (22) to get ๐œ•๐‘ข๐‘• ๐œ•๐‘ข๐‘• . ๐‘…๐‘– โˆ’ ๐‘…0 + 2 . ๐œ•๐‘…ฮ ๐‘• ๐œ•๐œŽฮ ๐‘• ๐‘ ๐‘—

=1 ๐œ”๐‘—๐‘•

๐œŽ๐‘–๐‘— =0 ๐œ™๐‘•

for ๐‘– = 1, โ€ฆ , ๐‘.

Consider two of these equations for two different assets ๐‘– and ๐‘˜, and divide one equation by the other to obtain ๐œ•๐‘ข๐‘• ๐œ•๐‘ข๐‘• . ๐‘…๐‘– โˆ’ ๐‘…0 2 . ๐œ•๐œŽฮ ๐‘• ๐œ•๐‘…ฮ ๐‘• = ๐œ•๐‘ข๐‘• ๐œ•๐‘ข๐‘• . ๐‘…๐‘˜ โˆ’ ๐‘…0 2 . ๐œ•๐‘…ฮ ๐‘• ๐œ•๐œŽฮ ๐‘• ๐‘
๐œ”๐‘—๐‘• ๐œŽ๐‘–๐‘— ๐‘— =1 ๐œ™ ๐‘• ๐‘ ๐‘— =1 ๐œ”๐‘—๐‘•

๐œŽ๐‘˜๐‘— . ๐œ™๐‘•

The partial derivatives and ๐œ™๐‘• cancel each other out resulting in ๐‘…๐‘– โˆ’ ๐‘…0 = ๐‘…๐‘˜ โˆ’ ๐‘…0 ๐‘
๐‘— =1 ๐œ”๐‘—๐‘• ๐œŽ๐‘–๐‘— . ๐‘ ๐‘— =1 ๐œ”๐‘—๐‘• ๐œŽ๐‘˜๐‘—

This relationship holds for all investors and, hence, we sum all investors to arrive at ๐‘…๐‘– โˆ’ ๐‘…0 = ๐‘…๐‘˜ โˆ’ ๐‘…0 Rewriting the above as ๐‘…๐‘– โˆ’ ๐‘…0 ๐‘…๐‘˜ โˆ’ ๐‘…0 = ๐‘ ๐‘ ๐‘— =1 ๐œ‚๐‘— ๐œŽ๐‘–๐‘— ๐‘— =1 ๐œ‚๐‘— ๐œŽ๐‘˜๐‘— and since this must hold for all assets ๐‘– and ๐‘˜, the above expression must be equal to some constant ๐›พ. That is ๐‘…๐‘– โˆ’ ๐‘…0 = ๐›พ. ๐‘ ๐‘— =1 ๐œ‚๐‘— ๐œŽ๐‘–๐‘— Multiplying the numerator and denominator by ๐œ‚๐‘– and sum all ๐‘– assets to obtain (24) ๐‘
๐‘— =1 ๐‘ ๐‘— =1 ๐‘ ๐‘— =1 ๐œ”๐‘—๐‘• ๐œŽ๐‘–๐‘— ๐‘ ๐‘— =1 ๐œ”๐‘—๐‘• ๐œŽ๐‘˜๐‘—

= ๐‘

๐‘— =1 ๐‘ ๐‘— =1 ๐‘

๐‘—=1 ๐œ”๐‘—๐‘• ๐‘ ๐‘— =1 ๐œ”๐‘—๐‘• ๐œŽ๐‘–๐‘—

๐œŽ๐‘˜๐‘—

= ๐‘

๐‘— =1 ๐œ‚๐‘— ๐œŽ๐‘–๐‘— . ๐‘ ๐‘— =1 ๐œ‚๐‘— ๐œŽ๐‘˜๐‘—

45 ๐‘ ๐‘ ๐‘ ๐œ‚๐‘–

๐‘…๐‘– โˆ’ ๐‘…0 = ๐›พ ๐‘–
=1 ๐‘–=1 ๐‘— =1 ๐œ‚๐‘—

๐œ‚๐‘˜ ๐œŽ๐‘–๐‘— .

(25)

Using the definitions of expected return and variance of the optimal market portfolio ๐‘€, ๐‘ ๐‘…๐‘€

= ๐‘–
=1 ๐‘…๐‘–

and ๐‘
2 ๐œŽ๐‘€ ๐‘

= ๐‘–
=1 ๐‘— =1 ๐œ‚๐‘—

๐œ‚๐‘˜ ๐œŽ๐‘–๐‘—

and the fact that ๐‘ ๐œ‚๐‘–

= 1, ๐‘–
=1

simplifying equation (25) yields
2 ๐‘…๐‘€ โˆ’ ๐‘…0 = ๐›พ๐œŽ๐‘€ .

(26)

Since covariances are additive, the covariance of asset ๐‘– with the optimal market portfolio ๐‘€ is ๐‘ ๐œŽ๐‘–๐‘€

= ๐‘—
=1 ๐œ‚๐‘—

๐œŽ๐‘–๐‘— .

Therefore rearranging equation (26) in terms of ๐›พ and substituting this into equation (24) to get ๐‘…๐‘– โˆ’ ๐‘…0 ๐‘…๐‘€ โˆ’ ๐‘…0 = 2 ๐œŽ๐‘–๐‘€ ๐œŽ๐‘€ and rearranging to arrive at ๐‘…๐‘– = ๐‘…0 + ๐‘…๐‘€ โˆ’ ๐‘…0 ๐œŽ๐‘–๐‘€ = ๐‘…0 + ๐‘…๐‘€ โˆ’ ๐‘…0 ๐›ฝ๐‘– 2 ๐œŽ๐‘€

where ๐›ฝ๐‘– is the beta of asset ๐‘– , which is covariance of asset ๐‘– with the market portfolio divided by the variance of the market portfolio. The above expression is the Security Market Line (SML) and it represents the expected return for all assets and portfolios of assets in the whole economy. The linear relationship between the expected return and on any two assets is determined by their betas. The higher the ๐›ฝ๐‘– , the equilibrium return is higher, and vice versa. The above result can be applied to a portfolio ฮ  of riskless and risky assets by summing over all risky assets. The CML can be derived from equation (20). First note that since ๐›ฝ๐‘– are covariances of asset ๐‘– with the market ๐‘€ then ๐›ฝฮ  can be obtained from the weighted sum of ๐›ฝ๐‘– s. Thus, the expression is 46 ๐‘ ๐›ฝ

ฮ  = ๐‘–
=1 ๐œ‚๐‘–

๐›ฝ๐‘– .

Therefore, the expected return on portfolio ฮ  for investor ๐‘• is ๐‘ ๐‘…

ฮ ๐‘• = ๐‘–
=0 ๐‘ ๐œ”๐‘–๐‘•

๐‘… ๐œ™๐‘–๐‘• ๐‘– ๐œ”๐‘–๐‘• ๐‘…๐‘€ โˆ’ ๐‘…0 ๐‘…0 + ๐‘…0 + ๐œ™๐‘–๐‘• ๐œŽ๐‘€ ๐œ”๐‘–๐‘• ๐‘…๐‘€ โˆ’ ๐‘…0 ๐‘…0 + ๐œ™๐‘–๐‘• ๐œŽ๐‘€ ๐‘…๐‘€ โˆ’ ๐‘…0 ๐œŽ๐‘€ ๐‘
๐‘

= ๐‘–
=0 ๐‘ ๐œŽ๐‘–๐‘€

2 ๐œŽ๐‘€ ๐œ”๐‘–๐‘• ๐œ™๐‘–๐‘• ๐œŽ๐‘–๐‘€ 2 ๐œŽ๐‘€

= ๐‘–
=0 ๐‘–

=0

= ๐‘…0 + ๐‘…๐‘€

โˆ’ ๐‘…0 = ๐‘…0 + ๐œŽฮ ๐‘• . ๐œŽ๐‘€ The above equation is the CML. Recall from section 3.4, the gradient ๐‘…๐‘€ โˆ’ ๐‘…0 ๐œŽ๐‘€ is the market price of risk and point ๐‘€ on the CML which is tangent to the efficient frontier of the opportunity set of risky assets is the optimal portfolio at a given rate of interest (see Figure 9 in section 3.4). Thus, this maximises the market price of risk. All investors should select this part of portfolio, although the choice of proportion in relation to the risky asset depends on their preferences. We know from section 3.3.3, if ๐œ†0๐‘• < 0 and ๐œ†๐‘–๐‘• > 1 , the investor is borrowing from the bank at ๐‘…0 and investing the loan in more risky assets. Conversely, if ๐œ†0๐‘• > 1 and ๐œ†๐‘–๐‘• < 0, short selling of risky assets takes place and the proceeds are deposited in the bank account to earn interest at ๐‘…0 . If ๐œ†0๐‘• = ๐œ†๐‘–๐‘• = 0, then no investment in the risky assets takes place. This completes the derivation of CAPM. ๐‘–

=0 ๐œ”๐‘–๐‘•

๐œ™๐‘–๐‘• ๐œŽ๐‘–๐‘€

2 ๐œŽ๐‘€

47

A.2. Quadratic Programming
This appendix is based heavily on (Elton, Gruber, Brown, & Goetzmann, 2007, pp. 122-125) and (Brandimarte, 2002, pp. 154-159). Also based on (Jucker & de Faro, 1975), (Pardolos, Sandstrรถm, & Zopounidis, 1994, pp. 240-241) and (Jacobs, Levy, & Markowitz, 2005). Recall from sections 3.5.3 to 3.5.5 that the maximisation and minimisation of the objective functions subject to inequality constraints can only be solved through the use of quadratic programming. This programming method is called Kuhn-Tucker conditions (also known as Karushโ€“Kuhnโ€“Tucker conditions). This basically restricts ๐œ†๐‘– to positive values. If the maximum of ๐œ†๐‘– is negative, then the feasible maximum of ๐œƒ is at ๐œ†๐‘– = 0 where ๐œ•๐œƒ ๐œ•๐œ†๐‘– < 0. Whereas if the maximum of ๐œ†๐‘– is positive then the maximum of ๐œƒ is at some ๐œ†๐‘– > 0 where ๐œ•๐œƒ ๐œ•๐œ†๐‘– = 0. Before we proceed further, we will define this formally. Consider the following problem: min s. t. ๐‘“ ๐ฑ ๐‘•๐‘– ๐ฑ = 0 ๐‘”๐‘– ๐ฑ โ‰ค 0 ๐‘– โˆˆ ๐ธ ๐‘– โˆˆ ๐ผ,

where ๐‘“, ๐‘”๐‘– and ๐‘•๐‘– are functions and ๐ฑ is a vector. The stationary, which means variables are constant over time, of ๐‘“ has no role in finding the optimal solution but the Lagrangian function โ„’ does have one. Consider the function โ„’ ๐ฑ, ๐€, ๐ = ๐‘“ ๐ฑ + ๐‘–
โˆˆ๐ธ ๐œ†๐‘–

