i
BACHELOR/MASTER Master THESIS WITHIN: Finance NUMBER OF CREDITS: 30
PROGRAMME OF STUDY: Civilekonom AUTHOR: Anton Pettersson
TUTOR: Kristoffer Månsson JÖNKÖPING 05-20
Investing made easy
Make investment decision for your
premium pension based on one single
factor.
i
Master thesis in Finance
Title: Investing made easyAuthor: Anton Pettersson Tutor: Kristoffer Månsson Date: 18-05-2020
Key terms: Funds, Finance, Premium pension, Pension, Management fee, Performance
Abstract
As a part of the Swedish pension system all participants in the working population are given the control of an investment account, which is funded with their pension contributions. The purpose of the account is to allow everyone the possibility to tailor their pension investments in accordance with personal values about risk and expected returns. However, it is shown that many Swedish adults lack the knowledge needed to make informed financial decisions and are therefore at risk of hurting their pension when investing in these accounts. This thesis tries to identify whether a simple but effective investment criterion, namely the size of the management fee, can increase the probability of higher risk-adjusted returns for uninformed investors when participating in the Swedish pension plan. It uses the Graham-Harvey volatility matched approach to evaluate fund performance on an individual basis. The corresponding analysis is then conducted based on grouped statistics of the projected performance measures. The results indicate stronger risk-adjusted returns of low fee funds in two out of the three analyzed categories. Carhart computations of Jensen´s alpha is then used as a robustness check and it provides support for the results found using the Grayham-Harvey method. When testing for difference in performance using the entire sample, 1.2 percent higher risk-adjusted returns were concluded statistically significant in the sample of low fee funds.
ii
Table of Cont
ents
Abstract ... i
1.
Introduction ... 1
The Swedish Pension System ... 1
Theory ... 3 Problem description ... 3 Purpose ... 4
2.
Theoretical Framework ... 6
Returns ... 6 Fund risk ... 6 Risk-adjusted return ... 7Risk factor models ... 7
3.
Previous research ... 12
4.
Data ... 14
5.
Method ... 17
6.
Results ... 19
Robustness check ... 227.
Discussion ... 24
8.
Conclusion ... 27
Limitations and suggestions for further studies ... 27
1
1. Introduction
In Sweden, all individuals in the working population are given funded investment accounts as a part of the national pension system, which compels everyone to participate in the fund market. Without any consideration of prior experience, individuals are forced to choose one to five funds to invest parts of their retirement savings in. This situation is problematic for some, as not all Swedish adults are capable of making rational investment decisions (Almberg & Säve-Söderbergh, J 2011). A large number of individuals are therefore at risk of investing in poorly performing funds and thus harming their pension.
For decades, researchers have tried to identify fund characteristics that are associated with mutual funds which are consistently earning positive risk-adjusted returns. Multiple characteristics have been proposed over time, but none have had the same impact as the size of funds´ management fees as proposed by Sharpe (1966). He showed that good performance is closely related to low management fees, a relationship that is not consistent with the economic theory of competitive pricing. Literature on fund markets is extensive and covers most aspects of fund investing. However, research on the premium pension scheme seems non-existent despite its 100 percent participation rate (Fondbolagens Forening, 2018). This thesis will evaluate the fee-performance relationship and aims to contribute to the existing literature by exclusively focusing on funds licensed to participate in the Swedish premium pension system. Further contributions to existing literature will be made by using a non-traditional method to evaluate mutual fund performance.
The Swedish Pension System
The Swedish pension system is rather unique, Italy, Poland, and Latvia are the only countries that have adopted similar structures as of today (Holzmann, Palmer & Robalino, 2012). The system is known as a notional defined contribution plan (NDC) and was fully implemented in Sweden 1998 (Sundén, 2006). The characteristics of the Swedish system is very similar to the most commonly employed pension plan around the world, the defined contribution (DC) plan. The key difference between the heavily adopted DC to the NDC plan is that for the latter, the Government uses current contributions to finance payments to current pension beneficiaries. Contributions to the NDC plan are recorded on notional accounts instead of traditional savings accounts. But both systems later repays the sum of lifetime contributions as an annuity after retirement (Holzmann, 2017). The Swedish pension system relies upon three major constructs
2
that are available for everyone (Pensionsmyndigheten, 2020a). The first and the largest part comes from the national pension, a wage-based component which differs from person to person. Secondly, the occupational pension, a non-mandatory contribution made by the employer. Lastly it is recommended to privately save for retirement through individual savings accounts. Every tax-paying worker in Sweden is given the right to a pension through the national retirement plan. The size of the national pension is dependent on the total amount paid to the system and the rate of capital appreciation on the account over the course of the individual´s working life. Contributions to the national pension are determined by a mandatory contribution rate of 18.5 percent of gross income. Payment to the system are directly removed from the salary on a monthly basis into two different plans; the first and the largest part is known as the income pension and it is debited with 16 of the 18.5 percentage points deducted from the employee´s salary. (Sundén, 2006). This is the part of the pension known as the notional contribution plan as mentioned earlier. Contributions to this system are documented into notional accounts, but the actual money is given to, and then redistributed by the government to finance current pension benefits. The notional account compares well to a fictional fund, the development of the fund is not determined by the performance of financial markets, but instead it is subject to appreciation by a notional rate of return determined by the government. When the individual then retires, he will be given the documented after-appreciation value of his accumulated lifetime contributions paid out monthly as an annuity. (Holzmann, 2017).
The additional 2.5 percentage of the monthly salary is credited to the premium pension plan, a system with funded individual investment accounts (Sundén, 2006). The value of these accounts at the time of retirement does not only depend on the individual’s aggregated contributions but also the performance of financial markets. The governmental authority known as the Premium pension authority (PPM) runs the administration of the accounts, however, every participant has the choice to personally choose one to five funds to invest their own savings in. Funds available in the premium pension must hold a license, which is issued by the PPM. The license is given to funds that follow rules and restrictions on the reporting requirement, manager compensations, price reduction, and sustainable investing (Pensionsmyndigheten, 2018). There are currently 495 licensed funds with all types of investment objectives, strategies, and specializations, giving the participants a high degree of freedom when tailoring portfolios according to personal preferences.
3
The other two pillars of pension include the occupational pension scheme, which is additional removals from the salary into pension saving accounts made by the employer. This part is non-mandatory but is more often than not settled by agreement between employer and employee or by collective agreement contracts (Pensionsmyndigheten, 2020b). With occupation and national pensions together, individuals will receive monthly pension benefits that correspond to 60-80 percent of his historical average salary after retirement. As the third pillar, authorities recommend individuals also have personal savings for retirement to boost this further (Pensionsmyndigheten, 2020c).
