Performance of Socially Responsible Investing
Evidence from Dutch Global Equity Funds
Christoffer Idström
1177332BSc Business Administration Specialization: Finance
Supervisor: Magdalena Jurgiel Date: 15/07/2020
Abstract
This paper is aimed at investigating whether or not Dutch SRI funds that invest in global equities outperform their conventional counterparts. Two samples of conventional mutual funds and SRI mutual funds were compared. The SRI sample contained nineteen SRI funds, and the sample of conventional funds contained thirty funds. The returns of each fund were individually calculated both on a pre- and post-fee basis. Two asset pricing models were used to determine performance the Five Factor Asset pricing model of Fama and French (2015) and the traditional CAPM. The before- and after-fee alphas yielded for each of the fund types from each model were compared and the statistical significance was tested using a Paternoster Z-test (1998).
Additionally, Sharpe ratios of the funds were calculated individually and served as a basis for analysis. In order to determine the statistical significance of the difference between the Sharpe ratios of the funds and the market, the Jobson-Korkie Z-test was utilized. It was determined that both fund types mostly provide investors with worse risk-reward relationships compared to that of the market index as proxied by Fama French Developed Markets index (French, 2020).
The funds in the sample both SRI and conventional individually showed a trend of underperformance based on each of the regression models. When comparing the SRI funds’ alphas from panel regressions of the two models to the alphas of the conventional funds, no statistically significant difference between the two types’ risk reward relationship was found. The results yielded from the regressions cannot be considered to be conclusive however, as there were issues found within the data, which limit the conclusions drawn in this paper.
When comparing the Sharpe ratios of the two fund types with each other, it was yet again found that there is no statistically significant difference between the two types’ risk reward relationships, adding to the previous literature that there is a lack of difference in performance between SRI mutual funds and their conventional counterparts.
Statement of Originality
This document is written by Student Christoffer Idström 11773332, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
Abstract ... 1
Statement of Originality ... 2
Table of Contents ... 3
Introduction ... 4
Literature Review ... 6
Data...10
Methodology ...13
Hypotheses ...20
Results ...23
Discussion ...38
Conclusion ...43
Appendix ...45
References ...63
´Introduction
The world today has changed significantly in various ways in the recent decades since the beginning of a new millennia. The world is more connected than before, more economies have opened up and financial markets are now accessible to a vastly broader extent than ever in history. People travel more, cultures are interconnected and the political landscape is positively turning towards open democracies all over the world, where less people are impoverished and live in hardship (Acemoglu & Robinson, 2012). However, despite these positive changes however, there has also been negative developments. The global temperature has been increasing exponentially, multinational corporations have taken advantage of poor worker protection laws in developing countries and pollution levels along with CO2 emissions are increasing (Acemoglu & Robinson, 2012). Thankfully governments, individuals and institutions are realizing the devastating impact of these harmful practices, and are putting their money where their mouths are with socially responsible investing.
Even though Socially Responsible Investing (SRI) is a concept dating further back than most people realize, with roots in religious texts older than the middle ages, it has in recent decades experienced exponential growth since its modern-day conception in the 1970’s (Schröder, 2006). Today SRI is an important financial strategy for many large investors and individuals alike (USSIF, 2018). As people in general are getting more conscious towards social causes in their ways of living so are their investment decisions (Schröder, 2006). This has given important momentum for the SRI funds, and what previously considered a niche investment strategy is now a common instrument for investing (Leite & Cortez, 2014).
Today the SRI market is a multi-trillion-dollar industry, and more than 30 percent of total assets under management being SRI assets, growing exponentially since the 1990’s when SRI investing started growing as a trend (USSIF, 2018). SRI has a somewhat ambiguous definition, but in this thesis the definition used by the Forum for Sustainable and Responsible Investment will be used. They define SRI as follows: “an investment process that considers the social and environmental consequences of investments, both positive and negative, within the context of rigorous financial analysis.” (USSIF, 2020). Furthermore, it is important to highlight a few key differences between conventional and socially responsible investment mutual funds.
One key implication in this definition is the positive and negative consequences of investments, and based on these SRI funds apply screens. These screens are either positive or negative. A negative screen indicates that the firm under screening has a negative social impact, and is excluded from the scope of investment. A positive result of a screen indicates that the investment
has a net positive social impact and therefore gets a higher “investment” grade. (Auer & Schuhmacher, 2016).
As these screens are more time consuming compared to standard financial analyses, and the impact of this has been a major interest for researchers. In the existing literature it’s very commonly discussed and researched how the ethical screens are affecting the performance of the SRI funds, due to higher fees and less diversification possibilities. A large part of the existing literature has focused on comparisons between conventional and SRI funds. Despite there being multiple and extensive studies, the findings of under- or overperformance is mixed, with little consistent findings in either direction. (Bauer et. al, 2005).
The inconclusive evidence of either positive or negative performance of SRI mutual funds compared to conventional mutual funds leaves a lot to be desired for research to conclude. An important aspect has been the fact that many of the studies previously conducted are focusing on larger geographical areas and markets. This however, does not shed much specific knowledge about specific regions. That gives importance to examining and analysing individual smaller markets and funds with a specific limited investment universe, which this study is aimed at. There have been two previous studies conducted specifically on the performance of Dutch SRI funds, and a variety of studies covering the Netherlands as part of a larger sample, but yet again no evidence is found regarding consistent under- or overperformance (Scholtens, 2005), (Scholtens, 2007).
Due to the lack of consistent findings on this topic and the increase in importance of socially responsible investing, this study aims at further investigating the performance of SRI funds in a narrower context. This paper aims at answering the research question:
“Do the Dutch SRI mutual funds that invest into global equities outperform conventional Dutch Mutual funds that invest into global equities?”
Two regression models will be used in this paper to determine performance, the traditional CAPM and the more recent Fama French Five factor model (Fama & French, 2015). Furthermore, the Sharpe ratios (1964) of the individual funds will be calculated and compared.
This study finds no outperformance of the SRI mutual funds compared to the conventional mutual funds based on the two regression models, neither before nor after fees. Furthermore, it was concluded that both fund types mostly provided negative alphas based on the two regression models, both before and after management fees are included. The Sharpe ratios are consistently lower than those of the market index, the Fama French Developed Markets index. No statistically significant difference between the Sharpe ratios of the fund types was found.
This study first describes in further detail what previous studies found on the topic of SRI investing. First the common characteristics of SRI funds are presented, then different theories of fund performance are detailed. Then the previous literature and findings regarding performance of SRI funds is presented, and the existing evidence from the Dutch studies are presented in more detail. After the existing literature is discussed, the data used in this study is described, the process if data collection is presented and the limitations and potential issues with the data are discussed. After the presentation of data, the methodology for analysis is described in detail. After the data and methodology section, the hypotheses are formed, which will function as a means of answering the research question. Thereafter the results of the analyses are presented and discussed together with the key findings of the study. Finally, the study is concluded, providing a brief overview of the whole paper and pointing out the most important takeaways.