๐‘•๐‘– ๐ฑ + ๐‘–
โˆˆ๐ผ ๐œ‡๐‘–

๐‘”๐‘– ๐ฑ ,

where ๐œ† and ๐œ‡ are Lagrangian multipliers. The stationary of the above does play a role in the Kuhn-Tucker conditions. THEOREM 1 (Kuhn-Tucker conditions) Assume that ๐‘“ , ๐‘•๐‘– and ๐‘”๐‘– are continuously differentiable, and that ๐’™โˆ— is feasible and satisfies a constraint qualification condition. Then a necessary condition for the local optimality of ๐’™โˆ— is that there exist numbers ๐œ†โˆ— for ๐‘– โˆˆ ๐ธ and ๐‘– โˆ— ๐œ‡๐‘– โ‰ฅ 0 for ๐‘– โˆˆ ๐ผ such that โˆ‡๐‘“ ๐ฑ โˆ— ๐‘–
โˆˆ๐ธ ๐œ†

โˆ— โˆ‡๐‘•๐‘– ๐ฑ โˆ— + ๐‘– ๐‘–
โˆˆ๐ผ ๐œ‡๐‘–

โˆ— โˆ‡๐‘”๐‘– ๐ฑ โˆ— = ๐ŸŽ

and
โˆ— ๐œ‡๐‘– ๐‘”๐‘– ๐ฑ โˆ— = 0

โˆ€๐‘– โˆˆ ๐ผ.

All of these conditions has to be satisfied to ensure that the optimal solution is feasible. Back to where we were, we write ๐œ•๐œƒ โ‰ค 0. ๐œ•๐œ†๐‘– We make the above into an equality by writing ๐œ•๐œƒ + ๐‘ˆ๐‘– = 0 ๐œ•๐œ†๐‘– where ๐‘ˆ๐‘– is a dummy variable. This is the first Kuhn-Tucker condition for a maximum. If the optimum value of ๐œƒ exists when ๐œ†๐‘– > 0 , then ๐œ•๐œƒ ๐œ•๐œ†๐‘– = 0 and ๐‘ˆ๐‘– = 0 . In addition, if the 48

optimum exists at ๐œ†๐‘– = 0, then ๐œ•๐œƒ ๐œ•๐œ†๐‘– < 0 and ๐‘ˆ๐‘– > 0. The second, third and forth conditions are ๐œ†๐‘– ๐‘ˆ๐‘– = 0 ๐œ†๐‘– โ‰ฅ 0 and ๐‘ˆ๐‘– โ‰ฅ 0. If the proportions of ๐œ†๐‘– satisfy all Kuhn-Tucker conditions, then these give the optimum portfolio.

49

A.3. An example of calculating the Optimal Portfolio
This appendix is based heavily on (Howls & Fitt, 2009, pp. 36-37). Suppose that we have three risky assets ๐‘†1 , ๐‘†2 and ๐‘†3 and a riskless asset ๐‘†0 . The respective expected returns, standard deviations and correlations are ๐‘…0 = 0.05, ๐‘…1 = 0.14, ๐‘…2 = 0.08, ๐‘…3 = 0.2, ๐œŽ

1 = 0.06, and ๐œŒ12 = ๐œŒ21 = 0.5, Short selling is allowed. In this case we ๐œŽ

2 = 0.03, ๐œŽ

3 = 0.15 ๐œŒ

13 = ๐œŒ31 = 0.2, ๐œŒ

23 = ๐œŒ32 = 0.4. ๐œƒ

= subject to
3 ๐‘…๐‘ƒ

โˆ’ ๐‘…0 ๐œŽ๐‘ƒ ๐œ†๐‘–

= 1. ๐‘–
=1

First, we calculate the ๐œŽ๐‘–๐‘— to get ๐œŽ๐‘–๐‘— =
2 The portfolio variance ๐œŽฮ  is 3 2 ๐œŽฮ  3 3

1 36 9 9 9 104 18 18

18 18 . 225

= ๐‘–
=1 ๐œ†

2 ๐œŽ๐‘–2 ๐‘–

+ ๐‘–
=1 ๐‘— =1 ๐‘— โ‰ ๐‘– ๐œ†๐‘–

๐œ†๐‘— ๐œŽ๐‘–๐‘—

=

1 2 36๐œ†1 + 18๐œ†1 ๐œ†2 + 36๐œ†1 ๐œ†3 + 9๐œ†2 + 36๐œ†2 ๐œ†3 + 225๐œ†2 2 3 104

so the portfolio standard deviation ๐œŽฮ  is ๐œŽฮ  = We also have ๐‘

1 2 36๐œ†1 + 18๐œ†1 ๐œ†2 + 36๐œ†1 ๐œ†3 + 9๐œ†2 + 36๐œ†2 ๐œ†3 + 225๐œ†2 . 2 3 100 ๐‘…๐‘ƒ

โˆ’ ๐‘…0 = ๐‘–
=1 ๐œ†๐‘–

๐‘…๐‘– โˆ’ ๐‘…0 =

1 9๐œ†1 + 3๐œ†2 + 15๐œ†3 . 100

and so

50 ๐œƒ

=

9๐œ†1 + 3๐œ†2 + 15๐œ†3
2 36๐œ†1

+ 18๐œ†1 ๐œ†2 + 36๐œ†1 ๐œ†3 + 9๐œ†2 + 36๐œ†2 ๐œ†3 + 225๐œ†2 2 3

.

We differentiate with respect to each ๐œ†๐‘– to get ๐œ•๐œƒ = โˆ’๐›พ 36๐œ†1 + 9๐œ†2 + 18๐œ†3 + 14 โˆ’ 5 = 0 ๐‘‘๐œ†1 ๐œ•๐œƒ = โˆ’๐›พ 9๐œ†1 + 9๐œ†2 + 18๐œ†3 + 8 โˆ’ 5 = 0 ๐‘‘๐œ†2 ๐œ•๐œƒ = โˆ’๐›พ 18๐œ†1 + 18๐œ†2 + 225๐œ†3 + 20 โˆ’ 5 = 0 ๐‘‘๐œ†3 and rearranging in terms if ๐‘…๐‘– โˆ’ ๐‘…0 yields ๐›พ 36๐œ†1 + 9๐œ†2 + 18๐œ†3 = 9 ๐›พ 9๐œ†1 + 9๐œ†2 + 18๐œ†3 = 3 ๐›พ 18๐œ†1 + 18๐œ†2 + 225๐œ†3 = 15. Using the trick described in section 3.5.1 on the above, we obtain 36๐‘1 + 9๐‘2 + 18๐‘3 = 9 9๐‘1 + 9๐œ†2 + 18๐œ†3 = 3 18๐‘1 + 18๐‘2 + 225๐‘3 = 15. We simplify the system of equations to get 4๐‘1 + ๐‘2 + 2๐‘3 = 1 3๐‘1 + 3๐‘2 + 6๐‘3 = 1 6๐‘1 + 6๐‘2 + 75๐‘3 = 5. (26) (27) (28)

Then we solve for each ๐‘๐‘– . First, we multiply equation (26) by 3, subtract this from equation (27) and solve for ๐‘1 yields 2 9๐‘1 = 2 โ‡’ ๐‘1 = . 9 We substitute ๐‘1 into, say, equation (26) to acquire 4 2 + ๐‘2 + 2๐‘3 = 1, 9

multiply this by 6 and subtract from equation (27) to obtain 18 Then we solve for ๐‘3 yields ๐‘3 = 1 . 21 2 + 63๐‘3 = โˆ’1. 9

Substitute ๐‘1 and ๐‘3 into equation (26) and rearrange in terms of ๐‘2 results in

51

4 To summarise, we have

2 1 1 + ๐‘2 + 2 = 1 โ‡’ ๐‘2 = . 9 21 63

2 ๐‘1 = , 9 Note that
3 ๐‘–=1 ๐œ†๐‘– ๐‘

2 =

1 , 63 ๐‘

3 =

1 . 21 we find that the allocations are

= 1 so

3 ๐‘–=1 ๐‘๐‘–

= 2 7 . Using ๐œ†๐‘– = ๐‘๐‘– 14 , 18 ๐œ†2 = 1 , 18

3 ๐‘–=1 ๐‘๐‘– , ๐œ†

1 = ๐œ†

3 =

3 18

which give the optimal portfolio. The portfolio expected return is ๐‘…ฮ  = ๐œ†1 ๐‘…1 + ๐œ†2 ๐‘…2 + ๐œ†3 ๐‘…3 14 1 3 = 14 +8 + 20 18 18 18 11 = โ‰ˆ 0.1466 75 and risk is ๐œŽฮ  = 1 2 2 ๐œ†2 ๐œŽ 2 + ๐œ†1 ๐œ†2 ๐œŽ1 ๐œŽ2 + ๐œ†1 ๐œ†3 ๐œŽ1 ๐œŽ3 + ๐œ†2 ๐œŽ2 + ๐œ†2 ๐œ†3 ๐œŽ2 ๐œŽ3 + ๐œ†2 ๐œŽ3 2 3 100 1 1 1 14 2 14 1 14 3 1 2 1 = 36 + 18 + 36 +9 + 36 18 18 18 18 18 18 18 100