Theory
Since the topic was first introduced by Sharpe (1966), the role of fees in explaining returns have been heavily discussed. Contradicting theory and practice in the fee-return relationship is one of the reasons why this field has been given much attention, but the extent of which fees explain dispersion in performance is also of great importance. Pricing of mutual funds should reflect the relative ability to generate return to its investors, however, most research papers find that this is not the case. Instead studies show that funds charging high management fees, often underperform compared to its cheaper competitors (Tariq & Abdullahi, 2015; Harris, 1997). Some papers have identified a positive relationship, but a common trait in these studies is that they all have examined more niche populations. Funds from a sample with extremely high fees (Haslem, Baker, & Smith, 2007), or a sample of funds with a performance-based fee structure (Elton et al. 2003) are examples of these. Studies on the Swedish market show better risk-adjusted returns in low fee funds (Dalquist, Engström & Söderlind, 2000). The authors used regression analysis to show a significant difference in alphas of high and low fee funds. Their sample consisted of 126 equity and 42 money market funds based in Sweden and investing in the Swedish market over the period 1992-1997. To the knowledge of the author of this thesis, no research explicitly focusing on funds available in the premium pension exists, which is a bit surprising as it makes the entire Swedish working population participants in fund markets. Funds available in the premium pension are required to also operate outside the pension system. This means, the results from this paper are to some extent also applicable to other markets.
Problem description
Premium pension accounts for 13.5 percent of the monthly contributions to the national pension. But as the rate of return in the premium pension tends to be higher than the rate given on the notional accounts, the significance of premium pension would then rather account for
4
15-20 percent of the national pension at the time of retirement. The national income accounted for 69.4 percent of total pension in 2018 (Pensionsmyndigheten, 2018), meaning that the premium pension will on average, account for up to 15 percent of the total benefits after retirement. Also, one cannot forget that occupational pension is non-mandatory. There are therefore individuals whose pension completely relies on the benefits from the national pension, shredding no doubts on the significance of the returns in the premium pension.
With close to 500 funds available in the premium pension, it is not easy to select only five. Creating a fund portfolio that is well-performing and consistent with economic theory such as diversification and minimum variance optimization, may be possible for the most experienced investors, but for the majority this will not be the case. Almberg & Säve-Söderbergh (2011) show that many of Swedish adults lack the financial literacy required to make informed financial decisions. This puts a large number of individuals at risk of making financial mistakes and is therefore also at risk of hurting their pensions.
Since the introduction of the premium pension system, the share of participants who have made at least one change per year in his portfolio has fallen from 66 percent in 2000 to 5.7 percent in 2017 (Pensionsmyndigheterna, 2017). Lack of financial knowledge and fear of making economically damaging mistakes may be two possible reasons why active participation in premium pension has fallen since the start of the system. The idea of allowing individuals to tailor his pension savings investments according to their risk tolerance and targeted return may be seen as a good thing. However, it assumes that individuals can draw accurate conclusions about returns and risk in the available types of funds. But as illustrated by Almberg, & Säve-Söderbergh (2011), this is not always the case. Normally, the fear of making economically damaging mistakes tends to scare away financially uninformed individuals from fund markets, but in premium pension they are forced to participate anyway. Without any other guidance, these uninformed individuals make investment decisions based on other factors than the expectations of risk and return. As a result of this irrational decision making, it is easy to believe that the pension will vary widely between individuals at the time of retirement.
Purpose
Despite having a 100 percent participation rate from the Swedish working population, little or no attention has been given to funds available in the premium pension system. Due to mandatory participation and a high number of individuals who lacks the knowledge needed to
5
make informed investment decisions, there will be large deviations in the benefits coming from the premium pension scheme.
This paper focuses on funds available in the Swedish premium pension. It aims to give pension savers a simple investment criterion, which is not based on the expectations of returns and risk. It should be a criterion that everyone understands, no matter previous education, and it should be easily employed and require no calculations. Despite being simple, the criteria must also be associated with an increased probability of higher risk-adjusted returns, and thus contributing towards greater stability in future pension benefits.
Fees have been identified as a factor explaining deviations in return. It is easily identified and requires no further analysis and could, therefore, be the simple investment criteria uninformed investors want. By evaluating the relationship between funds´ management fees, historical return, and risk, this paper aims to answer the following research question:
Is it possible for participants in the premium pension scheme to select funds solely based on the management fee and still have a greater probability of superior performance?
For complete beginners in the financial markets it is all about keeping it simple. From this perspective, it is interesting to see whether funds with low fees do overperform to the more costly alternative, or if the high fee funds complying with the restrictions imposed by the PPM are priced on its ability to reward investors.
Similar studies have been made on most fund markets, but the premium pension is not one of them. This thesis, therefore, wants to contribute to the existing literature by solely focusing on funds licensed to participate in the Swedish premium pension scheme. A further contribution to existing literature will be made by employing a non-traditional method of measuring performance.
6
2. Theoretical Framework
The fund market is a widely researched area in finance. It discusses many aspects: problems, advantages, and disadvantages of fund saving. The fees associated with a fund and its ability to deliver return is one of the topics that have been discussed many times, especially in later years when digitalization has increased availability and with it, also the competition between asset management companies. Traditionally, informed investors have used linear regression models to estimate abnormal returns of funds relative to a given benchmark, but this paper will use a non-traditional method to evaluate fund performance. This section will present all the relevant theoretical frameworks to the methods used in this paper.