Literature Review
Characteristics of SRI Funds
The most important factor that differs SRI funds from conventional ones are the applications of screens. SRI funds either use positive or negative screens when selecting their investments (Bollen, 2007). Negative screens are screens that will exclude a certain stock from the investment selection, due to having negative ethical impacts, such as firms operating in the tobacco or pornography industries. SRI funds tend to be more exposed to growth-oriented stocks compared to value-stocks (Bauer et. al, 2005). SRI funds also tend to be more prone to invest in small-cap stocks (Schröder, 2006).
Some studies have also found that the investors in SRI funds tend to differ from investors in conventional funds, where lower performance is accepted. This is due to the willingness to give up personal gain for an ethical cause, supported by the argument that SRI investors are more slowly withdrawing their capital compared to conventional investors (Geczy, Stambaugh & Levin, 2005).
One important factor affecting performance of SRI funds, and which is consistently different from conventional funds are the higher management fees (Leite & Cortez, 2014). As the cost of portfolio construction is a key driver of fund performance, this is an important characteristic for analysis. (Revelli & Viviani, 2014). A large body of research conducted on the characteristics of SRI
funds find that the fees are significantly higher compared to conventional mutual funds (Geczy et. al, 2005), (Capelle-Blancard & Monjon, 2014).
Theory of Fund Performance
Modern Portfolio Theory
Modern portfolio theory discusses the advantage of diversification when selecting the investment universe. Greater diversification of investments, will also provide the investor with lower variance for the same expected returns, as long as the investments are not perfectly correlated (Markovitz, 1952). When the investment universe is restricted, so are the diversification benefits and hence the portfolio will deliver lower returns (Markovitz, 1952). This would suggest that limitations to the investment universe provides a lower minimum variance frontier, and hence a worse risk-return relationship (Gil-Bazo, Ruiz-Verdú & Santo, 2009). This implies that limiting the investment universe to only domestic markets will have a negative impact on the performance of the fund, as the same degree of diversification is not reached. Funds investing globally should hence achieve greater diversification benefits compared to those only investing in a single market (Bodie, Kane & Marcus, 2018). However, this also has implications for funds applying criteria that limit the available investments. By limiting one’s investment universe the same degree of diversification is no longer possible, and hence the risk-return relationship is worsened (Hickman, Teets & Kohls, 1999). As SRI funds apply screens when selecting investments, they exclude a certain portion of the available securities on the market. This means that SRI investors should have to bear a so-called diversification cost. (Revelli & Viviani, 2014). This is supported in some research where it was found that returns for investors applying screens to their investment selection had lower annual returns compared to those not applying restrictions. (Gesczy, Stambaugh, & Levin, 2006)
Good Management Theory
One argument predicting SRI funds outperforming conventional funds is based on the performance of the underlying assets of the fund, where it is theorized that companies engaging in Corporate Social Responsibility (CSR) will outperform companies who do not. (Bollen, 2007) This is discussed in some previous studies and expressed as the “good management theory” (Waddock & Graves, 1997). The theory states that management engaging in CSR are outperforming the managers who do not engage in CSR, which could explain why socially responsible firms in general performing
better than the firms who chose not to act ethically (Revelli & Viviani, 2014). The findings regarding this theory are mixed and there is little persistent support for this claim (Leite & Cortez, 2014).
Costs of Portfolio Creation
Previous literature argues that that the costs of the portfolio creation will lead to underperformance of SRI funds. As SRI mutual funds require more rigorous screening processes compared to conventional funds, the management costs have been found to increase and lead to underperformance of the funds (Revelli & Viviani, 2014). It has been found that financial performance is reduced relative to the number of screens applied (Capelle-Blancard & Monjon, 2014). As SRI funds apply screening processes that are more rigorous and limit their investment universes compared to conventional funds, and the costs might outweigh the benefits from a financial perspective (Geczy et. al, 2005)
To conclude, theory suggests that in efficient markets, fundamental analysis will not provide the investor with positive alpha returns. Hence SRI funds should not outperform the market nor their conventional counterparts. If markets are inefficient there is opportunity to find under- or over-priced investment opportunities. The theory also suggests that diversification provides investors with an improved risk-return relationship, which indicates that international funds, both conventional and SRI will outperform funds with an investment universe limited to domestic markets. The diversification theory implies that screens which exclude certain opportunities will provide less of a diversification benefit, and provide worse risk return relationships. As SRI funds apply screens limiting their investment universe they should underperform compared to the conventional funds. As more intense screening processes are applied to actively managed funds compared to passive funds, any alphas that might arise from these screens are diminished by increased fees. This implies that SRI portfolios should underperform counterparts if the value generated by the more rigorous screening process do not exceed the costs of implementing these screens
Empirical Evidence of SRI Fund Performance
International SRI performance
There is a significant amount of research conducted regarding the topic of ethical investing, largely due to the fast-increasing development and popularity of this type of investing. The results however are not overwhelmingly conclusive to either under- or outperformance of SRI funds. This section is aimed at providing insights regarding the findings of existing literature.
The largest part of research conducted, has concluded that there is little difference between SRI funds and their conventional counterparts. Hamilton et. Al (1993) found that US based SRI funds perform slightly better than their conventional counterparts, but not on a statistically significant level. The study also found that both conventional and SRI funds to underperform their relative indexes. Another, and more recent US based study also found a slightly higher alphas among the SRI funds compared to the conventional counterparts, but yet again not statistically significant and both types of funds underperforming their indexes (Statman, 2000). More recent studies have also yielded similar results of no significance difference in performance between SRI mutual funds and conventional mutual funds. There is one other US based study that found that SRI mutual funds offering equal risk-return relationships to those of the conventional funds (Kiymaz, 2019).
Studies conducted based on European mutual funds have found a wider variety of performance among the SRI funds. A study conducted in 2017 finds that the green funds significantly underperformed their conventional counterparts. The green funds however, improve their performance significantly over time (Ibikunle & Steffen, 2017). These findings are in line with a previous paper written over a decade earlier. The study found a significant underperformance, and implications that investors are paying a premium for ethics in Europe (Renneboog et. al, 2005). Auer and Schuhmacher found that European SRI funds offer worse returns than the conventional mutual funds in the same area implying a premium for ethics. The study also finds that SRI funds in the Asia-Pacific area outperform the conventional ones, with the implication that the more intense screening process is more rewarding in areas with less developed and efficient financial markets (Auer & Schuhmacher, 2016).
There are other studies on the other hand that find no difference in performance between conventional funds and SRI funds. A previous study conducted with a primary focus on Europe study found no statistically significant difference in the performance between the conventional and SRI funds, indicating no under- or outperformance of either conventional nor SRI mutual funds. The study found that SRI funds in two countries providing positive alphas compared to their counterparts (Cortez, Silva & Areal, 2012). These findings are further supported in a later study, where it was concluded that there was no significant difference in performance between conventional and SRI mutual funds (Leite & Cortez, 2014).