3 18

+ 225

3 2 18

and this leads to 1218 โ‰ˆ 0.05816. 600 ๐œŽ

ฮ  = The market price of risk is ๐œƒ = 29 2 ๐‘…ฮ  =

609 โ‰ˆ 1.66 and the CML is given by

1 29 2 + ๐œŽ โ‰ˆ 0.05 + 1.66๐œŽ. 20 609

52

A.4. Transforming Non-Positive Semi-Definite Covariance Matrices
This appendix is based on (Rousseeuw & Molenberghs, 1993, pp. 971-973). Before we explain the method, we will go through the concept of transforming a non-positive semidefinite correlation matrix into positive semi-definite one. Note that non-positive semi-definite correlation matrices occur due to the method the researcher collected the data. Since the correlation matrix ๐šธ is symmetric and positive semi-definite we can write ๐šธ = ๐Ž๐ƒ๐Ž๐‘‡ where ๐ƒ is the diagonal matrix containing the eigenvalues of ๐šธ and ๐Ž is the orthogonal matrix of matching eigenvectors. Suppose that the eigenvalues appear in decreasing order. Let ๐‘Œ๐‘– be standardised versions of initial variables. Then an orthonormal set of eigenvectors ๐‘๐‘— corresponds to the linear combinations of ๐‘Œ๐‘– . These linear combinations are uncorrelated and have decreasing variance. In the case of ๐šธ is positive definite, the eigenvalues are strictly positive. When ๐ non-positive semi-definite, due to some eigenvalues are negative, but symmetric, the above expression still holds. Usually, these negative values will not have large absolute values. We can simply replace the negative eigenvalues with their absolute values. Now we let ๐ƒโ€ฒ be diagonal matrix containing the absolute eigenvalues. Then we calculate the positive definite correlation matrix ๐šธโ€ฒ = ๐Ž๐ƒโ€ฒ๐Ž๐‘‡ . If the diagonal elements of ๐šธโ€ฒ are not equal to 1, we transform ๐šธโ€ฒ to ๐šธ = ๐ƒ1 ๐‘โ€ฒ ๐ƒ1 ,
โ€ฒ where ๐ƒ1 is the diagonal matrix with diagonal elements 1 ๐‘๐‘–๐‘– . In our case, we need to transform a non-positive semi-definite covariance matrix into positive semi-definite one by transforming the former into a correlation matrix by using the cov2corr command. Then we use the above to calculate the positive definite correlation matrix. Then we the corr2cov command to convert the matrix into positive semi-definite covariance matrix. Then we check the covariance matrix by using the eig command to ensure that it is positive definite. Surprisingly, we find that the matrix is positive semi-definite since there were no negative eigenvalues.

53

A.5. Experiment Instructions
A.5.1. Contrarian Investment Strategy 1) On the Digital Look website, click on โ€˜Visual Toolsโ€™ link under โ€˜Investor Toolboxโ€™ menu. Then click on โ€˜Heat Mapsโ€™ link.

Figure 11: Visual Tools webpage on Digital Look website. 'Heat Maps' link is highlighted by the underline.

Figure 12: Heat maps webpage on the Digital Look website.

54

2) Click on โ€˜Sector Averagesโ€™ in the โ€˜Sectorโ€™ drop-down menu. 3) Find the average values of sectors by selecting the following ratios; P/E, P/BV, P/CF and Dividend Yield. The first three are found by clicking on โ€˜Valuation Ratiosโ€™. The last is under โ€˜Income Ratiosโ€™. Both under โ€˜Data Plot Categoryโ€™ drop-down menu. Click on the relevant ratios in the โ€˜Data Plotโ€™ drop-down menu. An example is shown in Figure 13.

Figure 13: Average P/E values of sectors defined by LSE on the Digital Look website.

4) Click on โ€˜Screening Toolsโ€™ link under โ€˜Investor Toolboxโ€™ menu. Then click on โ€˜Screenerโ€™.

Figure 14: Screener webpage with โ€˜full fundamental screenerโ€™ link highlighted on the Digital Look website.

55

5) Click on โ€˜full fundamental screenerโ€™ link under โ€˜Getting Startedโ€™ (see Figure 14). Ensure that UK flag is displayed on the next screen. 6) Select the required sector under โ€˜Sectorโ€™ menu.

Figure 15: Full Screener webpage.

7) Enter the market values from the required sector into the โ€˜Maxโ€™ fields of the following ratios: P/E, P/BV and P/CF (This is labelled as โ€˜Share Price / Operating Cashflow per Shareโ€™). These are under Valuation Ratios section.

Figure 16: Example result of application of steps 1 to 10 for Contrarian Investment Strategy.

56

8) Enter the market value for dividend yield into the โ€˜Minโ€™ field. This is under โ€˜Income Ratiosโ€™ section. 9) Enter 0.1 in the โ€˜Minโ€™ fields of the following ratios to weed out zero and negative values: P/CF and P/S. Enter 1.5 in the โ€˜Minโ€™ field of dividend cover. 10) Click on โ€˜Run Screenโ€™ at the bottom of the webpage. An example is shown in Figure 16.

Figure 17: Heat map showing 12 companies with positive P/E values. The values are arranged in decreasing order so the last two are in the two lowest quintiles.

Figure 18: Summary page of a company.

57

11) Use โ€˜Heat Mapsโ€™ to check if companies lie in the two lowest quintiles by value for P/E, P/BV and P/CF and the two highest quintiles for dividend yield. This is done by selecting the relevant sector the companies are listed under in LSE. Arrange the companies in value and tick the box to hide โ€˜n/aโ€™ values. Each row has up to four companies. Count the rows with four companies and add the number of companies appearing on the last row with less than four appearing. For example, there are ten rows of four companies and one row of two. Therefore, there are 42. In this case there are 12 as shown in Figure 17.

Figure 19: Final results showing accounts for a company going back up to five years.

Figure 20: 'UK Shares' webpage showing indices. 'FTSE 100' is highlighted by the underline.

58

12) To check if the companies have current ratio of at least 1: a) Click on link of a company. This loads the summary page (see Figure 18). b) Click on โ€˜Financialsโ€™ tab on the company webpage. c) Click on โ€˜Fundamentalsโ€™. This loads the final results accounts for the company. This is indicated by the highlight of the โ€˜Final Resultsโ€™ tab in light blue (see Figure 19). d) Look for values for the current assets and current liabilities in the latest year and calculate the current ratio. 13) To check if the companiesโ€™ forecast 1 year EPS is greater than that of the FTSE100: a) Click on โ€˜UK Marketโ€™ link under โ€˜Select a Marketโ€™ menu. b) Click on โ€˜FTSE 100โ€™ link (see Figure 20). 14) If all mandatory requirements are satisfied, execute the trade by using the portfolio features on the website. Otherwise repeat steps 4 to 14 until all mandatory requirements are satisfied.

59

A.5.2. Value Investing 1) On the Digital Look website, click on โ€˜Screening Toolsโ€™ link under โ€˜Investor Toolboxโ€™ menu. Then click on Screener. 2) Click on โ€˜full fundamental screenerโ€™ link under โ€˜Getting Startedโ€™. 3) Enter the relevant values in the โ€˜Minโ€™ field for the following: 30 for revenue (in ยฃm). 33 for EPS growth (average) (in percent) and select 4 years in the drop down menu on the right hand side. The former is under โ€˜Profit/Loss Accountโ€™ and the latter is under โ€˜Growth Ratiosโ€™. 4) Enter the values in the โ€˜Maxโ€™ fields for the following: 1.5 for P/BV and 15 for P/E. 5) Click on โ€˜Run Screenโ€™ at the bottom of the webpage. 6) Click on any company.

Figure 21: Example result of application of steps 1 to 5 for Value Investing.

7) To check if the company satisfies other criteria: a) Click on link of a company. This loads the summary page. b) Click on โ€˜Financialsโ€™ tab on the company webpage. c) Click on โ€˜Fundamentalsโ€™. This loads the final results accounts for the company. This is indicated by the highlight of the โ€˜Final Resultsโ€™ tab in light blue. d) For general and defensive strategies: i) Check if there are dividends paid and no earnings loss for the past five years under. ii) Current ratio is equal to at least 2. e) For defensive strategy: i) For manufacturers, check if liabilities due in more than one year are less than net current assets. ii) For utility companies, ensure that liabilities are less than double the net asset value. f) For enterprising strategy check if: i) There are at least some dividends paid for the past five years under. ii) Current ratio is equal to at least 1.5. iii) For manufacturers, debt must be at most 110% of net current assets. 60

iv) Market capitalisation is less than 120% of net tangible assets. 8) If all mandatory requirements are satisfied, execute the trade by using the portfolio features on the website. Otherwise repeat steps 6 and 7 until all mandatory requirements are satisfied.

61

A.5.3. Zulu Principle 1) On the Digital Look website, click on โ€˜Screening Toolsโ€™ link under โ€˜Investor Toolboxโ€™ menu. Then click on Screener. 2) Click on โ€˜full fundamental screenerโ€™ link under โ€˜Getting Startedโ€™. 3) Enter values in the โ€˜Maxโ€™ fields for the following: 1 for PEG, 20 for P/E and 50% for net gearing. 4) Enter values in the โ€˜Minโ€™ fields for the following: 83.33 for RSI, 0 for P/S and P/BV. This is because to easily check if the bonus criteria are satisfied of the latter two. 5) Click on โ€˜Run Screenโ€™ at the bottom of the webpage. 6) Select any company.

Figure 22: Example of application of steps 1 to 5 for the Zulu Principle.