Returns
The core of this thesis is the fund returns. Computations of fund returns are projected using the funds per share accumulated asset value less than` liabilities. The net asset value or NAV of a fund represents the price of the fund at a given point in time. For the given sample of funds, all dividends are reinvested into the fund. Hence the return on investment is only dependent on changes in the NAV. Returns are computed on a monthly basis and it captures monthly increases or decreases in the value of his investment. Returns are projecting by the following formula:
𝑅, = , ,
,
(Eq. 1) Fund risk
Risk is related to the uncertainty that actual returns turn out to be different from the expected. Volatility is the most commonly used estimator of risk and it measures the deviations between the realized return compared to the historical average. With this measure both the upside potential and the downside risk are captured as both upside and downside deviations are included in the measure. Increasing the volatility increases the probability of both greater wins and bigger losses at the same time (Hundal et al. 2019). If the actual returns do not deviate from the historical average, the fund is considered less risky, the opposite holds true for funds with large variations in returns. Volatility is the square root of portfolio variance:
𝜎 = ∑(𝑅 − 𝑅 )
𝑁 − 1 Eq (2)
7
Risk-adjusted return
Investors are generally risk-averse, meaning that they not only care about high absolute returns, but they also care about how much risk is needed to achieve those returns. Ultimately the goal of investors is to maximize return but at the minimum amount of risk, but an inverse relationship between the two makes it difficult. Financial markets are characterized by a risk-return tradeoff (Sharpe, 1964), where investors must accept a higher probability of a loss to have the chance of higher returns. Because of this trade-off, it is not accurate to compare performance solely based on the ability to generate absolute returns as it does not say anything about the associated risk. Investors prefer certainty in their investments, hence a fund manager returning ten percent on investment at five percent risk has done a better job than a manager who generates the same return but instead at twenty percent risk. Therefore, risk-adjusted measures have been developed to more accurately measure and compare performance.
The most commonly employed risk-adjusted measure is the linear computation of Jensen’s alpha (Jensen, 1968). It measures the spread between the actual and the expected return, given the systematic risk of the asset. The expected return is specified using linear modeling, typically using models such as the CAPM (Sharpe, 1964) or some of its extensions which is discussed in the next section. The alpha is the intercept in these linear models and captures the fund’s expected risk-adjusted return, higher or lower than the expected net return of the benchmark. Jensen´s alpha is projected using the return of the evaluated fund (Ri), and the returns from the
corresponding benchmark (Rb,i) and additional risk factors (rn) discussed in the next section.
Alpha is computed using the following relationship.
𝛼 = 𝑅 − 𝑅 , + 𝑓(𝑟 ) (Eq. 3)
Risk factor models
The risk-return tradeoff has been widely studied throughout history. The first linear method to assess performance was developed in the ’60s by Treynor (1961, 1962), Sharpe (1964), Lintner (1965), and Mossin (1966) who independently built on the earlier work of Stuart & Markowitz (1959). Together they provided a simple but useful explanation of stock return as a function of systematic risk. Today the model is known as The Capital asset pricing model or CAPM. It is built on the assumption that investors are only rewarded on systematic risk as unsystematic risk can be eliminated with diversification. (Hundal et al. 2019). The model regresses the funds´
8
excess returns against the excess return of the market, where the estimated intercept shows the systematic risk-adjusted return above or below the market. The intercept is the alpha (αi) of the
fund to the market and the coefficient or beta (β), shows the sensitivity of fund returns to changes returns of the market minus the risk-free rate denoted (Rm-rf). CAPM is computed by
the following formula:
𝑅 − 𝑟 = 𝛼 + 𝛽 (𝑅 − 𝑟 ) (Eq. 4)
The model was later developed into incorporating multiple risk factors to better explain stock returns. Fama & French (1992) identified three common risk factors that had better explanatory power than the CAPM. The three risk factors include market risk (Rm – rf), overperformance of
small stock over large (SMB), and the overperformance of companies with high book-to-market ratio compared to companies with a low ratio (HML). The interpretation of the three-factor regression is the same as CAPM. Over or underperformance is identified by the alpha, and betas show the sensitivity of fund returns to the changes in returns of the corresponding risk factor.
R − r = α + β ,(R − r ) + β ,SMB + β ,HML (Eq. 5)
Later a momentum (MOM) factor was added to the model. Jegadeesh & Titman (1993) observed that buying the stocks that have performed the best over the last six months and short selling the stocks that have performed the worst realizes a significant abnormal return. Carhart (1997) added this abnormality into the Fama-French three-factor model as a measurement of fund manager performance. Carhart's four factors have become standard in mutual fund literature to measure risk-adjusted performance.
The full Carhart four-factor model is express as following:
R − r = α + β .(R − r ) + β ,SMB + β ,HML + β ,MOM (Eq. 6)
Points of critique have been given to traditional models due to the potential bias that may arise in linear models, (see example Grinblatt & Titman, 1989; Ferson & Schadt, 1996). In recent years, researches have tried to identify alternative ways to evaluate fund performance. In more recent years, it has become more common for researchers to avoid linear modeling in their research. However, many fields within the fund literature, including the literature on the
9
Swedish markets have not yet been researched, after the trend of using non-linear specifications in the analysis appeared.
A more recent risk measure of risk-adjusted performance is the Graham & Harvey (1997), measure (GHM). The measure is similar to Jensen’s alpha of the factor models, but it does not depend on a linear specification and is therefore free from bias that may arise from such a model. The measure is computed by constructing a portfolio using a benchmark index and the risk-free rate. The two assets are combined so that the portfolio has the same volatility as the evaluated fund. GHM is the difference between the fund return and the return of the levered or unleveraged benchmark. Figure 1 shows the return of two funds in relation to the benchmark portfolio. The line shows the risk and return for all combinations of benchmark and the risk-free rate, and the points illustrated by X shows the return of the combination of benchmark and risk-free rate that has the same volatility as funds A and B. In the figure, GHM(A) Shows a fund that has a lower return than the unleveraged benchmark. Fund B shows a positive GHM measure indicating superior performance over the benchmark portfolio given the same amount of risk. When leveraging, funds are borrowed at the risk-free rate to have a position in benchmark greater than 100 percent. When deleveraging, the position in the benchmark is smaller than 100 percent, the extra funds are lent at the risk-free rate. The net position in the benchmark portfolio is always 100 percent.
10
GHM is measured as the difference between the actual return of the fund and the return of a volatility matched benchmark. The matched portfolio is constructed by combining the benchmark index and the risk-free rate so that the average volatility equals the volatility of the evaluated fund. The risk-free rate is assumed to have zero volatility; hence the portfolio volatility (𝜎 ) is only dependent on the weighted volatility of the benchmark (𝜎 ), hence 𝜎 = 𝑤𝜎 . The weight, (𝑤), is the weight which sets the average volatility of the portfolio equal to the average volatility of the evaluated fund (𝜎 ), and it is determined by the relationship of the volatility of the fund to the benchmark: 𝑤 = . The return of the volatility matched portfolio is the sum of the weighted return of the benchmark index and the risk-free rate. The return of the benchmark portfolio is calculated using the following formula:
The GHM is then projected by subtracting the average return of the fund (𝑅 ) with the average return of the benchmark portfolio (𝑅 ).