Dutch SRI Performance
The existing literature on the Dutch SRI fund performance is quite limited, however the author still makes important findings. It is found that the previously mentioned pattern of lack of either under
or overperformance remains also in the Dutch SRI market (Scholtens, 2005). It was found however, that the SRI funds were more inclined to hold value stocks rather than growth stocks which has been the case in other SRI portfolios across the world (Scholtens, 2005). A later study by Scholtens (2007), found that SRI funds would benefit from green investments due to tax benefits. The same study eventually concluded however, that SRI funds underperformed their indexes. (Scholtens, 2007) Another study conducted about the SRI funds in Europe found that the SRI funds in the Netherlands underperformed their conventional counterparts with around 1.5% annually (Renneboog et. al, 2005). A later study found that the Dutch SRI funds outperformed their conventional counterparts, providing higher risk-return relationships (Cortez et. al, 2011). The conflicting evidence of performance among the Dutch funds leaves a gap in the current research.
This study aims at further investigating the performance of the Dutch SRI mutual funds compared to conventional mutual funds. Studying returns of SRI funds and conventional funds a decade later might provide new insights into the development and direction of the market.
Data
Fund data
The data collected for the funds used in this research paper is collected from the Refinitiv Lipper Database (2020). The information collected consists of changes in net asset values (NAV) from conventional and SRI mutual funds based in the Netherlands, which are investing in global equities. The data was obtained by selecting “Netherlands” for domicile, “Mutual Fund” for asset type, “Global” for geographical focus. To distinguish SRI funds from conventional funds, ethical funds were filtered and extracted from the total sample. This yields two initial samples. The SRI fund sample consisted of 23 funds and the conventional sample of 41 funds. All funds younger than three years, were excluded from the sample with an exception of one conventional fund having 34 months of observations. The final sample consists of 30 conventional mutual funds and 19 SRI funds. Monthly returns are calculated from the monthly net asset values and dividend payments. All returns from 2020 were excluded due to the extreme circumstances of the year. The Conventional fund sample dates back to June 2000 and the oldest SRI fund date to May 1993. Extreme returns due to structural changes within the funds were excluded in the analyses.
Management Fees
The dataset is supplemented with the yearly expense ratios, derived from the funds’ own websites and from Morningstar (2020). These yearly management fees are then calculated as monthly management fees. Monthly after-fee returns are then derived by subtracting the monthly management fees from the monthly excess returns. These are only the management fees however. Hidden costs such as costs of sales or brokerage costs are not included. Fund names, fund fees can be found in the appendix in tables 1 and 2 for the SRI funds and the conventional funds respectively. Summary statistics for the two fund types before and after fees can be found in tables 3 and 4 for the SRI funds, and tables 5 and 6 for the conventional funds. In this sample there is little difference in the annual management fees between the two types of funds. Mostly the management fees were around 0.5% annually, with the highest being 1% and the lowest 0%. This indicates a difference from what previous research has found (Leite & Cortez, 2014). The higher fee structures were previously discussed in the literature review.
Market Data
The data sampled for the risk-free rate is the one-month T-Bill rate, downloaded from the website of Kenneth French. (2020). The market performance is proxied by the Fama-French Developed Markets index, which was collected from the website (2020). This index functions as a proxy for the global market, and functions as a benchmark for analysis. The index is broad, including equity from small-, medium- and large-cap stocks. This was selected in order to have a non-biased benchmark for global equity markets.
Factors
The monthly factors were obtained from Kenneth French’s website. (2020) The data set for the Five Factor Developed Markets will be used. This data set cover the largest area, hence will improve reliability.
The Small-Minus-Big (SMB) factor is the average return on nine small cap stock portfolios minus the average return on nine big stock portfolios:
SMB(B/M) =
SMB(OP) =
1/3 (Small Value + Small Neutral + Small Growth) - 1/3 (Big Value + Big Neutral + Big Growth).
SMB(INV) =
SMB =
1/3 (Small Robust + Small Neutral + Small Weak) - 1/3 (Big Robust + Big Neutral + Big Weak).
1/3 (Small Conservative + Small Neutral + Small Aggressive) - 1/3 (Big Conservative + Big Neutral + Big Aggressive).
1/3 (SMB(B/M) + SMB(OP) + SMB(INV) ).
HML (High Minus Low) is the average return on the two value portfolios minus the average return on the two growth portfolios:
HML = 1/2 (Small Value + Big Value) - 1/2 (Small Growth + Big Growth).
RMW (Robust Minus Weak) is the average return on the two robust operating profitability portfolios minus the average return on the two weak operating profitability portfolios:
RMW = 1/2 (Small Robust + Big Robust) - 1/2 (Small Weak + Big Weak).
CMA (Conservative Minus Aggressive) is the average return on the two conservative investment portfolios minus the average return on the two aggressive investment portfolios:
CMA = 1/2 (Small Conservative + Big Conservative)
- 1/2 (Small Aggressive + Big Aggressive)
Limitations to the data
In this dataset there are a few limitations that are worth mentioning. The deficiencies are important to consider as they pertain to the results of the analyses. The first quite clear issues with the data are sample size and sampling. As the total sample of SRI funds is 19, it is below the threshold
of sample size at 30. The small size of the sample limits the generalizability of the findings in this research and threatens the external validity of the research.
The data set available from the Dutch market also suffers from survivorship bias. This is the bias in fund performance that arises when the funds that have ceased to exist are excluded from the analysis (Bodie, Kane & Marcus, 2014). The exclusion of the “dead” funds is likely to paint a better picture of the funds’ performance, as only the highest performing funds tend to survive (Bodie, et. al., 2014). The implications of this will be discussed later on in the discussion section.
Another important limit is the lack of data regarding the “hidden fees”. These are fees that arise from turnover rates, marketing costs and other operating costs (Bodie, et. al., 2014). These costs can drastically alter the performance of a fund and are important to include to accurately assess differing performances, and the sources of these differences. (Bodie, et. al., 2014).
Methodology
Performance measures
In order to test the hypothesis mentioned two regression models will be used and statistically tested. The Sharpe Ratio developed in 1964 will also serve as a basis of measuring fund performance (Sharpe, 1964). Ordinary Least Squares (OLS) factor model regression will be utilized for the Fama French model. Jensen’s Alpha is used to determine fund overperformance versus the exposure to certain risks. In his study from 1968, Jensen derived the alpha by utilizing one of the most widely used models, which still today serves as the basis for many more developed models today: The Capital Asset Pricing Model (Jensen, 1968). Equations are numbered and the formula number is displayed before each equation.