7) To check if the company satisfies other criteria: a) Click on link of a company. This loads the summary page. Market capitalisation can be found on the page. b) Click on โ€˜Financialsโ€™ tab on the company webpage. c) Click on โ€˜Fundamentalsโ€™. This loads the final results accounts for the company. This is indicated by the highlight of the โ€˜Final Resultsโ€™ tab in light blue. d) Check if the cash flow per share greater than EPS for last reported year and average for the last five years. e) Check if EPS increases every year for the past five years. f) On the screener results webpage, click on โ€˜Director Dealsโ€™ link to find out if a group of directors have bought and sold shares in the past month. g) Look at the latest annual reports. These contain information regarding brand names, industries the company operates in and other important information. 8) If all mandatory requirements are satisfied, execute the trade by using the portfolio features on the website. Otherwise repeat steps 6 and 7 until all mandatory requirements are satisfied.

62

A.5.4. Recording trades 1) On the Digital Look website, click on โ€˜Portfolios & Watchlistsโ€™ under โ€˜Investor Toolboxโ€™. This loads the Portfolio Valuation screen. 2) Click on โ€˜Add to portfolioโ€™. 3) Enter the name or ticker of the company. Select LSE in the โ€˜Stock Exchangeโ€™ drop-down menu (see Figure 23).

Figure 23: 'Portfolio Selection' window with 'Portfolio Valuation' webpage in the background.

Figure 24: 'Trade Security' webpage to record transactions.

63

Click on โ€˜Searchโ€™. Select the required company. Click on โ€˜add sharesโ€™. On the โ€˜Trade Securityโ€™ webpage (see Figure 24), enter the required number of shares taking into account of the bid, or offer price, commission of ยฃ10, stamp duty and, if the value of shares alone is over ยฃ10,000, PTM levy of ยฃ1. If over ยฃ15,000, ยฃ1.50. Also enter the date of trade. 8) Select โ€˜Buyโ€™ or โ€˜Sellโ€™ in the drop-down menu under โ€˜Actionโ€™, depending on the type of transaction. 4) 5) 6) 7)

64

A.5.5. Executing trades 1) On the BullBearings website, click on โ€˜Portfolio summaryโ€™. 2) Click on โ€˜Fantasy Stocksโ€™ (see figure 25).

Figure 25: 'Portfolio summary' webpage displaying types of portfolios with their values.

3) Click on โ€˜PLACE A TRADEโ€™ link on the โ€˜Share Trading Portfolioโ€™ webpage.

Figure 26: โ€˜Share Trading Portfolioโ€™ webpage showing stocks held in the portfolio with 'PLACE A TRADE' link above.

65

4) One the โ€˜Place a tradeโ€™ webpage, enter the company name or ticker in the relevant field and press enter (see Figure 27). 5)

Figure 27: Illustration of step 4.

6) Since the ยฃ100,000 is split into equal proportions, enter the cash amount equal to that in โ€˜Cash Amountโ€™ field. For example, an investor has ยฃ100,000 to invest in 10 stocks in equal proportions so ยฃ10,000 would be spent on each stock. Select โ€˜At Bestโ€™ and click on โ€˜Nextโ€™ (see Figure 28).

Figure 28: 'Stock Trading' webpage illustrating an example of a trade to buy ยฃ10,000 in Royal Bank of Scotland Group plc.

66

A.6. Experiment Data
A.6.1. Data of companies matching relevant criteria for Investment Strategies
Table 9: Companies with the relevant ratios which satisfy at least all compulsory critera for Contrarian Investment Strategy
Company Ratios Company Ticker P/E Beazley Group Elementis GETECH Group Inchcape Kazakhmys KCOM Group Raymarine Royal Dutch Shell 'A' SIG Titan Europe Vitec Group Yell Group BEZ ELM GTC INCH KAZ KCOM RAY RDSA SHI TSW VTC YELL 4.9 2.6 7.8 1 1.2 3.7 0.8 4.8 1.4 0.7 2.8 0.4 P/BV 0.78 0.4 7.65 0.2 0.31 0.48 0.49 0.57 0.25 0.03 0.51 0.07 P/CF 0.6 2.21 4.21 43.15 10.84 1.76 0.37 1.67 1.37 0.3 1.99 0.36 Dividend Yield 7.22% 12.08% 7.65% 0.84% 1.71% 16.59% 25.92% 7.73% 24.84% 50.00% 11.49% 77.42% ROCE 15.17% 18.34% 17.50% 31.20% 35.82% 10.87% 40.80% 29.70% 30.76% 11.15% 20.89% 71.28% P/S 0.37 0.3 1.2 0.03 0.4 0.2 0.1 0.2 0.1 0.03 0.2 0.05 ROA 2.15% 3.53% 9.17% 6.72% 19.25% 3.39% 5.29% 9.30% 5.55% 3.11% 6.54% 3.04% Current Ratio 1.2 1.94 2 1.2 6.05 1 1.93 1.1 2.72 1.23 1.41 1.31 Dividend cover 2.85 3.14 1.7 2.35 7.41 1.62 4.9 2.7 2.8 2.8 3.05 3.14 0.1 0.02 0.1 0.08 0.1 0.4 0.3 PEG

All companies have P/E, P/BV, P/CF and dividend yield values which are below the sector counterparts (see Table 3) at the time of purchase. They are also in the two lowest quintiles for the first three ratios and the top two for dividend yield. Note that the 1 year forecase EPS growth rate of 39.7% for Titan Europe is greater than that of FTSE 100, which the rate is 19.7%. Table 10: Sector average values for the Contrarian ratios
Sector Ratios Ticker BEZ ELM GTC INCH KAZ KCOM RAY RDSA SHI TSW VTC YELL Sector in which the Company is listed in P/E Non-life Insurance Chemicals Oil Equipment & Services General Retailers Mining Fixed Line Telecommunications Electronic & Electrical Equipment Oil & Gas Producers Support Services General Industrials Industrial Engineering Media 11.78 10.17 15 11.12 25.59 9.3 11.68 234.8 8.49 11.79 1083.89 8.68 P/BV 1.54 1.47 1.57 1.62 1.71 1.45 1.58 1.08 2.23 0.62 1.16 1.61 P/CF 6.04 5.59 7.95 3.5 18.73 23.42 7 10.1 2.6 29.75 3.54 14.91 Dividend Yield 6.90% 9.70% 5.10% 15.70% 7.80% 12.00% 9.90% 3.60% 8.20% 14% 10.30% 15.50%

67

Table 11: Companies with the relevant ratios which satisfy at least all compulsory criteria for Value Investing
Company Accident Exchange Group Aero Inventory Blacks Leisure Bloomsbury Publishing Delta Jersey Electricity 'A' Lonmin Ticker P/E Average P/E (4 Years) 0.9 0.86 P/BV P/BV x 2 Current Ratio 2.59 Market Capitalisation ยฃ0.1500

ACE

0.15

0.3

AI.

2.2

3.26

0.42

0.84

4.91

ยฃ1.5600

BSLA

9.1

146

0.2

0.4

1.50

ยฃ0.3950

BMY

8.4

9.78

0.99

2.34

2.34

ยฃ1.3475

DTLA

7.6

13.7

0.62

1.24

4.82

ยฃ1.0500

JEL

10.5

10

0.55

1.1

2.80

ยฃ67.0000

LMI

6.3

13.23

1.35

2.7

2.25

ยฃ1.6300

Man Group Ocean Wilsons

EMG

3.3

11.2

1.12

2.24

3.55

ยฃ2.4300

OCN

1.1

4.07

0.46

0.92

5.72

ยฃ5.4750

Market Capitalisation x 1.2 Accident Exchange Group Aero Inventory Blacks Leisure Bloomsbury Publishing Delta Jersey Electricity 'A' Lonmin ACE ยฃ0.1800

Net Tangible Assets ยฃ0.6621

Net Tangible Assets x 1.1 ยฃ0.7283

Revenue (m)

Net Assets (m)

Total Liabilities (m)

ยฃ114.58

AI.

ยฃ1.8720

ยฃ3.1999

ยฃ3.5199

ยฃ305.56

BSLA

ยฃ0.4740

ยฃ1.0930

ยฃ1.2023

ยฃ294.41

BMY

ยฃ1.6170

ยฃ1.1170

ยฃ1.2287

ยฃ150.00

DTLA

ยฃ1.2600

ยฃ1.6434

ยฃ1.8077

ยฃ266.40

ยฃ258.80

ยฃ63.20

JEL

ยฃ80.4000

ยฃ121.0818

ยฃ133.1900

82.00

ยฃ133.26

ยฃ39.24

LMI

ยฃ1.9560

ยฃ6.8063

ยฃ7.4869

ยฃ15.49

Man Group Ocean Wilsons

EMG

ยฃ2.9160

ยฃ1.2905

ยฃ1.4196

ยฃ310.56

OCN

ยฃ6.5700

ยฃ11.7410

ยฃ12.9151

ยฃ280.56

Note that the net assets and total liabilities only apply to Delta and Jersey Electricity, which the former is a manufacturer and the latter is a utility

68

Table 12: Criteria satisfied by companies for respective sub-strategies for Value Investing Criteria numbers satisfied Ticker ACE AI. BSLA BMY DTLA JEL LMI EMG OCN General All All 1, 3, 4, 5, 6 All All All All All All Defensive All All 1, 3, 4, 5, 6 All All All All All All Enterprising All All All 1,2,3 All All All 1,2,3 All

Table 13: Companies with the relevant ratios which satisfy at least all compulsory criteria for the Zulu Principle Company Cash Converters Dewhurst Jersey Electricity 'A' Prologic Visonic Zytronic Ticker CCVU DWHT JEL PGC VSC ZYT PEG 0.3 0.1 0.4 0.1 0.01 0.1 P/E 5.8 4.8 10.9 2.7 7 14.8 EPS 6.28AUยข 38.92p 641p 13.24p 8.9USยข 8.4p Average EPS (5 year) 3.92AUยข 27.78p 373.92p 8.8p1 9.38USยข 5.02p CF per share 13.12AUยข 92.56p 1359.34p 24.6p 8.12USยข 13.55p RSI 88.24% 100% 100% 100% 100% 90% Net Gearing 12.25% 22.85% 25.53% 34.74% 17.99% 47.69%

Current Ratio Cash Converters Dewhurst Jersey Electricity 'A' Prologic Visonic Zytronic CCVU DWHT JEL PGC VSC ZYT 2.4 2.98 2.8 1.17 2.4 1.92

Quick Ratio 2.2 2.23 2.31 1.16 1.7 1.22

P/S 0.7 0.8 0.94 0.31 0.3 1.24

Dividend Yield 8.2% 3.1% 3.3% 5.6% 3.4% 3.2%

P/BV 0.7 0.3 0.58 0.32 0.57 2.05

P/R

3.17 4.76

1

2 year average EPS was used.