𝑅 = 𝑤𝑅 + (1 − 𝑤)𝑟 (Eq. 7)
11
𝐺𝐻𝑀 = 𝑅 − 𝑅 (Eq. 8)
A positive GHM indicates stronger risk-adjusted returns compared to its benchmark and vice versa.
12
3. Previous research
There is consistency amongst researchers; funds with high fees should reward investors with a higher adjusted return. In equilibrium, all funds should have expected zero, after-fee, risk-adjusted return. The reasoning is that if investors can identify fund alpha, and it is greater than the fee (fi) there would be excessive demand for the fund, causing appreciation pressure on the
pricing until the fund edge disappears into the equilibrium condition: 𝛼 − 𝑓 = 0 (Berk & Green, 2004). If investors instead do not know alphas, the only factor determining price is the expectations of adjusted returns. Bil-gazo & Ruiz-Verdu (2004), proved that the equilibrium condition does not hold for the US fund market, and thus that the American funds are not competitively priced. In a report by The Swedish Competition Authority (Konkurenserket, 2015), it became evident that similar patterns exist in the pricing of Swedish funds. It concludes that despite high market concentration, companies tend to have high profit margins which should have depreciated away in the case of competitive pricing.
Sellers in a monopolistic market compete with differentiated products based on the quality of products and branding instead of price. The price of funds is reflected in the management fee and it means that companies can offer funds at a high fee despite a relatively poor performance compared to its competitors (Haslem, 2004). The pricing of mutual funds is a widely researched topic with varying conclusions. Haslem et al. (2007) investigated the mentioned relationship between management fee and fund performance in the US, and their findings suggest that funds with extremely high management fees may add value to its investors. Their sampling procedure divided the funds into groups based on the size of the management fee. The bands were determined based on how many standard deviations away the fee was from the average. Two performance measures were used in the analysis, the Sharpe ratio and Jensen´s alpha from CAPM were computed for each of the funds. Their empirical findings show that funds with management fees above three standard deviations from its mean have both higher alpha and sharp ratio than the funds with fees within one standard deviation from the mean, suggesting that the most expensive funds can add value to investors. On the contrary Gil-Bazo & Ruiz-Verdu (2009), found that in the general market there is instead a negative relationship between fee and return in the US. Using Carhart four-factor regressions, the authors showed that even before fees were removed, funds with high fees returned a smaller alpha compared to low fee funds. Considering the after-fee returns, it is obvious investors are even worse off investing in the high fee funds.
13
Despite the hypothesized positive relationship, most studies on the US but also international markets suggest that funds with higher fees on average return lower risk-adjusted returns to investors compared to the cheaper competitors. Fraś (2018) found clear evidence of a negative relationship between returns and management fees in the Polish funds based on a sample of 93 funds. The author used CAPM regression to compute abnormal returns before and after fees were deducted. According to their research, a one percentage point increase in the management fee was associated with a 0.6 percent decline in the before fee return on an annual basis. They found an even more statistically significant relationship between the fund variability and the fee, indicating that funds charging high fees typically have a more aggressive investment strategy, which does not reward its investors. The results seem to be consistent in most geographical markets. Tariq & Abdullahi (2015) used Fama-French three factors regressions to compute the abnormal return of Australian mutual funds. The authors show that Australian high fee funds generate weaker before fee alpha compared to the cheaper alternatives, both in poorer economic conditions as well as in normal economic conditions.
The tendency of low fee funds to generate higher returns also seems so exist in Swedish funds (Dalquist et al. 2000). They found evidence of this relationship using regression analysis based on a sample of 126 Sweden-based equity and 42 money market funds during the sample period 1992-1997. They divided the sample into a low and a high fee group separated on the 40th and
60th percentile respectively based on the size of the management fee. Estimations of Jensen´s
alpha showed a 1.4 percentage points higher in the low fee sample of equity and an alpha 0.24 percentage point higher in the money market funds.
It is clear that literature on the fee-performance relationship seems to rely on linear models in the performance analysis. Rodriguez (2010) used a different method to compare the performance of socially responsible with conventional funds. The method known as the Graham-Harvey measure is typically used in performance evaluation of market timing strategies, but despite being very similar to Jensen´s alpha in its nature, very little attention has been given to it in mutual fund performance analysis.
14
4. Data
The funds included in this study are sampled using the filtering tool on the independent investment advisor Morningstar Sweden´s website. The filtering procedure used the Morningstar Categories to identify equity funds investing in three of the most popular markets for Swedish investors; Global, Europe, and Sweden funds. Investments in the premium pension scheme are long-term, which is why this paper aims to maximize the sampling period while
maintaining enough funds in the sample to have statistical significance. Ten years turned out to be the longest possible period in order to maintain a relatively high number of funds. The sampled funds were all available in the premium pension and had a ten-year history of operations between January 2010 and December 2019. The filtering process returned a list of 61 funds from three different categories, with fees ranging from zero to 2.45 percent. The categories included 18 global funds, 19 European funds, and 24 Swedish funds. The denotations “Global” “European” and “Swedish” refer to the geographical equity markets funds invest in, not to be confused with the funds’ country of origin. The author is aware that the thesis can be exposed to survivorship bias as the sample only contains funds that are in business at the time of writing. Movements in or out of the premium pension are not taken into consideration, nor are changes in funds´ fees over time as no data of these two aspects could be found.
Monthly data of all fund NAVs as well as the index values of the geographical benchmarks were downloaded using the financial software Thomson Reuters DataStream. The assigned benchmarks were issued by Morgan Stanley Capital International (MSCI) which is the market leader when it comes to providing investors with geographical indices. MSCI is one of the market-leading index constructors with a business history of over 40 years. The benchmark indices in this thesis include MSCI: All-Country World Index (ACWI), Europe index, and Sweden index. The 1-month Swedish repo rate is used as the risk-free rate. It is the average between the lending and borrowing rate to the Swedish national bank, data for the repo rate was downloaded on the national bank´s official website. Data for the Carhart four-factors for the three markets was downloaded using Kenneth R. French data library.
To compare high with low fee funds, funds were divided into two groups based on the size of the fee. Similar to the study of Dalquist et al (2000), funds were ranked according to the management fee and then separated into two groups. The funds below the 40th percentile form
15
procedure was done for each of the three categories. Group size and fee statistics for each of the six groups are shown in Table 1.