CAPM
The basic capital asset pricing model will also be utilized. This is one of the most basic models used within finance, and is a fundamental model for evaluating the theoretical required return for level of risk exposure. The model was first introduced by Traynor (1961, 1962) and further developed by Sharpe (1964), Lintner (1965) and Mossin (1966). The model measures the rate of return required based on an assets exposure to market risk (Sharpe, 1964). Due to its simplicity it is widely used even today, almost 60 years later. However due to this simplicity it is naturally lacking in explanatory power, and is based on limiting assumptions, which is the cause for the lack of explanatory power. (Fama & French, 2004). The formula for CAPM is:
(1) 𝑅p = Rf+𝛽p*(𝑅𝑚 − 𝑅𝑓) + 𝑒𝑝
Where:
Rp stands for expected return of portfolio
Rf stands for the risk-free rate (Usually derived from a three-month T-Bill) ßp is the beta of the fund, compared to the market or benchmark
Rm is the expected return of the market fund
Jensen’s Alpha
In 1968 Jensen used the CAPM when he derived his model (Jensen, 1968). The alpha is the excess return a portfolio yields, above its expected, predicted return. Positive alphas indicate overperformance, and negative alphas indicate underperforming portfolios. The model measures the performance of a fund or portfolio compared to a single benchmark portfolio. Jensen’s alpha is derived from:
(2) 𝑅p − 𝑅𝑓 = 𝛼𝑝 + 𝛽p (𝑅𝑚 − 𝑅𝑓) + 𝑒𝑝
Where
𝛼p stands for Jensen’s alpha
Rp stands for the return of the portfolio Rf stands for the risk-free rate
ßp is the beta of the fund, compared to the market
Rm is the expected return of the market fund or benchmark. 𝑒𝑝 is the error term
The Fama-French Five Factor Model
Another model that has been further derived from the CAPM is the model developed by Eugene Fama and Kenneth French (Fama & French, 2015). This is a much more complicated arbitrage pricing theory model, which considers different important factors, when determining fund, portfolio or stock performance. The five-factor model is given by:
Where:
Rp stands for the return of the portfolio Rf stands for the risk-free rate
α𝑝 is the alpha of the excess return
ßp is the beta of the fund, compared to the market
Rm is the expected return of the market fund or benchmark SMB controls for the size effect
HML controls for the value effect
RMW is the average return of the weak and strong performing portfolios CMA is the average returns of aggressive versus conservative portfolios.
Sharpe Ratio
One means of answering the research question will be by utilizing a standard financial measure the: Sharpe ratio. The Sharpe ratio measures the risk-reward relationship of an investment, a portfolio or a fund by dividing the excess returns of the funds, that is the returns after subtracting the risk-free rate for the corresponding time period, by the standard deviation of the returns. Note that the standard deviation is calculated from the returns, not the excess returns.
Deriving the Fund Sharpe Ratios
The Sharpe ratios for the SRI funds and the conventional funds, including the after-fee ratios are calculated. The monthly returns are calculated for each fund, and the standard deviation derived from these monthly returns. The monthly standard deviation has been annualized.
The monthly excess returns are calculated by subtracting the corresponding risk-free rate from each month. The average monthly excess return will then be calculated for each fund. These calculations are done individually for each fund, both for each SRI fund and each conventional fund in the sample. The average monthly excess returns are then annualized. The final Sharpe ratio for each fund will be calculated for each fund individually by dividing the annualized excess return of the fund by the annualized standard deviation. These Sharpe Ratios are hereafter denoted as Sharpe Ratios Pre-Fee or Pre-Fee Sharpe Ratio. The Sharpe ratios can be found in the appendix.
After calculating the Pre-Fee Sharpe Ratios, the Sharpe Ratios of the funds after the management fees will be derived. First the monthly after-fee return for each fund is calculated by subtracting the monthly management fee of each individual fund, both SRI and conventional, for each
month. Then the same procedure as for the Pre-Fee Sharpe ratios is followed. Note that the monthly management fees are constants, and the standard deviations do not change. These Sharpe Ratios will hereafter be denoted as Sharpe Ratio Post-Fee or Post-Fee Sharpe Ratios.
For further analysis, market Sharpe Ratios will be calculated. The market ratio will be calculated using the Fama-French Developed Markets Index taken from their website (2020). This index is considered to be a fitting benchmark for global equity investments. The market Sharpe ratio will be calculated using the same procedures mentioned in the section above. The Sharpe ratio of the market will be calculated individually for each fund, corresponding the same duration as the funds. This will allow for a more accurate estimate of under- or outperformance of the funds compared to the market.
These Sharpe ratios will function as a basis for answering the research question by comparing the Sharpe Ratios between the funds, and examining the relative performance of the funds with the market. The procedures for comparison and statistical analyses are mentioned in the following section. The formula of the Sharpe Ratio is as follows:
(4)
Sharpe Ratio =
ABCADEBWhere:
Rp Stands for the return of the portfolio Rf Stands for the risk-free rate
sp stands for the standard dev
Statistical Tests
OLS Regressions
For the two capital asset pricing models, the CAPM and the Fama-French Five Factor Model OLS regression will be utilized. The Betas of each individual fund towards the factors and the alphas from the regressions will be investigated. the alphas will be important points for analysis for the individual funds’ performance, and determine whether or not they are providing positive performance (Fama & French, 2015). In order to be able to derive significant results from OLS regressions a few assumptions must be met.
The data must respect the assumption of homoscedasticity or equal variances, and to test for this a White’s test will be utilized. The null hypothesis of the test is that there is no heteroskedasticity, and the h1 is that there is heteroscedasticity. (White, 1980). In order to ensure this assumption is met, robust standard will be used in the regressions (Stock & Watson, 2015). The results of the White’s test will be discussed in the results section.
Another assumption that needs to be met for the results of the regression to be efficient is the assumption of no autocorrelation or serial correlation (Stock & Watson, 2015). This will be tested for using a Breusch-Godfrey test for the regressions. The null hypothesis of the test is that there is no serial correlation, and the H1 is that there is serial correlation (Breusch, 1978), (Godfrey, 1978). The results of this test will be discussed in the results section.
The last assumption that must be met is the residuals of the regressions must be normally distributed. In order to test for this a Shapiro-Wilk test will be used (Shapiro & Wilk, 1965). The null hypothesis of the test is that the data is normally distributed, and the H1 is that it is. The null hypothesis will be rejected at a 5% significance level. The result of the test will be discussed in the results section.
Hausman Test for Pooled Regressions
In order to determine whether fixed effects or random effects will be used, a Hausman’s test will be used. This will be used for the pooled CAPM regressions and the pooled Fama French Five factor regressions, of both the SRI funds and the conventional. The H0 of the test is that random effects pooled regression is the appropriate and the H1 is that fixed effects is the appropriate method (Hausman, 1978).