69

Table 14: Information which satisfies the Zulu Principle Ticker CCVU DWHT JEL PGC ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท ๏‚ท Other Information Strong brand Well established in pawn broking Opening a new branch in Norwich Specialist in input devices New product launches Legal monopoly of electricity market in Jersey Niche market โ€“ provides computer systems and consultancy to business in fashion and lifestyle sector Directors bought shares on 31/3/2009 Stock taking time drops by 50% at FatFace after new system goes live Strong brand in the security market In possession of 75 unique patents 45 patents pending Niche market โ€“ specialists in touch screens, laminated products and processes. In possession of patents for the above Directors bought shares on 20/2/2009 and 30/3/2009. Table 15: Parts of the Zulu Principle satisfied by companies Ticker CCVU DWHT JEL PGC VSC ZYT Criteria numbers satisfied Mandatory All All All All All All Highly Desirable 1, 4 1, 4 1, 2, 4 1, 2, 4 1, 4 1, 2, 4 Bonus 1, 2 1, 2, 4 1 All 1 2

VSC

ZYT

70

A.6.2. Portfolio Transaction Costs
Table 16: Transaction costs for the Contrarian Portfolio
Buy Date 20/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 20/03/2009 09/03/2009 09/03/2009 Ticker BEZ ELM GTC INCH KCOM KAZ RAY RDSA SHI TSW VTC YELL Quantity 15,341 40,990 53,731 26,158 56,801 3,768 77,963 657 9,161 104,147 6,098 65,182 Price ยฃ0.8825 ยฃ0.2425 ยฃ0.1850 ยฃ0.3800 ยฃ0.1750 ยฃ2.6375 ยฃ0.1275 ยฃ15.1100 ยฃ1.0375 ยฃ0.1300 ยฃ1.6300 ยฃ0.1525 Value excluding Stamp Duty ยฃ13,538.43 ยฃ9,940.08 ยฃ9,940.24 ยฃ9,940.04 ยฃ9,940.18 ยฃ9,938.10 ยฃ9,940.28 ยฃ9,927.27 ยฃ9,504.54 ยฃ13,539.11 ยฃ9,939.74 ยฃ9,940.26 Stamp Duty ยฃ67.69 ยฃ49.70 ยฃ49.70 ยฃ49.70 ยฃ49.70 ยฃ49.69 ยฃ49.70 ยฃ49.64 ยฃ47.52 ยฃ67.70 ยฃ49.70 ยฃ49.70 Value ยฃ13,606.12 ยฃ9,989.78 ยฃ9,989.94 ยฃ9,989.74 ยฃ9,989.88 ยฃ9,987.79 ยฃ9,989.98 ยฃ9,976.91 ยฃ9,552.06 ยฃ13,606.81 ยฃ9,989.44 ยฃ9,989.96 PTM Levy ยฃ1.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ13,617.12 ยฃ9,999.78 ยฃ9,999.94 ยฃ9,999.74 ยฃ9,999.88 ยฃ9,997.79 ยฃ9,999.98 ยฃ9,986.91 ยฃ9,562.06 ยฃ13,617.81 ยฃ9,999.44 ยฃ9,999.96

Sell Date 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 Ticker BEZ ELM GTC INCH KCOM KAZ RAY RDSA SHI TSW VTC YELL Quantity 15,341 40,990 53,731 26,158 56,801 3,768 77,963 657 9,161 104,147 6,098 65,182 Price ยฃ0.8800 ยฃ0.2500 ยฃ0.1500 ยฃ0.6775 ยฃ0.1700 ยฃ3.6800 ยฃ0.1100 ยฃ15.6800 ยฃ1.0375 ยฃ0.14 ยฃ1.5500 ยฃ0.1375 Value excluding Stamp Duty ยฃ13,500.08 ยฃ10,247.50 ยฃ8,059.65 ยฃ17,722.05 ยฃ9,656.17 ยฃ13,866.24 ยฃ8,575.93 ยฃ10,301.76 ยฃ9,504.54 ยฃ14,059.85 ยฃ9,451.90 ยฃ8,962.53 Stamp Duty ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Value ยฃ13,500.08 ยฃ10,247.50 ยฃ8,059.65 ยฃ17,722.05 ยฃ9,656.17 ยฃ13,866.24 ยฃ8,575.93 ยฃ10,301.76 ยฃ9,504.54 ยฃ14,059.85 ยฃ9,451.90 ยฃ8,962.53 PTM Levy ยฃ1.00 ยฃ1.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ13,489.08 ยฃ10,236.50 ยฃ8,049.65 ยฃ17,710.55 ยฃ9,646.17 ยฃ13,855.24 ยฃ8,565.93 ยฃ10,290.76 ยฃ9,494.54 ยฃ14,048.85 ยฃ9,441.90 ยฃ8,952.53

Total Spent Total Sold

ยฃ126,780.39 ยฃ133,781.68

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,545.46 ยฃ106,576.60

71

Table 17: Transaction costs for the Value Investing Portfolio
Buy Date 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 26/03/2009 Ticker ACE AI. BSLA BMY JEL LMI EMG OCN DTLA Quantity 77,667 7,815 31,460 9,205 172 813 5,773 2,239 10,756 Price ยฃ0.16 ยฃ1.59 ยฃ0.40 ยฃ1.35 ยฃ72.00 ยฃ15.28 ยฃ2.15 ยฃ5.55 ยฃ1.08 Value excluding Stamp Duty ยฃ12,426.72 ยฃ12,425.85 ยฃ12,426.70 ยฃ12,426.75 ยฃ12,384.00 ยฃ12,422.64 ยฃ12,426.38 ยฃ12,426.45 ยฃ11,562.70 Stamp Duty ยฃ62.13 ยฃ62.13 ยฃ62.13 ยฃ62.13 ยฃ61.92 ยฃ62.11 ยฃ62.13 ยฃ62.13 ยฃ57.81 Value ยฃ12,488.85 ยฃ12,487.98 ยฃ12,488.83 ยฃ12,488.88 ยฃ12,445.92 ยฃ12,484.75 ยฃ12,488.51 ยฃ12,488.58 ยฃ11,620.51 PTM Levy ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ12,499.85 ยฃ12,498.98 ยฃ12,499.83 ยฃ12,499.88 ยฃ12,456.92 ยฃ12,495.75 ยฃ12,499.51 ยฃ12,499.58 ยฃ11,631.51

Sell Date 26/03/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 Ticker OCN ACE AI. BSLA BMY JEL LMI EMG DTLA Quantity 2239 77,667 7,815 31,460 9,205 172 813 5,773 10,756 Price ยฃ5.20 ยฃ0.11 ยฃ1.50 ยฃ0.33 ยฃ1.20 ยฃ65.00 ยฃ14.86 ยฃ2.47 ยฃ1.15 Value excluding Stamp Duty ยฃ11,642.80 ยฃ8,155.04 ยฃ11,722.50 ยฃ10,381.80 ยฃ11,046.00 ยฃ11,180.00 ยฃ12,081.18 ยฃ14,230.45 ยฃ12,369.40 Stamp Duty ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Value ยฃ11,642.80 ยฃ8,155.04 ยฃ11,722.50 ยฃ10,381.80 ยฃ11,046.00 ยฃ11,180.00 ยฃ12,081.18 ยฃ14,230.45 ยฃ12,369.40 PTM Levy ยฃ1.00 ยฃ0.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 ยฃ1.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ11,631.80 ยฃ8,145.04 ยฃ11,711.50 ยฃ10,370.80 ยฃ11,035.00 ยฃ11,169.00 ยฃ12,070.18 ยฃ14,219.45 ยฃ12,358.40

Total Spent Total Sold

ยฃ111,581.83 ยฃ102,711.16

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,950.32 ยฃ91,079.36

72

Table 18: Transaction costs for the Zulu Principle Portfolio
Buy Date 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 Ticker CCVU DWHT JEL PGC VSC ZYT Quantity 87,222 8,722 220 41,430 36,827 12,947 Price ยฃ0.19 ยฃ1.90 ยฃ75.00 ยฃ0.40 ยฃ0.45 ยฃ1.28 Value excluding Stamp Duty ยฃ16,572.18 ยฃ16,571.80 ยฃ16,500.00 ยฃ16,572.00 ยฃ16,572.15 ยฃ16,572.16 Stamp Duty ยฃ82.86 ยฃ82.86 ยฃ82.50 ยฃ82.86 ยฃ82.86 ยฃ82.86 Value ยฃ16,655.04 ยฃ16,654.66 ยฃ16,582.50 ยฃ16,654.86 ยฃ16,655.01 ยฃ16,655.02 PTM Levy ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ16,666.54 ยฃ16,666.16 ยฃ16,594.00 ยฃ16,666.36 ยฃ16,666.51 ยฃ16,666.52

Sell Date 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 Ticker CCVU DWHT JEL PGC VSC ZYT Quantity 87,222 8,722 220 41,430 36,827 12,947 Price ยฃ0.17 ยฃ1.82 ยฃ65.00 ยฃ0.32 ยฃ0.42 ยฃ1.22 Value excluding Stamp Duty ยฃ14,827.74 ยฃ15,874.04 ยฃ14,300.00 ยฃ13,257.60 ยฃ15,467.34 ยฃ15,795.34 Stamp Duty ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Value ยฃ14,827.74 ยฃ15,874.04 ยฃ14,300.00 ยฃ13,257.60 ยฃ15,467.34 ยฃ15,795.34 PTM Levy ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 ยฃ1.50 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ14,816.24 ยฃ15,862.54 ยฃ14,288.50 ยฃ13,246.10 ยฃ15,455.84 ยฃ15,783.84