Table 1. Group and fee statistics
Group Name Number of funds Average fee
(%) Minimum Fee (%) Maximum Fee (%) Global High 7 1.79 1.45 2.46 Global Low 7 0.38 0.08 0.53 European High 8 1.73 1.5 2.10 European Low 8 0.50 0.20 1.4 Swedish High 10 1.48 1.26 1.76 Swedish Low 10 0.33 0.00 0.62 Total sample 61 1.03 0.00 2.45
The data of the NAVs for each fund and index were then manipulated to project monthly returns using equation 1. Statistics of group average return and standard deviation are presented in Table 2.
Table 2 shows the average, median and standard deviation of monthly returns, as well as the monthly maximum and minimum for each group over the sample period. It shows that the average return of the groups varies from 0.67 percent in European high fee sample to 0.92 percent per month in low fee Global funds. Swedish funds have the highest median return with 1.46 and 1.45 percent for low and high fee groups respectively. Swedish equity funds also experienced the largest deviations in its returns where both high and low fee groups experienced volatility of 3.99 percent on a monthly basis. Global low fee funds were the least risky of the sample groups with monthly volatility of 3.48 percent. The highest maximum return was recorded in the global high fee group and the minimum return was recorded in the European high group.
Low fee samples have higher raw returns than high fee groups for Global and European funds. For the Swedish funds, the high fee sample manages to generate a higher average return during this period. Like for the returns, both low fee funds in the Global and European categories performed better than the high fee counterpart in the measurement of risk. The high fee funds experience higher volatility in its returns, additional risk that does not seem to generate an additional return to the investors. The risk-return relationship does not seem to exist in this
16
sample of funds, here investors instead are rewarded by long-term stability. The sample instead shows that the maximum and minimum values correlate with volatility. For this sample, high fee funds have equal or higher volatility than the low fee group. It is also the high fee group within each category that generally returns the highest maximum and lowest minimum of monthly returns. There is one exception though, the Swedish low fee sample has a lower minimum than the high fee counterpart.
Table 2. Descriptive statistics of monthly returns (%)
Group Name Mean Median Standard
deviation Maximum Minimum Global High 0.77 1.15 3.70 12.19 -11.75 Global Low 0.92 1.11 3.48 11.52 -11.43 European High 0.67 1.11 3.77 10.95 -15.60 European Low 0.69 1.01 3.59 9.48 -12.09 Swedish High 0.84 1.45 3.99 10.89 -12.99 Swedish Low 0.78 1.46 3.99 8.61 -14.12 MSCI ACWI 0.81 1.09 3.65 10.57 -9.59 MSCI Europe 0.58 1.03 3.51 8.29 -10.20 MSCI Sweden 0.53 0.67 3.97 8.58 -12.05
17
5. Method
This thesis uses a quantitative approach to analyze the relationship between management fees and performance. It will use the Graham-Harvey volatility match approach to measure fund performance relative to a matched benchmark portfolio. The measure was first introduced by Graham & Harvey (1997) when analyzing the performance of market timing newsletters. They used the method to analyze the performance of investing according to newsletter recommendations in comparison to passive index investing. Rodriguez (2010) then employed the measure in mutual fund literature when evaluating the performance of socially responsible funds. The Graham-Harvey measure was computed for two samples of funds, the analysis was then conducted on grouped statistics of the two samples. He then used the Welch´s t-test to determine whether a statistical difference in the performance of socially responsible and conventional funds existed.
This thesis will follow the methodology used by Rodriguez (2010). To answer the research question, GHM will be computed for the six samples of funds. The analysis will then be conducted based on the sample statistics of the measure for each group and market. Welch T-test for differences in mean will determine whether there is a statistically significant difference in risk-adjusted return between the high and low fee funds within each geographical market. The matched portfolio is constructed by combining the category benchmark and the one-month risk-free interest rate so that the volatility of the portfolio equals the volatility of the evaluated fund. Returns of the matched portfolios are calculated as the weighted return of the benchmark and the risk-free rate (Equation 7). The Graham-Harvey performance measure is then calculated by subtracting the average return of the benchmark portfolio from the average return of the fund. GHM is calculated using Equation 8.
Most academic papers on fund performance use computations of Jensen’s alpha or some of its extensions to compute abnormal returns. The Carhart (1997) four-factor model has become the standard in fund literature. This thesis wants to relax the linear assumption made in such models and thus uses the Graham-Harvey procedure as the main method, but will also compute Carhart alphas as a check for robustness in the results of the more unconventional method. The models are similar in their nature as they both compute the risk-adjusted abnormal returns, hence the results from both models are expected to be rather similar. The included factors include excess market return, size, book to market value, and momentum. An error term is also added to the model as it is now an estimation of a statistical relationship.
18
R − r = α + β ,(R − r ) + β ,SMB + β ,HML + β ,MOM + 𝜀 (Eq. 6)
Similar to the analysis of the main method, the alpha will be computed for each fund. The analysis will be conducted on the grouped statistics of the estimated alphas, and Welch t-test will then determine whether statistical differences exist between high and low fee samples.
19
6. Results
Fund performance can be evaluated in various ways, each providing a different perspective of performance. Figure 2 shows the long-run average return of each fund individually relative to the set of combinations of the assigned benchmark and the risk-free rate. A point lying above the lines indicates funds with superior portfolio management in relation to index investing. The figure shows how most the data points lie above the benchmark line, indicating strong management skills of most fund managers. Figure 2a shows the performance of European funds relative to the MSCI Europe index. 14 out of 16 funds lie above the benchmark, with one underperforming fund from each of the two groups. Figure 2b shows larger deviations in the performance of the sampled Global funds. 11 out of 14 funds have higher risk-adjusted return compared to the benchmark. Amongst the underperforming Global funds, 2 were “high fee” and one was “low”. Figure 2c shows the overperformance of the Swedish funds compared to its benchmark. Only one fund failed to beat the benchmark, and it came from the high fee group.
Figure 2. Annualized mean returns and standard deviation for sample of funds 2c. Sweden funds
20
As illustrated, most of the sampled funds generate higher risk-adjusted returns to investors compared to their benchmark. The Graham-Harvey measure quantifies the overperformance, by calculating the spread between data points and the benchmark portfolio of risk-free rate and index. Figure 3 shows the GHM of all funds individually plotted against its management fee. Figure 3a shows a weak negative relationship between GHM and the management fee of European funds. It also shows more stability in the performance of cheaper funds. The best and worst performing funds are both found in the high fee group. Figure 3b shows a strong negative relationship between performance and management fees. The low fee funds generally have positive GHM, while high fee funds are centered around zero but are also represented by the sample losers. The sample of Swedish funds was the only sample indicating a positive relationship between the fee and risk-adjusted performance. Figure 3c illustrates stability in the returns of low fee funds, while high fee funds, despite being all positive has greater variability in its returns. The figure also shows a strong outlier in the sample of high fee funds. Figure 3d shows the GHM and management fee of the 40th and 60th percentile of the entire sample of 61
funds. The figure shows a negative relationship between the GHM and the management fee. The majority of funds with a negative GHM are from the above 60th percentile, the sample
losers also come from the more costly group. It is also shown that variations in returns from the low fee percentile are lower than in the more costly counterpart.