Testing for significance
In order to determine differences of the coefficients, a statistical test will be utilized. The Z-test developed in 1998 for Z-testing for significant difference in coefficients (Paternoster, et al., 1998). The null hypothesis of the test is that there is no statistical difference between the coefficients. And the alternative hypothesis is that there is indeed a difference. After the panel regression is conducted for the SRI funds, both Pre- and Post-fee, and the conventional funds pre- and post-fee, the differences between coefficients will be determined. This will be done for the Market coefficient, for the SMB factor, RMW factor, CMA factor and the HML factor. The alphas will also be studied. These analyses will function as a means of analysing differences in exposure towards the factors and differences in alphas. If the Z-statistics of each difference yields values of 1.96 or above the null hypothesis is
rejected at a 5% significance level. This will function as a means of answering the research question and determine if the SRI funds do outperform their conventional counterparts. The formula of the Z-test is as follows:
(5) Z = 𝑏1 − 𝑏2 J(𝑆𝐸1M+ 𝑆𝐸2M)
Where:
b1 is the coefficient for a factor, for the SRI funds.
b2 is the coefficient for the same factor, for the conventional funds SE1 is the Standard Error for the factor coefficient for the SRI funds
SE2 is the Standard error for the factor coefficient for the conventional funds.
Sharpe Ratio Comparison
The Sharpe Ratios will be analysed in two different ways. First of all, the Sharpe Ratios of the funds will be individually compared to the corresponding market Sharpe Ratios. This will determine whether or not the funds are providing comparable risk-reward relationships to the market index, the Fama-French Developed Market Index (French, 2020). This will be done both for the Pre-Fee Sharpe ratios of the funds and the Post-fee Sharpe Ratios of the funds. The Jobson-Korkie Z-test will be utilized to determine the statistical significance of the funds’ differences in their individual Sharpe ratio and the market Sharpe ratio.
The null hypothesis of the Jobson-Korkie Z-test is, that there is no statistically significant difference between the Sharpe ratio of the market and the fund (Jobson & Korkie, 1981). H0 will be rejected if the difference is significant at a 5% level. In order to reject the H0, the Z-statistic will have to reach a critical value of 1.96, which is associated with the previously mentioned 5% significance level. This test will determine whether or not the funds are providing similar risk-reward relationships as the market. The Jobson-Korkie Z-test is as follows (Jobson & Korkie, 1981):
𝑍 = 𝑆𝑅𝑓 − 𝑆𝑅𝑚
Where:
SRf Stands for the Sharpe ratio of the fund.
SRm Stands for the Sharpe ratio of the market, for the same time period T is the number of observations, in months
Pf, m s the correlation between the fund returns and the market returns.
The Hypothesis of the Z-test: H0: SRf – SRm = 0 H1: SRf- SRm ¹ 0
The second analysis of the Sharpe Ratios is the comparison of the Sharpe Ratios of the SRI funds and the Sharpe ratios of the conventional funds. They will be tested against each other to examine whether or not the SRI funds offer similar risk-reward relationships as the conventional funds. The test will also provide the difference between the ratios of the two types of funds, which will allow to answer the research question if Dutch SRI funds do indeed outperform their conventional counterparts. This test will be conducted on both the Pre-Fee Sharpe Ratios and the Post-Fee Sharpe Ratios.
The test that will be utilized is a two-sample t-test. The null hypothesis of the t-test is that there is no statistically significant difference between the two measured means. The H1 is that the Sharpe ratios are large among the SRI funds, the H2 is that the difference between the funds’ Sharpe Ratios are different from zero and finally, the H3 is that the Sharpe Ratios are higher among the conventional funds. In order to reject the H0, the result must be statistically significant at a 5% level. The hypotheses of the t-test are summarized as follows:
H0: Sharpe (SRI) = Sharpe (conventional) H1: Sharpe (SRI) ¹ Sharpe (Conventional) H2: Sharpe (SRI) > Sharpe (conventional) H3: Sharpe (SRI) < Sharpe (conventional)
A Levene’s test for equality of variances will be performed, in order to determine whether or not equal variances in the samples can be assumed (Levene, 1960). This test will be performed both on the Sharpe Ratios Pre-fee and Post-fee. The null hypothesis of Levene’s test is that there is no
difference in variances between the two samples. The H1 of the test is that there is a difference in variances of the Sharpe ratios. This will allow to perform the t-test mentioned in the previous section, assuming either equal or unequal variances. In order to reject the null hypothesis and assume unequal variances the p-value of the test must be below 0.05.
Hypotheses
Based on previous literature and studies, and the methodology of conducting analyses on this topic six hypotheses have been formed in order to answer the research question. These will be presented in the following section.
As previous literature has found evidence suggesting both under- and outperformance of both SRI funds and conventional funds in terms of risk-reward relationships, this will be tested in this sample. This will function as a means of examining the performance of individual funds for the risk of the same fund. The Capital Asset Pricing Model and the Fama-French factor model will be used to determine the performance of the funds using Jensen’s alpha. The hypothesis will be that Dutch SRI funds and Conventional funds have alphas differing from zero. The null hypothesis will be that they are zero. If the alphas, yielded by the regressions, are statistically significant at a 5% level, the null hypothesis will be rejected. The mathematical expression of the first hypothesis is as follows:
1. Hypothesis regarding alphas, pre-fee H0: α(fund) = 0
H1: α(fund) ≠ 0
Furthermore, it is found in previous research is that there is a lack of performance difference between the conventional funds and SRI funds (Leite & Cortez, 2014). In order to arrive at conclusions regarding the performance between these funds, pooled regressions will be utilized in combination with the individual regression analyses, in order to determine the overall performance of both fund types.
The pooled regression will yield “pooled” alphas for both the conventional and the SRI funds. These alphas will be compared in order to determine differences in performance, using the Z-test discussed in the methods section above (Paternoster, et al., 1998).
The null hypothesis will be, that there is not a difference between the pooled alphas of the conventional and SRI funds. The H1 will be that there is no difference. In order to reject the null
hypothesis, the difference between the alphas must be significant at a 5% level. The mathematical expression of the second hypothesis is as follows:
2. Hypothesis regarding alphas, Pre-fee: H0: α(SRI) - α(Conventional) = 0 H0: α(SRI) - α(Conventional) ≠ 0
The literature is extensively discussing the importance of fees as a key driver of performance. It is found in the bulk of the literature that SRI funds tend to have higher fees than their conventional counterparts, and that these higher fees do not translate into higher performance. In other words, the higher fees required by the SRI funds do not achieve higher performance and reduce the returns for the investors (Revelli & Viviani, 2014), (Capelle-Blancard & Monjon, 2014).
Based on previous literature, this assumption will be the base for the third hypothesis. As the fees of the SRI’s tend to be higher, with no improved returns offered, it is expected that the SRI funds will perform worse after fees are accounted for. A Pooled Fama- French five factor and CAPM regressions will be used to determine the after-fee alphas of both the conventional and SRI funds. This difference will be measured, and the significance tested by using the Z-test developed by Paternoster, et al. (1998).