Total Spent Total Sold

ยฃ99,926.09 ยฃ89,453.06

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,926.09 ยฃ89,453.06

73

Table 19: Transaction costs for the Optimal Portfolio 1
Buy Date 09/03/2009 09/03/2009 09/03/2009 09/03/2009 09/03/2009 Ticker ABC AD. ADV AEC RGN Quantity 44 327,127 11 671 3,779,278 Price ยฃ5.6750 ยฃ0.2300 ยฃ1.4000 ยฃ0.2000 ยฃ0.0063 Value excluding Stamp Duty ยฃ249.70 ยฃ75,239.21 ยฃ15.40 ยฃ134.20 ยฃ23,809.45 Stamp Duty ยฃ1.25 ยฃ376.20 ยฃ0.08 ยฃ0.67 ยฃ119.05 Value ยฃ250.95 ยฃ75,615.41 ยฃ15.48 ยฃ134.87 ยฃ23,928.50 PTM Levy ยฃ0.00 ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ1.50 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ260.95 ยฃ75,626.91 ยฃ25.48 ยฃ144.87 ยฃ23,940.00

Sell Date 20/03/2009 20/03/2009 20/03/2009 20/03/2009 20/03/2009 Ticker ABC AD. ADV AEC RGN Quantity 44 327,127 11 671 3,779,278 Price ยฃ5.7750 ยฃ0.1800 ยฃ1.1000 ยฃ0.1700 ยฃ0.0023 Value excluding Stamp Duty ยฃ254.10 ยฃ58,882.86 ยฃ12.10 ยฃ114.07 ยฃ8,692.34 Stamp Duty ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Value ยฃ254.10 ยฃ58,882.86 ยฃ12.10 ยฃ114.07 ยฃ8,692.34 PTM Levy ยฃ0.00 ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ1.50 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ244.10 ยฃ58,871.36 ยฃ2.10 ยฃ104.07 ยฃ8,680.84

Total Spent Total Sold

ยฃ99,926.09 ยฃ89,453.06

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,926.09 ยฃ89,453.06

74

Table 20: Transaction costs for the Optimal Portfolio 2
Buy Date 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 23/03/2009 Ticker ABC ACC AD. ASW AMS ADV Quantity 79 42,454 395,178 7,342 10,652 1,316 Price ยฃ6.4000 ยฃ0.0263 ยฃ0.2300 ยฃ0.2175 ยฃ0.3275 ยฃ1.4000 Value excluding Stamp Duty ยฃ505.60 ยฃ1,114.42 ยฃ90,890.94 ยฃ1,596.89 ยฃ3,488.53 ยฃ1,842.40 Stamp Duty ยฃ2.53 ยฃ5.57 ยฃ454.45 ยฃ7.98 ยฃ17.44 ยฃ9.21 Value ยฃ508.13 ยฃ1,120.00 ยฃ91,345.39 ยฃ1,604.87 ยฃ3,505.97 ยฃ1,851.61 PTM Levy ยฃ0.00 ยฃ0.00 ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ518.13 ยฃ1,129.99 ยฃ91,356.89 ยฃ1,614.87 ยฃ3,515.97 ยฃ1,861.61

Sell Date 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 03/04/2009 Ticker ABC ACC AD. ASW AMS ADV Quantity 79 42,454 395,178 7,342 10,652 1,316 Price ยฃ5.5900 ยฃ0.0238 ยฃ0.1800 ยฃ0.2025 ยฃ0.3250 ยฃ1.2750 Value excluding Stamp Duty ยฃ441.61 ยฃ1,008.28 ยฃ71,132.04 ยฃ1,486.76 ยฃ3,461.90 ยฃ1,677.90 Stamp Duty ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ1.00 ยฃ0.00 Value ยฃ441.61 ยฃ1,008.28 ยฃ71,132.04 ยฃ1,486.76 ยฃ3,462.90 ยฃ1,677.90 PTM Levy ยฃ0.00 ยฃ0.00 ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ431.61 ยฃ998.28 ยฃ71,120.54 ยฃ1,476.76 ยฃ3,452.90 ยฃ1,667.90

Total Spent Total Sold

ยฃ99,997.47 ยฃ79,147.99

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,997.47 ยฃ79,147.99

75

Table 21: Transaction costs for the Optimal Portfolio 3
Buy Date 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 06/04/2009 Ticker AD. ADM ASW ADV AEC ABM ADS ANR Quantity 383,848 44 9,518 1,608 8,079 1,217 212,146 5,510 Price ยฃ0.2300 ยฃ8.8750 ยฃ0.2325 ยฃ1.5750 ยฃ0.2000 ยฃ2.0400 ยฃ0.0080 ยฃ0.0372 Value excluding Stamp Duty ยฃ88,285.04 ยฃ390.50 ยฃ2,212.94 ยฃ2,532.60 ยฃ1,615.80 ยฃ2,482.68 ยฃ1,686.56 ยฃ204.97 Stamp Duty ยฃ441.43 ยฃ1.95 ยฃ11.06 ยฃ12.66 ยฃ8.08 ยฃ12.41 ยฃ8.43 ยฃ1.02 Value ยฃ88,726.47 ยฃ392.45 ยฃ2,224.00 ยฃ2,545.26 ยฃ1,623.88 ยฃ2,495.09 ยฃ1,694.99 ยฃ206.00 PTM Levy ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ88,737.97 ยฃ402.45 ยฃ2,234.00 ยฃ2,555.26 ยฃ1,633.88 ยฃ2,505.09 ยฃ1,704.99 ยฃ216.00

Sell Date 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 17/04/2009 Ticker AD. ADM ASW ADV AEC ABM ADS ANR Quantity 383,848 44 9,518 1,608 8,079 1,217 212,146 5,510 Price ยฃ0.1700 ยฃ9.1650 ยฃ0.2350 ยฃ1.2500 ยฃ0.1600 ยฃ1.8900 ยฃ0.0045 ยฃ0.0352 Value excluding Stamp Duty ยฃ65,254.16 ยฃ403.26 ยฃ2,236.73 ยฃ2,010.00 ยฃ1,292.64 ยฃ2,300.13 ยฃ944.05 ยฃ193.95 Stamp Duty ยฃ326.27 ยฃ2.02 ยฃ11.18 ยฃ10.05 ยฃ6.46 ยฃ11.50 ยฃ4.72 ยฃ0.97 Value ยฃ65,580.43 ยฃ405.28 ยฃ2,247.91 ยฃ2,020.05 ยฃ1,299.10 ยฃ2,311.63 ยฃ948.77 ยฃ194.92 PTM Levy ยฃ1.50 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 ยฃ0.00 Commission ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 ยฃ10.00 Total ยฃ65,568.93 ยฃ395.28 ยฃ2,237.91 ยฃ2,010.05 ยฃ1,289.10 ยฃ2,301.63 ยฃ938.77 ยฃ184.92

Total Spent Total Sold

ยฃ99,989.64 ยฃ74,926.60

Total Spent at start of holding period Total Sold at end of holding period

ยฃ99,989.64 ยฃ74,926.60

76

A.6.3. Portfolio Valuation at the end of Holding Period
Table 22: Portfolio valuation of Contrarian Portfolio at the end of the holding period before selling
Security Name Beazley Group Elementis GETECH Group KCOM Group Kazakhmys Raymarine Royal Dutch Shell 'A' Titan Europe Vitec Group Yell Group Totals Ticker BEZ ELM GTC KCOM KAZ RAY RDSA TSW VTC YELL Quantity 15,341 40,990 53,731 56,801 3,768 77,963 657 104,147 6,098 65,182 Market Price ยฃ0.8800 ยฃ0.2500 ยฃ0.1650 ยฃ0.1725 ยฃ3.6850 ยฃ0.1125 ยฃ15.6800 ยฃ0.1350 ยฃ1.5500 ยฃ0.1375 Holding Value ยฃ13,500.08 ยฃ10,247.50 ยฃ8,865.61 ยฃ9,798.17 ยฃ13,885.08 ยฃ8,770.84 ยฃ10,301.76 ยฃ14,059.85 ยฃ9,451.90 ยฃ8,962.53 ยฃ107,843.32 Profit / Loss -ยฃ38.35 ยฃ307.42 -ยฃ1,074.62 -ยฃ142.00 ยฃ3,946.98 -ยฃ1,169.44 ยฃ374.49 ยฃ520.74 -ยฃ487.84 -ยฃ977.73 ยฃ1,259.65

Table 23: Portfolio valuation of Value Investing Portfolio at the end of the holding period before selling
Security Name Accident Exchange Group Aero Inventory Blacks Leisure Bloomsbury Publishing Delta Jersey Electricity 'A' Lonmin Man Group Totals Ticker ACE AI. BSLA BMY DTLA JEL LMI EMG Quantity 77,667 7,815 31,460 9,205 10,756 172 813 5,773 Market Price ยฃ0.1175 ยฃ1.5375 ยฃ0.3450 ยฃ1.2000 ยฃ1.1600 ยฃ70.0000 ยฃ14.8500 ยฃ2.4675 Holding Value ยฃ9,125.87 ยฃ12,015.56 ยฃ10,853.70 ยฃ11,046.00 ยฃ12,476.96 ยฃ12,040.00 ยฃ12,073.05 ยฃ14,244.88 ยฃ93,876.02 Profit / Loss -ยฃ3,300.85 -ยฃ410.29 -ยฃ1,573.00 -ยฃ1,380.75 ยฃ914.26 -ยฃ344.00 -ยฃ349.59 ยฃ1,818.50 -ยฃ4,625.72