21
Table 3 presents the group results of GHM estimations for the different samples of funds. It shows, the average and standard deviation of GHM scores, maximum and minimum values as well as the share of positive values. With over 66 percent of high fee funds and almost 90 percent of low fee funds returning a positive GHM, the results indicate strong overperformance of funds compared to their benchmarks. Swedish funds showed the highest risk-adjusted return above the benchmark with over three percent highly statistically significant abnormal return for the high fee group and 2.8 percent in the low fee sample. Global and European funds show similar characteristics of GHM, a negative value for high fee group, and a positive value for the low. For the two categories, low fee funds show a higher share of positive values, indicating low fee funds have a higher probability of overperformance to its benchmark.
When testing for differences between sample means within each category, 2.3 percent statistically significant higher GHM was found in global low fee funds over global high fee funds. Similarly, European low fee funds had 0.9 percent higher risk-adjusted return compared to the high fee group, although significance was failed to be concluded for this category. For the Swedish sample, high fee group had 0.4 percent higher GHM than the low fee sample, but again without statistical significance. When then testing for the entire sample by including all the three categories, 1.2 percent higher return was observed in low fee funds, overperformance was then concluded at ten percent significance.
3c. Sweden funds 3d. Total sample
22 Table 3. Group statistics of annualized GHM
Global European Swedish Aggregated
High Low High Low High Low High Low
Mean -0.006 0.018** -0.003 0.005 0.032*** 0.028*** 0.007*** 0.019 Standard Deviation 0.021 0.018 0.027 0.014 0.018 0.007 0.028 0.014 Max 0.025 0.040 0.028 0.014 0.074 0.037 0.074 0.040 Min -0.033 -0.017 -0.063 -0.031 0.006 0.018 -0.064 -0.016 % positive values 57.1 85.7 50.0 62.5 100 100 66.7 87.5 Difference in means -0.023* -0.009 0.004 -0.012*
Notes: This table presents the distribution of the results for the GHM estimations. *,** and *** denote statistical significance at ten, five and one percent respectively.
Robustness check
To have a clearer picture of the risk-adjusted performance of the two groups an additional test is performed. Alphas for all funds are estimated individually using the Carhart (1997) four-factor model. Table 4 presents the grouped average results of regressions. All groups of funds have a high, statistically significant alpha, hence the evidence of higher risk-adjusted returns of funds over the respective benchmark continues. Like in the GHM estimations it is shown that low fee global and European funds beat the high fee alternative, but the identified difference is only significant in the global category. The opposite direction of the relationship is once again observed for the Swedish category, where a highly statistically significant difference in alpha of 0.7 percent where identified in favor of high fee funds. When looking at the entire sample, a small overperformance of 0.3 percent annually is found in the low fee group at ten percent significance.
23 Table 4. Group statistics of annualized alpha
Global European Swedish Aggregated
High Low High Low High Low High Low
Mean 0.038*** 0.065*** 0.048*** 0.049*** 0.042*** 0.035*** 0.045*** 0.048*** Standard Deviation 0.004 0.013 0.005 0.002 0.007 0.002 0.008 0.004 Max 0.007 0.011 0.006 0.005 0.007 0.004 0.011 0.006 Min 0.003 -0.002 0.002 0.002 0.001 0.002 -0.002 0.002 % positive values 100 85.7 100 100 100 100 0.958 100 Difference in means -0.027** -0.0002 0.007*** -0.003*
Notes: This table presents the distribution of the results for the Carhart alpha estimations. *, ** and *** denotes statistical significance at ten, five and one percent respectively.
24
7. Discussion
The results show a robust negative relationship between fee and risk-adjusted return for two out of the three categories of funds. From the GHM estimations, Global funds, as well as the test for the entire sample indicates significant differences in the average risk-adjusted returns in favor of low fee funds over high. The low fee sample performs better in Europe funds, but the difference does not have statistical support. The results from Carhart's regression gives robustness to all the identified relationships in the first method. Estimations of Jensen´s alpha show similar directions of overperformance for all categories, also the magnitudes of differences in risk-adjusted returns are similar for both methods. The significance of group differences using the sample averages from Carhart regressions was higher than in the Graham-Harvey methods. This could be explained by the linear specifications in the model reducing the volatility in the projected abnormal returns. The findings in this paper are consistent with the results from Fraś (2018), Gil-Bazo & Ruiz-Verdu (2009), Tariq & Abdullahi, (2015), Harris, J. (1997) and Dalquist et al (2000).
A puzzling positive relationship was identified in the funds investing in Swedish Equity. Contradicting to previous studies and the hypothesized results, the high fee group of Swedish funds robustly returned higher average risk-adjusted returns compared to the low fee sample. One possible explanation for the identified relationship is the sensitiveness to outliers due to the relatively small sample size. When looking into the estimations of individual funds an extreme outlier in the high fee group with a GHM more than twice the size of the mean is identified. A test of the sensitiveness of the results to this outlier support this theory. Reducing the GHM of that fund to the average of the other funds within the same category changes the direction of overperformance to the low fee sample.
Another potential explanation could come with expertise from investing in the domestic market. The management fee is often tied to the degree of active management in the funds (Haslem et al. 2007). Cheaper funds tend to more closely replicate the benchmark index, while active funds tend to rely more on stock picking or market timing. Expertise in local businesses could potentially increase managers´ stock-picking abilities and thus increase the probability of beating the cheaper more index alike funds in the domestic market.