The null hypothesis, is that there is no difference between the post-fee alphas. The H1 is that the Dutch SRI funds pooled post-fee alpha will be lower than the pooled conventional post-fee alpha, which is in line with the previous literature (Capelle-Blancard & Monjon, 2014). This will be an important way of examining the difference in after fee performance of SRI funds, and allow for answering the research question. In order to reject the null hypothesis, the difference between the post-fee alphas must be significant at a 5% level. The mathematical expression of the hypothesis regarding post-fee alphas is as follows:
3. Hypothesis regarding Post-fee Alphas H0: α(SRI) - α(Conventional) = 0 H1: α(SRI) - α(Conventional) < 0
Another commonly used measure in determining fund performance are the Sharpe ratios. Each funds’ ratio, both SRI and conventional, will be compared with the market Sharpe ratio in order to examine if the funds are providing similar risk-return relationships as the market. (Sharpe, 1964). This is however not expected, as it has been overwhelmingly found that both SRI and conventional
funds underperform the market (Hamilton et al.,1993), (Renneboog et. al, 2005). Hence the first hypothesis regarding the Sharpe Ratios is that the Sharpe ratios of the funds differ from the Sharpe Ratio of the market. The statistical significance of the difference between the market and the funds’ ratios will be tested using the Jobson & Korkie Z-test discussed in the methodology section (Jobson & Korkie, 1981). The null hypothesis will be rejected if the Z-statistic is above the critical value of 1.96 or below the critical value -1.96. The mathematical expression of the fourth hypothesis is:
4. Hypothesis Pre-Fee Sharpe Ratios of the Funds versus the Market H0: Sharpe Ratio Fund = Sharpe Ratio Market
H1: Sharpe Ratio Fund ≠ Sharpe Ratio Market
In order to determine whether the risk-return relationships are higher among SRI funds or not, the Sharpe ratios of both types of funds will be compared. As the previous literature finds a lack of difference between the funds, the null hypothesis will be that there is no difference between the Sharpe ratios of the SRI funds and the Sharpe Ratios of the conventional funds. The H1 is that there is a difference between the two types of funds. This will be tested by utilizing a T-test, of comparison of means, where unequal variances are assumed. The null hypothesis is, that there is no difference in the means. The null hypothesis will be rejected if the difference is statistically significant at 5% level. This test will also determine whether or not the SRI funds provide improved risk-return relationships compared to the conventional counterparts, by examining the difference. The t-test is explained in the methodology section. The mathematical expression is as follows:
5. Hypothesis regarding differences in Sharpe ratios, Pre-Fee, between the funds. H0: Sharpe Ratio SRI Funds = Sharpe Ratio Conventional
H1: Sharpe Ratio SRI Funds ≠ Sharpe Ratio Conventional
The same T-Test utilized in the fifth hypothesis will also be utilized for the final hypothesis. This will test whether or not the difference between after-fee Sharpe ratios have either become significant, depending on if it was not in the previous test, or whether it still is non-significant. The after-fee Sharpe ratios should offer worse risk-return relationships among the SRI funds compared to the conventional funds. The null hypothesis is yet again that the Sharpe ratios are equal between the fund categories, and the H1 is that the Sharpe ratios differ. The null hypothesis will be rejected if the difference is statistically significant at a 5% level. The mathematical expression is the following:
6. Hypothesis regarding differences in Sharpe ratios, Post-Fee, between the funds. H0: Sharpe Ratio SRI Funds, after-fee = Sharpe Ratio Conventional, after-fee H1: Sharpe Ratio SRI Funds, after-fee ≠ Sharpe Ratio Conventional, after-fee
Results
In this section, the results of the analyses will be presented and the hypotheses will be tested in the order presented in the previous section. First the results and the implications of the White’s tests, the Breusch-Godfrey tests, and the Shapiro-Wilk’s tests will be discussed and presented.
Secondly, the individual OLS CAPM and Fama-French Five factor regression results will be discussed and the implications presented. Thereafter, the pooled regression results will presented be for both CAPM and the Fama-French Five Factor models and the results from the Paternoster, et al., Z-test will be discussed. This will allow for conclusions to be made regarding the three first hypotheses.
After the regression models and the statistical tests have been presented, the Sharpe ratios will be summarized and presented. First, the results of the comparison between the fund Sharpe ratios and the corresponding market Sharpe ratios, together with their statistical significance derived from the results of the Jobson-Korkie Z-tests will be presented, and allowing for either accepting or rejecting the fourth hypothesis.
Finally, the results of the analysis between the Sharpe ratios of the conventional funds and the SRI funds, both before- and after-fee, will be presented. The result of the T-test of comparison of means will be discussed, allowing for conclusions to be drawn regarding the fifth and the sixth hypotheses.
OLS Assumption Testing
The OLS regression test results for the Breusch-Godfrey tests, the White’s tests and the Shapiro-Wilk’s tests can be found in the appendix, in tables 19 and 20.
The Shapiro-Wilk’s test, for the SRI funds and the Conventional funds, shows that the data for many of the funds is not normally distributed. Of the SRI funds fifteen out of nineteen funds have a non-normal distribution. Of the conventional funds, twenty out of thirty had non-normal distributions.
An explanation for the non-normal data, are extreme events more frequently occurring in the financial markets than a normal distribution would allow for. These events are for example economic downturns or crashes such as in 2007-2009, which often are followed by strong economic upswings (Sheikh & Qiao, 2010). This causes financial returns to be non-normally distributed, leading to a skewed distribution (Sheikh & Qiao, 2010). In this sample, the non-normality of the data is also prevalent in the market proxy, the Developed World Index of Fama and French (2020).
The non-normal distribution of the fund returns limits the conclusions that can be drawn from the results but still allows for some judgement from the results. This will be further elaborated in the discussion section.
The Breusch-Godfrey test for autocorrelation or serial correlation, with the H0 that there is no serial correlation between variables, mixed results. For the SRI funds, it is found that eight individual SRI fund regressions for CAPM do suffer from serial correlation and for the Fama-French Five factor regressions, twelve suffer from serial correlation.
For the conventional funds’ individual regressions, twenty three out of thirty funds do not suffer from autocorrelation of the CAPM regressions, and twenty-one of the Fama-French Five Factor regressions do not suffer from autocorrelation.
Autocorrelation indicates that the monthly observations in the sample are correlated, which violates the assumption of independence of observations. This causes the predictions of the OLS regressions to be unbiased and consistent, but inefficient (Mukherjee, Laha, 2019). Yet again this will be taken into consideration, and the implications of this will be presented in the discussion section.
The White’s test shows that there are funds in the sample that have heteroskedastic data. Of the SRI CAPM regressions, six funds show heteroskedasticity, and six of the Fama-French regressions. Of the conventional funds, six of the CAPM regressions showed heteroskedasticity and fourteen of the Fama-French Five factor regressions show heteroskedasticity. The issue of heteroskedasticity will be corrected for by using robust standard error in all regressions.