Table 24: Portfolio valuation of Zulu Principle Portfolio at the end of the holding period before selling
Security Name Cash Converters Dewhurst Jersey Electricity 'A... Prologic Visonic Zytronic Totals Ticker CCVU DWHT JEL PGC VSC ZYT Quantity 87,225 8,722 220 41,432 36,828 12,947 Market Price ยฃ0.1800 ยฃ1.8600 ยฃ70.0000 ยฃ0.3600 ยฃ0.4350 ยฃ1.2700 Holding Value ยฃ15,700.50 ยฃ16,222.92 ยฃ15,400.00 ยฃ14,915.52 ยฃ16,020.18 ยฃ16,442.69 ยฃ94,701.81 Profit / Loss -ยฃ872.25 -ยฃ348.88 -ยฃ1,100.00 -ยฃ1,657.28 -ยฃ552.42 -ยฃ129.47 -ยฃ4,660.30

77

Table 25: Portfolio valuation of Optimal Portfolio 1 at the end of the holding period before selling
Security Name Abcam ADL Advent Capital (Holdings) AEC Education Regenesis Group Totals Ticker ABC AD. ADV AEC RGN Quantity 44 327,127 11 671 3,779,278 Market Price ยฃ6.0750 ยฃ0.2050 ยฃ1.2500 ยฃ0.1850 ยฃ0.0043 Holding Value ยฃ267.30 ยฃ67,061.04 ยฃ13.75 ยฃ124.14 ยฃ16,250.90 ยฃ83,717.12 Profit / Loss ยฃ6.35 -ยฃ8,565.87 -ยฃ11.73 -ยฃ20.74 -ยฃ7,689.10 -ยฃ16,281.09

Table 26: Portfolio valuation of Optimal Portfolio 2 at the end of the holding period before selling
Security Name Abcam Access Intelligence ADL Advanced Computer Software Advanced Medical Solutions Group Advent Capital (Holdings) Totals Ticker ABC ACC AD. ASW AMS ADV Quantity 79 424,543 395,178 7,342 10,652 1,316 Market Price ยฃ5.8900 ยฃ0.0025 ยฃ0.2050 ยฃ0.2150 ยฃ0.3625 ยฃ1.4250 Holding Value ยฃ465.31 ยฃ1,061.36 ยฃ81,011.49 ยฃ1,578.53 ยฃ3,861.35 ยฃ1,875.30 ยฃ89,853.34 Profit / Loss -ยฃ52.82 -ยฃ68.64 -ยฃ10,345.40 -ยฃ36.34 ยฃ345.38 ยฃ13.69 -ยฃ10,144.14

Table 27: Portfolio valuation of Optimal Portfolio 3 at the end of the holding period before selling
Security Name ADL Admiral Group Advanced Computer Software Advent Capital (Holdings) AEC Education Albemarle and Bond Holdings Alexander David Securities GRP Altona Energy Totals Ticker AD. ADM ASW ADV AEC ABM ADS ANR Quantity 383,848 44 9,518 1,608 8,079 1,217 212,146 5,510 Market Price ยฃ0.1950 ยฃ9.1700 ยฃ0.2475 ยฃ1.4000 ยฃ0.1750 ยฃ1.9200 ยฃ0.0062 ยฃ0.0362 Holding Value ยฃ74,850.36 ยฃ403.48 ยฃ2,355.71 ยฃ2,251.20 ยฃ1,413.83 ยฃ2,336.64 ยฃ1,315.31 ยฃ199.46 ยฃ85,125.98 Profit / Loss -ยฃ13,887.61 ยฃ1.03 ยฃ121.71 -ยฃ304.06 -ยฃ220.05 -ยฃ168.45 -ยฃ389.69 -ยฃ16.53 -ยฃ14,863.67

78

A.6.4. Market, Bid and Offer prices
Table 28: Market, Bid and Offer prices at the time of sale of stocks for the Contrarian Portfolio Name Beazley Group Elementis GETECH Group Inchcape Kazakhmys KCOM Group Raymarine Royal Dutch Shell 'A' SIG Titan Europe Vitec Group Yell Group
Ticker

Bid ยฃ0.8800 ยฃ0.2500 ยฃ0.1500 ยฃ0.6775 ยฃ0.1700 ยฃ3.6800 ยฃ0.1100 ยฃ15.6800 ยฃ1.0375 ยฃ0.1350 ยฃ1.5500 ยฃ0.1375

Mid ยฃ0.8800 ยฃ0.2500 ยฃ0.1650 ยฃ0.6775 ยฃ0.1725 ยฃ3.6850 ยฃ0.1125 ยฃ15.6800 ยฃ1.0400 ยฃ0.1350 ยฃ1.5500 ยฃ0.1375

Offer ยฃ0.8825 ยฃ0.2525 ยฃ0.1700 ยฃ0.6800 ยฃ0.1800 ยฃ3.6875 ยฃ0.1150 ยฃ15.6900 ยฃ1.0500 ยฃ0.1550 ยฃ1.5750 ยฃ0.1400

Spread ยฃ0.0025 ยฃ0.0025 ยฃ0.0200 ยฃ0.0025 ยฃ0.0100 ยฃ0.0075 ยฃ0.0050 ยฃ0.0100 ยฃ0.0125 ยฃ0.0200 ยฃ0.0250 ยฃ0.0025

BEZ ELM GTC INCH KCOM KAZ RAY RDSA SHI TSW VTC YELL

Table 29: Market, Bid and Offer prices at the time of sale of stocks for the Value Investing Portfolio Name Accident Exchange Group Aero Inventory Blacks Leisure Bloomsbury Publishing Delta Jersey Electricity 'A' Lonmin Man Group Ocean Wilsons AI. BSLA BMY DTLA JEL LMI EMG OCN Ticker ACE Bid ยฃ0.1050 ยฃ1.5000 ยฃ0.3300 ยฃ1.2000 ยฃ1.1500 ยฃ65.0000 ยฃ14.8600 ยฃ2.4650 ยฃ5.2000 Mid ยฃ0.1175 ยฃ1.5375 ยฃ0.3450 ยฃ1.2000 ยฃ1.1600 ยฃ70.0000 ยฃ14.8500 ยฃ2.4675 ยฃ5.2750 Offer ยฃ0.1350 ยฃ1.5450 ยฃ0.3650 ยฃ1.2250 ยฃ1.1525 ยฃ75.0000 ยฃ14.8700 ยฃ2.4625 ยฃ5.3500 Spread ยฃ0.0300 ยฃ0.0150 ยฃ0.0350 ยฃ0.0250 ยฃ0.0025 ยฃ10.0000 ยฃ0.0100 ยฃ0.0025 ยฃ0.1500

Table 30: Market, Bid and Offer prices at the time of sale of stocks for the Zulu Principle Portfolio Name Cash Converters Dewhurst Jersey Electricity 'A' Prologic Visonic Zytronic Ticker CCVU DWHT JEL PGC VSC ZYT Bid ยฃ0.1750 ยฃ1.8150 ยฃ65.0000 ยฃ0.3200 ยฃ0.4200 ยฃ1.2550 Mid ยฃ0.1800 ยฃ1.8600 ยฃ70.0000 ยฃ0.3600 ยฃ0.4350 ยฃ1.2700 Offer ยฃ0.1900 ยฃ1.9050 ยฃ75.0000 ยฃ0.4000 ยฃ0.4400 ยฃ1.2850 Spread ยฃ0.0250 ยฃ0.0900 ยฃ10.0000 ยฃ0.0800 ยฃ0.0200 ยฃ0.0300

79

Table 31: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 1 Name Abcam ADL Advent Capital AEC Education Regenesis Group Ticker ABC AD. ADV AEC RGN Bid ยฃ5.7750 ยฃ0.1800 ยฃ1.1000 ยฃ0.1700 ยฃ0.0023 Mid ยฃ6.0750 ยฃ0.2050 ยฃ1.2500 ยฃ0.1850 ยฃ0.0043 Offer ยฃ6.1700 ยฃ0.2300 ยฃ1.4000 ยฃ0.2000 ยฃ0.0630 Spread ยฃ0.3950 ยฃ0.0500 ยฃ0.3000 ยฃ0.0300 ยฃ0.0040

Table 32: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 2 Name Abcam Access Intelligence ADL Advanced Computer Software Advanced Medical Solutions Advent Capital Ticker ABC ACC AD. ASW AMS ADV Bid ยฃ5.5900 ยฃ0.0024 ยฃ0.1800 ยฃ0.2025 ยฃ0.3250 ยฃ1.2750 Mid ยฃ5.8900 ยฃ0.0025 ยฃ0.2050 ยฃ0.2150 ยฃ0.3625 ยฃ1.4250 Offer ยฃ5.9850 ยฃ0.0099 ยฃ0.2300 ยฃ0.2325 ยฃ0.3350 ยฃ1.5750 Spread ยฃ0.3950 ยฃ0.0075 ยฃ0.0500 ยฃ0.0300 ยฃ0.0100 ยฃ0.3000

Table 33: Market, Bid and Offer prices at the time of sale of stocks for the Optimal Portfolio 3 Name ADL Admiral Group Advanced Computer Software Advent Capital AEC Education Albemarle & Bond Alexander David Securities Group Altona Energy Ticker AD. ADM ASW ADV AEC ABM ADS ANR Bid ยฃ0.1700 ยฃ9.1650 ยฃ0.2350 ยฃ1.2500 ยฃ0.1600 ยฃ1.8900 ยฃ0.0045 ยฃ0.0352 Mid ยฃ0.1950 ยฃ9.1700 ยฃ0.2475 ยฃ1.4000 ยฃ0.1750 ยฃ1.9200 ยฃ0.0062 ยฃ0.0362 Offer ยฃ0.2200 ยฃ9.1800 ยฃ0.2650 ยฃ1.5500 ยฃ0.1900 ยฃ1.9500 ยฃ0.0080 ยฃ0.0372 Spread ยฃ0.0500 ยฃ0.0150 ยฃ0.0300 ยฃ0.3000 ยฃ0.0300 ยฃ0.0600 ยฃ0.0035 ยฃ0.0020