The sampling was done by creating two groups using the 40 percent cheapest and 40 percent most expensive funds for each category. The idea of using this procedure is to best capture the
25
hypothesized relationship by excluding funds with fees in the middle range. However, as the category samples are rather small, this procedure has its flaws. For the Europe category, the low fee group consisted of funds with fees up to 1.4 percent. For the other categories this rate would be considered in the middle or even in the high range, meaning it is not 100 percent representative for the low fee group and thus possibly inferring the results. When looking into the size of the management fees, the majority of the cheaper funds have fees clustering around 0.20 – 0.55 percent, meaning that including funds with fees up to 1.4 percent may skew the data. Similarly, the fees from the “high” group cluster around 1.30-1.70 percent, hence it could be more representative to separate the two groups based on those rates. However, it was not possible to do so in this paper as the within category samples would become unjustifiably small. Previous papers have provided several hypotheses about why this negative fee-performance relationship can be observed despite the contradicting theory. Dispersion of fees should in principle reflect the dispersion in the performance expectations but conducted research does not support this view. Christoffersen & Musto (1996) provided evidence that demand curve variations can explain fee variations that some cost considerations cannot. They argue that there are two types of investors, performance sensitive and performance insensitive. Sensitive investors will move their money to better prospects in case of poor performance while the insensitive investors will not, which will result in a high density of insensitive investors in the poorly performing funds. In the end, the fee is only a reduction in the performance and the investors that are still investing in the funds with poor performance are known to be insensitive to changes in returns. Companies can, therefore, raise the management fee without risking substantial money outflows. They are incentivized to do so if they want to maintain a similar level of revenues as before the sensitive investors withdrawing their money. It shows how funds with poor performance become “high fee” to save their revenues, but it also explains why funds with high fees have poorer historical performance. Also, the momentum effect could have a role in this. It is shown that winners during one period are likely to remain amongst the winners also the next, similarly, it is shown that the worst performing funds during a period are also likely to remain amongst the losers also in the next (Jegadeesh & Titman, 1993). If funds charge higher fees as they have become one of the losers, they are likely to remain amongst the losers also after raising the fee.
Another possible explanation to the negative fee-performance relationship is that funds with poorer performance require more marketing to attract investors, a cost that could directly be
26
borne by investors through an increased management fee (Haslem et al. 2007). Companies often use graphs of the relative performance of their funds to the benchmark as a marketing tool to show the funds that are on a “hot streak”. The results of this thesis showed that a large proportion of funds beat the benchmarks, which puts the asset management companies in a position where they easily could take advantage of such a marketing strategy. The same authors also motivate that fund managers do not earn most of its profits from fund performance, but rather through sales and asset growth through increased deposits from investors. Managers who prioritize is own salary then could shift their focus on marketing practices instead of giving full effort to maximizing investor value. Other papers suggest that actively managed (and thus often more costly) funds underperform to passive index funds due to the inability of timing the market but also as a result from lower returns due to non-stock holdings (Graham & Harvey, 1997). This thesis hypothesizes that just like the inelastic demand from performance insensitive investors, financial illiteracy can also motive acceptance of higher pricing. Outside financial markets there is a common misconception of a price-quality relationship amongst consumers (Smith & Natesanm, 1999). It is known that financially uninformed participants make investment decisions based on other factors than the expectations of risk and return. It is therefore believed that this misconception could also exist in the financial markets and thus explain inflow to the more costly funds. Financial illiteracy could potentially explain why some investors select costly funds with the wrong expectation of higher returns. This hypothesis also coincides with the economic theory of funds in a competitive market being priced by its ability to generate value to its investors (Gil-Bazo. & Ruiz-Verdu, 2004).
27
8. Conclusion
This thesis examines if a relationship exists between the management fee and risk-adjusted return in funds operating in the Swedish premium pension system. The aim is to identify if it is possible to have an increased probability of higher risk-adjusted returns, by investing in a portfolio wholly determined by looking on the size of the funds´ management fee. It examines the performance of a sample of 61 mutual funds split into three different categories based on the geographical equity markets the funds are investing in. To measure risk-adjusted performance, a volatility matching methodology is used to compare fund performance relative to its geographical benchmark. This paper concludes a robust relationship between low fees and stronger returns in two of the three categories of funds. When testing for the difference using the full sample of 61 funds, a difference of 1.2 percent was found using the Graham-Harvey measure and 0.3 percent from the Carhart regressions. Both measures favor the low fee sample with statistical significance stronger than ten percent. For the funds investing in Swedish equity, robust higher risk-adjusted returns are found in the sample of high fee funds. This relationship also has strong statistical support from the Carhart regressions.
Despite a theoretical positive relationship between price and performance, most research papers show clear evidence of underperformance of more expensive funds. Using two different methods, this thesis finds robustness and consistency of a ten percent significant relationship of overperformance of low fee funds, findings that also have strong support from previous literature. Due to robustness from two models as well as the strong statistical relationship found using Carhart computations of alpha, it is believed that cheaper funds do on average increase the probability of higher returns also in the Swedish premium pension system.
This thesis wanted to provide financially uninformed participants in the premium pension scheme, a simple but yet effective criteria for choosing funds to their premium pension. The results show that for the given sample it is better to invest in funds with low management fees as they on average generate 1.2 percent higher risk-adjusted return to its investors. If investors want to explicitly focus on one of the categories of funds, the strongest and most statistically significant overperformance of low fee funds was found in the global category.
Limitations and suggestions for further studies
Some limitations were encountered when writing this thesis. The author is aware that the thesis can be exposed to survivorship bias, as the sample only contains funds that are in business at the time of writing. Funds of any group that may have been forced out of business most likely
28
would have significant underperformance and therefore would have depreciated the results of any of the group, possibly changing the analysis. Neither is movements in or out of the premium pension scheme taken into consideration nor are changes in funds´ fees over time as no data of this could be found.
This thesis concluded a robust but weakly significant relationship between the fee and risk-adjusted return. Funds available in the premium pension with over ten years in business were scarce. It is therefore believed that the amount of gathered data was not enough to conclude a statistical relationship in all three categories. To correct for the weak results, it is suggested to increase the number of funds within each category. With sufficiently many more funds available, the procedure of sampling based on the 40/60 percentiles most likely would yield stronger results than what is found in this thesis. However, a good idea might also be to consider sampling based on the fee ranges identified in section 8.
Suggestion for further research is to analyze whether changes in the fee are associated with changes in abnormal performance before or after the change in fee. This would then answer if poor performance amongst funds happens before or after fees are raised. It will also answer whether the fee variations in the Swedish markets are explained by demand curve variations as proposed by Christoffersen & Musto (1996). It could also be of interest to analyze if funds available in the premium pension experience higher demand curve variations than funds outside the pension scheme. It would provide insights on whether the increased number of uninformed investors in the premium pension results in increasing insensitiveness to performance and thus explain the survivorship of poorly performing high fee funds.