The Hausman’s test resulted in not rejecting the null hypothesis. This was the case for both the SRI funds before and after fees, and the conventional funds before and after fees. This test result indicates that random effects regressions will be used for all pooled regressions.
Individual Regression analysis SRI Pre-fee CAPM & Fama French Five Factor
Table 7, containing the explained variances (R2) for each of the CAPM regressions and the Fama-French Five Factor regressions and as well as alphas of the SRI funds pre-fee can be found in the appendix.
The CAPM regressions yielded high R2 for all individual regressions. The highest R2 in the pre-fee SRI sample was 0.993 and the lowest 0.738. This indicates a good fit of the regression model for the data.
The alphas provided by the CAPM regressions were mostly negative. Of the 19 SRI funds only four had positive pre-fee alphas. None of these alphas were statistically significant however at a 10% level or lower. Of all the alphas from the CAPM regressions in total three were statistically significant at a 1% level, a total of five statistically significant at a 5% level and a total of six at a 5% level. All of the statistically significant alphas level of 10% or lower were negative.
As only a total of five funds out of nineteen had statistically significant pre-fee negative alphas, based on the CAPM model, this does not allow for rejecting the H0 of the first hypothesis which states the alphas differ from zero among the Dutch SRI funds. However, this hypothesis was rejected for five individual funds, all of which had statistically significant negative pre-fee alphas at a 5% level. The implications of the CAPM regressions will be discussed in the discussion section.
The Fama-French Five Factor model yielded similar results for the SRI funds pre-fees as the CAPM. The R2 of the individual regressions were continuously high, indicating a good fit of the model to the data. The highest R2 was 0.997 and the lowest 0.759, and the R2 for every individual regression was higher for the Fama French five factor model compared to the CAPM. This shows that the five-factor model has higher explanatory power than the CAPM, and that the CAPM suffers from omitted variable bias. This will be further elaborated upon in the discussion.
The pre-fee alphas of the SRI funds yielded from the Fama French Five Factor model were also negative. Of the nineteen SRI funds, three funds had positive pre-fee alphas none of which were statistically significant at a 10% level or lower.
A total of three out of nineteen pre-fee alphas were statistically significant at a 1% level, a total of six out of nineteen pre-fee alphas were statistically significant at a 5% level and finally a total of eight out of nineteen pre-fee alphas were statistically significant at a 10% level. This indicates that the five-factor model has more statistical power compared to the CAPM.
We cannot reject the first hypothesis stating that pre-fee alphas being different from zero among SRI funds overall, based on the Fama-French Five factor model. This hypothesis was rejected for the six of the SRI funds for which it is concluded that the pre-fee alphas were negative.
Individual Regression analysis SRI Post-fee CAPM & Fama French Five Factor
Table 8 containing the R2 for each of the CAPM regressions and the Fama-French Five Factor regressions as well as alphas of the SRI funds post-fee can be found in the appendix.
The post-fee CAPM regressions yielded high R2 for all individual regressions. The highest R2 in the post-fee SRI sample was 0.993 and the lowest 0.738. This indicates a good fit of the regression model for the data.
The post-fee alphas of the CAPM regressions were lower compared to those before fees, however the total statistically significant alphas increased compared to the pre-fee CAPM regressions. Of the post-fee alphas from the CAPM regressions a total four post-fee alphas were statistically significant at a 1% level, a total of six post-fee alphas at a 5% level and a total of seven post-fee alphas were statistically significant at a 10% level. All of the statistically significant post-fee alphas yielded from the CAPM regressions were negative. This allows for the H0 of the first hypothesis to be rejected for six out of the total of nineteen SRI funds, after fees. We cannot reject the hypothesis for the funds overall based on individual analysis alone, as thirteen of the nineteen SRI funds’ post-fee alphas were not significant.
The Fama-French Five factor model also yielded high R2 for the post-fee regressions. For each individual regression the R2 was higher using the Fama-French five factor model, yet again indicating a better fit of the model and omitted variable bias. (Stock & Watson, 2015). The highest R2 was 0.997 and lowest 0.756 (See table 8).
A total of five post-fee alphas of the individual five factor regressions were significant at a 1% level, a total of 8 were statistically significant at a 5% level and nine at a 10% level. All of the statistically significant post-fee alphas were negative at a 10% level or lower. Of the total sample of SRI post-fee regressions, only two post-fee alphas were above zero, but as previously mentioned none of the positive post-fee alphas were statistically significant.
This once again allows for rejecting the null hypothesis of the first hypothesis, that the fund alphas equal zero for eight out of nineteen of the post-fee alphas among the SRI funds. This however, does not allow for rejecting the H0 for all SRI funds after fees are accounted for. The implications of this will be further elaborated upon in the discussion section.
Individual Regression analysis Conventional Pre-fee CAPM & Fama French Five Factor
Table 9 containing the R2 for each of the CAPM regressions, the Fama-French Five Factor regressions, and the pre-fee alphas of the conventional funds can be found in the appendix.
The pre-fee CAPM regressions yielded mostly high R2 for all individual regressions of the conventional funds. The highest R2 in the pre-fee conventional fund sample was 0.977 and the lowest 0.58. This indicates a generally good fit of the model for the data.
Of the CAPM regressions a total of five out thirty pre-fee alphas were statistically significant at a 1% level, a total of eight out of thirty pre-fee alphas were statistically significant at a 5% level and a total of thirteen out of thirty pre-fee alphas were significant at a 10% level.
Of the thirty CAPM regressions only four funds had positive alphas, and only one of these were significant at 5% level and two in total at a 10% level. This does not allow for rejecting the H0 of the first hypothesis for the funds overall
The pre-fee Fama-French five factor regressions yielded mostly high R2 for all individual regressions of the conventional funds. The R2 in the pre-fee conventional fund sample ranged from 0.989 and 0.619. This indicates a good fit of the model for the data. The R2 for the five factor regressions were higher compared to the R2 of the CAPM. This indicating omitted variable bias, which will be discussed further.
A total of five out of thirty pre-fee alphas were significant at the 1% level, a total of ten out of thirty pre-fee alphas were significant at a 5% level and a total of fourteen pre-fee alphas were significant at a 10% level.
Three alphas out of thirty were positive, and one of these significant at a 5% level. This does not allow for rejecting the H0 of the first hypothesis for the pre-fee conventional funds based on the Fama French Five Factor Model for the conventional funds overall. However, this null hypothesis was rejected for ten of the funds, and the alphas were negative for nine of them.
Individual Regression analysis Conventional Post-fee CAPM & Fama French Five Factor
Table 10 containing the R2 for each of the CAPM regressions, the Fama-French Five Factor regressions and the post-fee alphas of the conventional funds can be found in the appendix.
The post-fee CAPM regressions yielded mostly high R2 for all individual regressions of the conventional funds. The highest R2 in the post-fee conventional fund sample was 0.977 and the lowest 0.586. This indicates a generally good fit of the model for the data.