80

A.6.5. Stocks Returns from Portfolios
Table 34: Stock Returns after transaction costs for the Contrarian Portfolio Company Name Beazley Group Elementis GETECH Group Inchcape KCOM Group Kazakhmys Raymarine Royal Dutch Shell 'A' SIG Titan Europe Vitec Group Yell Group BEZ ELM GTC INCH KCOM KAZ RAY RDSA SHI TSW VTC YELL Ticker Profit/Loss -ยฃ128.04 ยฃ236.72 -ยฃ1,950.29 ยฃ7,710.80 -ยฃ353.71 ยฃ3,857.45 -ยฃ1,434.05 ยฃ303.85 -ยฃ67.52 ยฃ431.04 -ยฃ557.54 -ยฃ1,047.43 ๐‘น๐’Š (%) -0.94% 2.37% -19.50% 77.11% -3.54% 38.58% -14.34% 3.04% -0.71% 3.17% -5.58% -10.47%

Table 35: Stock Returns after transaction costs for the Value Investing Portfolio Company Name Accident Exchange Group Aero Inventory Blacks Leisure Bloomsbury Publishing Delta Jersey Electricity 'A' Lonmin Man Group Ocean Wilsons ACE AI. BSLA BMY DTLA JEL LMI EMG OCN Ticker Profit/Loss -ยฃ4,354.82 -ยฃ787.48 -ยฃ2,129.03 -ยฃ1,464.88 ยฃ726.89 -ยฃ1,287.92 -ยฃ425.57 ยฃ1,719.93 -ยฃ867.78 ๐‘น๐’Š (%) -34.84% -6.30% -17.03% -11.72% 6.25% -10.34% -3.41% 13.76% -6.94%

Table 36: Stock Returns after transaction costs for the Zulu Principle Portfolio Company Name Cash Converters Dewhurst Jersey Electricity 'A' Prologic Visonic Zytronic Ticker CCVU DWHT JEL PGC VSC ZYT Profit/Loss -ยฃ1,850.30 -ยฃ803.62 -ยฃ2,305.50 -ยฃ3,420.26 -ยฃ1,210.67 -ยฃ882.68 ๐‘น๐’Š (%) -11.10% -4.82% -13.89% -20.52% -7.26% -5.30%

81

Table 37: Stock Returns after transaction costs for Optimal Portfolio 1 Company Name Abcam ADL Advent Capital (Holdings) AEC Education Regenesis Group ABC AD. ADV AEC RGN Ticker Profit/Loss -ยฃ16.85 -ยฃ16,755.55 -ยฃ23.38 -ยฃ40.80 -ยฃ15,259.16 ๐‘น๐’Š (%) -6.46% -22.16% -91.76% -28.16% -63.74%

Table 38: Stock Returns after transaction costs for Optimal Portfolio 2 Company Name Abcam Access Intelligence ADL Advanced Computer Software Advanced Medical Solutions Group Advent Capital (Holdings) ABC ACC AD. ASW AMS ADV Ticker Profit/Loss -ยฃ87.52 -ยฃ132.71 -ยฃ20,236.35 -ยฃ138.11 -ยฃ63.07 -ยฃ193.71 ๐‘น๐’Š (%) -16.89% -11.74% -22.15% -8.55% -1.79% -10.41%

Table 39: Stock Returns after transaction costs for Optimal Portfolio 3 Company Name ADL Admiral Group Advanced Computer Software Advent Capital (Holdings) AEC Education Albemarle and Bond Holdings Alexander David Securities GRP Altona Energy AD. ADM ASW ADV AEC ABM ADS ANR Ticker Profit/Loss -ยฃ23,169.03 -ยฃ7.18 ยฃ3.91 -ยฃ545.21 -ยฃ344.78 -ยฃ203.46 -ยฃ766.22 -ยฃ31.08 ๐‘น๐’Š (%) -26.11% -1.78% 0.18% -21.34% -21.10% -8.12% -44.94% -14.39%

82

A.7. MATLAB codes
function [R, ER, COV] = Statdata_and_find_eigenvalues(R); % Need to transpose returns data in order to get suitable solutions. This % is because we downloaded the data of stock prices with the weekly prices % in columns and stock in rows. Also If we try to download data with the stocks in columns, % MS Excel 2003 cannot fit 1914 stocks. R = transpose(R); % The following calculate the mean and variance from observed data, that is data that are not missing. ER = nanmean(R); COV = nancov(R); % Decomposes the matrix into a vector of eigenvalues. This is useful for % checking if the covariance matrix is positive semi-definite or not. Eig_of_COV = eig(COV); % Convert the covariance matrix into correlation matrix [ExpSigma,ExpCorrC] = cov2corr(COV); % Decompose correlation matrix into matrix of eigenvectors and diagonal % matrix of eigenvalues. [V,D] = eig(ExpCorrC); D = diag(D); % Find the vector of eigenvalues so that we can replace negative eigenvalues with their absolute values. This because we do not have sufficient experience of MATLAB to write a code which can do this automatically. end
Figure 29: MATLAB code for Statdata_and_find_eigenvalues.m.

83

function [COVa, Eig_of_COVa] = gettranscovmatrix(D, V, ExpCorrC); D = diag(diag(diag(D))); % Creates diagonal matrix from vector of eigenvalues (after replacing the negative ones with their absolute values manually). % The following computes the new correlation matrix and transforms it into % covariance matrix. ExpCorrC = V*D*transpose(V); Eig_of_ExpCorrC = eig(ExpCorrC); % We need to check manually if the matrix is positive semi-definite or not. COVa = corr2cov(ExpSigma, ExpCorrC); % Converts correlation matrix into covariance matrix % We need to check manually if the covariance matrix is positive semidefinite or not. Eig_of_COVa = eig(COVa); clear Eig_of_COV D D1 Eig_of_ExpCorrC Eig_of_COVa V R % Clears all unnecessary variables. end
Figure 30: MATLAB code for Gettranscovmatrix.m.

function optimalportolio(ER, COVa); % Define variables and specify the number of portfolios ExpReturn = ER; ExpCovariance = COVa; NumPorts = 20; % Compute efficient frontier. [PortRisk, PortReturn, PortWts] = portopt(ExpReturn, ExpCovariance, NumPorts); % Calculate optimal portfolio with riskless rate (that is the LIBOR in our % case) and without short sales. RisklessRate BorrowRate = = 0.007275; 0.007275;

[RiskyRisk, RiskyReturn, RiskyWts, RiskyFraction, OverallRisk, OverallReturn] = portalloc(PortRisk, PortReturn,PortWts, RisklessRate, BorrowRate); portalloc(PortRisk, PortReturn,PortWts, RisklessRate, BorrowRate); % Displays efficient frontier and optimal portfolio point. end
Figure 31: MATLAB code for Optimal_Portfolio.m.

84

A.8. Glossary of Terms
Asset โ€“ property which the company owns. Tangible assets are those that have a physical substance, such as buildings, equipment and workers, whereas intangible assets are those which do not exist, such as brand names and copyrights. Book value โ€“ valuation of the company if all of its assets are sold. Includes tangible and intangible assets. Capital Employed โ€“ measures the amount of money used to operate the company. Cash flow โ€“ measures the amount of cash goes in and out of the company. Current assets โ€“ assets of the company which is expected to be sold or otherwise used up in the near future, say within one year. Current liabilities โ€“ debts need to be paid off by the end of the company year. Current ratio โ€“ measures the ability of the company to pay off its debts. The formula is current ratio = Current assets โ„ Current liabilities. Dividend โ€“ payment from the net profits of the company. The investors are basically the owners of the company so they are entitled to take profits from a company. Dividend cover โ€“ measures the security of the dividend. The formula is dividend cover = Net profits โ„ Dividends = EPS โ„ DPS. Dividend per share (DPS) โ€“ The formula is DPS = Dividends โ„ Number of shares in issue. Dividend yield โ€“ percentage return from net profits per share, for example, owning 100 shares at ยฃ1 and a dividend of 5p per share is paid so the yield is 5%. Earnings per share (EPS) โ€“ The formula is EPS = Net profits โ„ Number of shares in issue. Gearing โ€“ the amount of loan the company has taken out. It is a percentage of net assets. This can either magnify profits and losses. Index โ€“ a measure of performance of a selection of stock in the stock market. Examples are the FTSE 100, FTSE 250, FTSE AIM 50 and Dow Jones Industrial Average. Market Capitalisation โ€“ notional value of the company listed on the stock market. Operating Profit โ€“ profit after costs from ongoing operations. Price to book value (P/BV) ratio โ€“ measures the value of assets of the company relative to the market value of the company. The formula is P/BV = Share price โ„ Book value per share. Price to cash flow (P/CF) ratio โ€“ measures the price of the volume cash flow per share at market value. The formula is P/CF = Share price โ„ Cash flow per share. Price to earnings (P/E) ratio โ€“ measures of the number of year it would take the company to earn net profits equal to the market value of the company. The formula is P/E = Share price โ„ EPS. 85

Price to earnings growth (PEG) ratio โ€“ measures the earnings growth in terms of P/E ratio. The formula is PEG = (P/E)โ„ EPS Growth. Price to research (P/R) ratio โ€“ measures the price of total expenditure of research and development per share. The formula is P/R = Market Capitalisation โ„ Turnover = Share price โ„ Research and Development (R&D) expenditure per share Price to sales (P/S) ratio โ€“ measures the price of turnover per share. The formula is P/S = Market Capitalisation / Turnover = Share price โ„ Turnover per share. Price to tangible book value (P/TBV) ratio โ€“ same as P/BV but only includes tangible assets. The formula is P/TBV = Share price / Tangible book value per share. Relative strength โ€“ measures the price trend of a stock which indicates its performing relative to other stocks in its industry. Return on Assets (ROA) โ€“ measure profit the company's assets in generating turnover. Return on Capital Employed (ROCE) โ€“ measures the profit the company makes for its capital. Sector โ€“ list of a business classification on the stock market. Each stock market listed company is allocated a relevant sector, for example, Barclays in โ€˜Banksโ€™, QinetiQ in โ€˜Aerospace and Defenceโ€™ and Sainsburys in โ€˜General Retailersโ€™. Turnover โ€“ the total of sales in one time period.

86