29
References
Berk, J. & Green, R. (2004). ‘Mutual fund flows and performance in rational markets’ Journal of Political Economy, vol. 112, no. 1, pp. 1269-1295.
Christoffersen. S & Musto, D. (2002) ‘Demand curves and the pricing of money management’ Review of Financial Studies, vol. 15, pp.1495-1524.
Dalquist, M., Engström, S. & Söderlind. P. (2000). ‘Performance and Characteristics of Swedish mutual funds’ Journal of financial and quantitative analysis, vol. 35, no. 3, pp. 409-432.
Elton, J., Gruber, M. & Blake, C. (2003). ‘Incentive fees and mutual funds’ Journal of Finance, vol. 58, no. 1, pp.779-804.
Fama, E. & French, K. (1992). ‘The cross-section of Expected Stock Returns’ The Journal of finance, vol. 47, no. 2, pp. 427-465.
Fama, E. & French, K. (2015). ‘A five factor asset pricing model’ Journal of financial economics, vol. 116, no, pp. 1-22.
Fonbolagens Förening, (2018). Det Svenska fondsparandet ur ett internationellt perspektiv – en översikt. [online] Available at: https://www.fondbolagen.se/fakta_index/studier-och-undersokningar/ [Accessed 30-04-20].
Fraś, A. (2018). ‘The relation between management fees and the mutual funds` performance in Poland in 2015’ Oeconomia Copernicana, vol. 9, no. 2, pp. 245–259.
Gil-Bazo, J. & Ruiz-Verdu, P. (2009). ‘The relationship between price and performance in the mutual fund industry’ The Journal of Finance vol. 64, no. 5, pp. 2153-2183.
Graham, J. & Harvey, C. (1997). ‘Grading the performance of market timing newsletters’ Financial Analysts Journal, Vol. 53, pp. 54-66.
Haslem, J. (2004). ‘Are mutual fund expenses too high? A commentary’, Journal of Investing, vol. 13, no. 2, pp. 8-12.
Haslem, J., Baker, H.K & Smith, D.M. (2007). ‘Identification and performance of equity mutual funds with high management fees and expense ratios’ Journal of Investing, vol. 16, no. 2, pp. 32-51.
30
Holzmann, R., Palmer, E. & Robalino, D. (2012) ‘Nonfinancial Defined Contribution Pension Schemes in a Changing Pension World’ The World Bank, pp 31-85.
Holzmann, R. (2019). ‘The ABCs of Nonfinacial defined Contribution (NDC) Schemes’, International social security review, vol. 70, no. 3, pp. 53-77.
Hundal, S., Eskola, A. & Tuan. D. (2019). ‘Risk return relationship in the Finnish stock market in the light of Capital Asset Pricing Model (CAPM) ’ Journal of Transactional Management, vol. 24, no. 4, pp. 305-322.
Jegadeesh, N. & Titman, S. (1993). ‘Returns to buying winners and selling losers:
Implications for stock market efficiency’ The Journal of Finance, vol. 48, no. 1, pp. 65–91. Jensen, M. (1967) ‘The Performance of Mutual Funds in the Period 1945-1964’ Journal of Finance, vol. 23, no. 2, pp. 389-416.
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, vol. 47, No. 1, pp. 13-37.
Mossin, J. (1966). ‘Equilibrium in a Capital Asset Market’ Econometrica, vol. 34, no. 4, pp. 768-783.
Pensionsmyndigheten. (2017). Pensionssparana och pensionärerna. [online] Availble at:
https://www.pensionsmyndigheten.se/statistik-och-rapporter/statistik/statistik-for-premiepension [Accessed 02-03-2020]
Pensionmyndigheten (2018). Fondavtal mellan pensionmyndigheten och fondförvaltare. [online] Available: https://www.pensionsmyndigheten.se/om-pensionsmyndigheten/for-fondbolag/ansokan-om-att-inga-fondavtal [Accessed 11-03-2020].
Pensionsmyndigheten. (2020a). Pensionens alla delar. [online] Available at:
https://www.pensionsmyndigheten.se/forsta-din-pension/sa-fungerar-pensionen/pensionens-alla-delar [Accessed 03-03-2020].
Pensionsmyndigheten (2020b). Det här är tjänstepension. [online] Available at:
31
Pensionsmyndigheten. (2020c). Eget sparande till din pension. [online] Available at:
https://www.pensionsmyndigheten.se/forsta-din-pension/sa-fungerar-pensionen/eget-sparande-till-pension [Accessed 02-03-2020]
Pensionsmyndigheten. (2020d). Total Pension. [online] Available at Pensionsmyndigheten:
https://www.pensionsmyndigheten.se/statistik/pensionsstatistik/?domain=tab-
6&report=report-6-2&columns=Apbelopp&rows=Time&sex=Samtliga&metrics=Medelbelopp&Alder=Samtliga
×eries=2005-01-01×eries=2018-01-01&complete=false&childrows-store=&reportGraphsType=line [Accessed 02-03-2020]
Rodriguez, J. (2010). ‘The performance of socially responsible mutual funds: a volatility match approach’ Review of Accounting and finance, vol. 9, no. 2, pp. 180-188.
Sharpe, W. (1964). ‘Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk’ The Journal of Finance, vol. 19, no. 3, pp. 425-442.
Sharpe, W. (1966). ‘Mutual fund performance’ Journal of Business, vol. 39, pp. 119-138. Smith, K & Natesan, C. (1999). ‘Consumer price-quality beliefs: Schema variables predicting individual differences’, Advances in consumer research, vol. 26, no. 1 pp. 562-568.
Stuart, A. & Markowitz, H. (1959). Portfolio Selection: Efficient Diversification of Investments, John Wiley & Sons, Inc. New York.
Sundén, A. (2006). ‘The Swedish experience with Pension reform’, Oxford Review of Economic Policy, vol. 22, no. 1, pp. 133-148.
Tariq, H. & Abdullahi, A. (2015). ‘The relationship between Australian mutual fund fees and risk-adjusted performance in different economic conditions’ Australian economic papers, vol. 54, no. 1, pp.1-22.
Treynor, J. (1961). Market Value, Time, and Risk [online] Available at:
https://ssrn.com/abstract=2600356 [Accessed: 12-05-2020].
Treynor, J. (1962) Jack Treynor's 'Toward a Theory of Market Value of Risky Assets. [online] Available at: https://ssrn.com/abstract=628187 [Accessed: 12-05-2020]