Of the CAPM regressions a total of eight out thirty post-fee alphas were statistically significant at a 1% level, a total of eleven out of thirty post-fee alphas were statistically significant at a 5% level and a total of sixteen out of thirty post-fee alphas were significant at a 10% level.
Of the thirty CAPM regressions only three funds had positive post-fee alphas, and only one of these were significant at 5%. This does not allow for rejecting the H0 of the first hypothesis for
the funds overall, that the post-fee alphas of the funds differ from zero, as only just over half the sample had statistically significant alphas, fifteen of which were negative.
The pre-fee Fama-French five factor regressions also yielded mostly high R2 for all individual regressions of the conventional funds. The highest R2 in the pre-fee conventional fund sample was 0.989 and the lowest 0.624. This indicates a generally good fit of the model for the data. The R2 for the five factor regressions were higher compared to the R2 of the CAPM. This indicating omitted variable bias in the CAPM, which will be discussed further.
A total of nine out of thirty post-fee alphas were significant at the 1% level, a total of fifteen out of thirty post-fee alphas were significant at a 5% level and a total of twenty post-fee alphas were significant at a 10% level.
Three alphas out of thirty were positive, and one of these significant at a 5% level. This does not allow for rejecting the H0 of the first hypothesis for the pre-fee conventional funds based on the Fama French Five Factor Model for the conventional funds overall, it does indicate a trend of underperformance of the conventional funds after fees however. Th null hypothesis was rejected for fifteen of the funds, and the post-fee alphas were negative for nine of them.
Pooled Regressions
The pooled regressions of SRI funds and conventional funds were compared with each other. First the results of the panel regressions of both CAPM and the Fama French Five factor model before fees will be presented individually for each type of fund. Thereafter the comparison between the two fund types and the result of the Paternoster Z-test will be presented. This will first be done for the two fund types pre-fee, then post-fees.
SRI Funds Pre-Fee Panel Regression
The panel regression results for the SRI funds before fees from the CAPM and the Fama French Five Factor model regressions is posted below.
Table: Pooled Regression Results, SRI funds Pre-Fee
Factors: SRI Factor Betas SE:
Market (CAPM 1.0233*** 0.0168
Alpha (CAPM) -0.0017*** 0.0003
R2 (CAPM) 0.8547 -
Market (Five Factor) 1.0216*** 0.0160
SMB -0.1103*** 0.0351
HML -0.0454 0.0359
RMW 0.0313 0.0808
CMA -0.0310 0.0368
Alpha (Five Factor) -0.0016*** 0.0005
R2(Five Factor) 0.8576 -
***P< 0.01, **P<0.05, *P<0.1. Where: The Factors are the factors loadings, the SRI Fund Betas are the pooled betas for each factor, the SE are the Standard Errors of each of the coefficients and the two different alphas are the alphas of the monthly returns from each pooled regression model and the R2 is below the alphas for each model.
For the CAPM panel regression the market factor is statistically significant at a 1% level and so pre-fee alpha is likewise significant at a 1% level. The alpha is negative at -0.0017, indicating that the SRI funds underperform with 0.16% per month, based on the risk exposure to the market. The R2 of the CAPM is 0.8547 indicating a good fit of the model for the data. Based on the panel regression results from the CAPM the null hypothesis of the first hypothesis of the SRI fund pre-fee alpha being equal to zero, at a significance level of 5%. It is concluded that the Dutch SRI funds in this sample do not deliver positive performance due to the alpha being negative.
The Fama-French Five-Factor model panel shows similar results as the CAPM for the pre-fee SRI funds. The SMB factor and Market factor are statistically significant at a 1% level while none of the other factors are. The SMB is negative at -0.1103, indicating that the Dutch SRI equity funds tend to be more exposed to large cap stocks and this has a negative impact on the returns of the funds. (Fama & French, 2015). The market factor coefficient is slightly above one at 1.0216, indicating that the SRI portfolios are more volatile, or have higher standard deviations, than the market (Berk & DeMarzo, 2016). The alpha is also statistically significant at a 1% level, indicating that the SRI funds underperforms with 0.17% per month, based on the exposure to the different factors.
The results are very similar between the two regressions. The reason for this being the similar exposure towards the market, having high explanatory power in both models. The small difference likely is due to the added variables. This will be further elaborated in the discussions section.
Conventional Funds Pre-Fee Panel Regression
The panel regression results for the conventional funds, before fees, from the CAPM and the Fama French Five Factor model regressions is posted below.
Table: Pooled Regression Results, Conventional funds Pre-Fee
Factors: Conventional Funds Betas SE
Market (Capm) 1.0418*** 0.039
Alpha (CAPM) -0.0019*** 0.000
R2 (CAPM) 0.8380 -
Market (Five Factor) 1.0382*** 0.030
SMB -0.0821*** 0.024
HML -0.0821 0.040
RMW 0.0701** 0.028
CMA -0.0929 0.088
Alpha (Five Factor) -0.0020*** 0.000
R2 (Five Factor) 0.8395 -
***P< 0.01, **P<0.05, *P<0.1
Where: The Factors are the factors loadings, the SRI Fund Betas are the pooled betas for each factor, the SE are the Standard Errors of each of the coefficients and the two different alphas are the alphas of the monthly returns. from each pooled regression model and the R2 is below the alphas for each model.
For the CAPM panel regression the market coefficient is statistically significant at a 1% level and the alpha is also significant at 1% level. The market coefficient is larger than one indicating that the conventional funds are more volatile compared to that of the market (Berk & DeMarzo, 2016). The negative pre-fee alpha from the CAPM indicates that the conventional funds underperform based on their exposure to market risk by 0.19% monthly. The CAPM yields high R2 indicating a well-fitting model. According to the CAPM the first H0 of the hypothesis can be rejected, and its concluded that the conventional funds have provide negative returns for their risk exposure.
The Fama-French Five Factor regression shows that the market factor, the SMB factor and the RMW factor are statistically significant. The Market factor is positive at 1.0382 and the SMB factor is negative at -0.0821, both statistically significant at a 1% level. The RMW factor is positive at 0.0701 and is statistically significant at a 5% level. The pre-fee alpha is negative at -0.002 and statistically significant at a1% level. The R2 is high also for this model, at 0.8395.
The positive RMW coefficient indicate higher exposure towards firms with higher profitability which leads to higher returns for the fund. The conventional funds’ negative SMB coefficient indicates larger exposure towards large cap stocks, and this lowers the monthly excess returns. The market coefficient is above 1, which means that the conventional funds are more volatile than the market index, the Fama-French Developed Markets Index (Fama, 2020).
Yet again this allows for rejecting the H0 of the first hypothesis, and concluded that the conventional funds’ alphas do differ from zero. And we can conclude that the conventional funds provide negative alphas, based on the Fama-French Five factor model.
