Word count: 12427
Performance and flow analysis of socially responsible investing mutual funds
A study on Dutch and German mutual funds
Sam Peppelman (S2908336) Master Thesis MSc Finance
University of Groningen
Department of Economics, Econometrics and Finance Under supervision of:
Dr. A. Plantinga
January, 2019 Abstract
This paper analyzes the performance and fund flow of Dutch and German socially responsible investing (SRI) mutual funds and compares it to a conventional portfolio. It finds no significant risk-adjusted returns for the self-assessed SRI portfolio whereas the top ranked quartile and decile SRI portfolios show significant underperformance. Fund flow results are in line with theory. Socially responsible investors irrationally chase self-assessed SRI funds, showing a linear-like flow-performance relation, whereas all other portfolios show the well-known convex flow-performance relation.
Keywords: Socially responsible investing, mutual fund performance, fund flow
2 1. Introduction
The importance of social responsibility arose when Freeman (1984) published his book ‘Strategic Management: A Stakeholder Approach’. This approach presented a break from the prevailing view, namely the social responsibility of business is to increase profits (Friedman, 1970). It prompted the research in corporate governance and stakeholder theory, changing the view on profit maximization. Furthermore, it showed that corporate governance could improve profits in its own way. SRI builds upon the stakeholder view, giving investors the ability to invest in firms that do good to its stakeholders. The Forum for Sustainable and Responsible Invent (US SIF) reported that the assets under management (AUM) of SRI mutual funds increased by 33% to $8.72 trillion since 2014. This highlights the rising importance of research on SRI funds. SRI mutual funds aim to obtain alpha by investing in stocks that are both undervalued and socially responsible. SRI mutual funds actively select securities issued by firms that are all screened for certain socially responsible criteria that matches the mutual fund’s focus. The most common and important screens are the environmental, social and governance (ESG) screens. Management companies can use either external ESG research or conduct their own. SRI mutual funds are assembled according to three different methods. The first method is to exclude firms based on specific themes, usually associated with the nature of the products or services provided by the firm (e.g. the exclusion of firms active within the war, tobacco or the adult entertainment industry). Inclusion is the method where all firms are selected based on positive ESG screens. The best-in-class method selects firms that score the best on certain ESG criteria that matches the mutual fund’s focus. Capelle-Blanchard and Monjon (2014) note that European SRI mutual funds typically use the best-in-class approach while American SRI funds use exclusion screening.
Although the SRI industry underwent significant growth over the past decades, the reason why is puzzling, as there is little evidence that active managed mutual funds outperform passive mutual funds (see e.g., Malkiel, 1995; Gruber, 1996; Wermers, 2000). Active management is defined as investing in stocks that are undervalued based on research and expertise of the mutual fund managers. Active management aims to outperform the benchmark, whereas passive management is defined as tracking a benchmark. Little evidence exists that active mutual fund managers outperform passive mutual funds, especially after considering transaction costs and mutual fund fees.
3 which is the set of portfolios that maximizes the expected return for a given level of risk. The Capital Asset Pricing Model (CAPM) by Sharpe (1964) builds upon the MPT and introduces systematic market risk (beta) and nonsystematic (firm-specific) risk. Nonsystematic risk can be eliminated through diversification, which introduces the main conflict. Since SRI mutual funds focus on socially responsible securities as opposed to the market portfolio, they are unable to fully diversify the nonsystematic risk away. Consequently, socially responsible investors are unable to achieve the mean variance efficient portfolio as they place themselves under the efficient frontier. Therefore, theory dictates that socially responsible investors need to accept either a lower return or a higher risk. However, this theory is based on the assumption that capital markets are efficient.
Although Jensen (1978) states that ‘there is no other proposition in economics which has more solid empirical evidence supporting it than the efficient market hypothesis’, the efficient market hypothesis (EMH) has been under a lot of scrutiny. Fama (1970) was the first to define market efficiency as one that ‘reflects all available information’. This means that there are no arbitrage opportunities and thus that SRI mutual funds are not able to obtain alpha by screening for social responsibility. However, the EMH is based on the assumption that investors are fully rational. Nowadays the EMH is not as commonly accepted as it was in the 1970s. It has become more common for investors to let go of the underlying fundamentals to become noise traders and chase trends, also referred to as the inefficiency hypothesis (Black, 1986). If investors are not rational, arbitrage is possible. Additionally, Ball (2009) argues that the popularity of the EHM causes investors to have too much faith on the set price and do little research themselves, leading to financial crises and ESG scandals. In the past decade we have seen Volkswagen, Facebook, Wells Fargo and many others being prone to scandals. SRI eliminates these risks by screening firms and working hand in hand with the firm to increase their ESG policy. Preventing negative events leads to an outperformance of SRI mutual funds, even though this should not be possible under the fundamental theory in finance.
Even though there is little evidence that mutual funds managers outperform passive benchmarks, investors chase high performing mutual funds. Chevalier and Ellison (1997) and Sirri and Tufano (1998) find a strong relation between lagged past performance and fund flow and refer to this as the flow-performance relation. This highlights the irrationality of investors as little evidence exists that mutual funds consistently outperform passive benchmarks. This raises the question whether the irrational flow-performance relation also exists for SRI mutual funds despite the little diversification possibilities.
4 many firms that try to set concepts, definitions and principles for firms that want to improve their social responsibility, like the Sustainability Accounting Standards Board (SASB), the International Integrated Reporting Committee, Global Reporting Initiative and many others. However, this does not lead to one clear quantifiable score, as there is no litigation regarding whether mutual funds can be defined as socially responsible. This introduces a vast amount of subjectivity. Using Thomson Reuters ESG data to rank mutual funds ensures that all mutual funds are subject to the same ESG screens.
In this paper I analyze the performance and fund flows of all Dutch and German domiciled socially responsible mutual funds. The Thomson Reuters ESG Database allows me to analyze mutual funds by creating one clear quantifiable score, based on screens equal for all mutual funds. This allows this paper to differentiate from previous literature by analyzing two SRI classifications; self-assessed SRI mutual funds and rated SRI mutual funds. I compare these two classifications to conventional investment (CI) mutual funds by calculating factor and alpha differentials. The main results show negative, albeit insignificant alpha differentials for the self-assessed SRI mutual funds and significant negative alpha differentials for the rated SRI mutual funds. The fund flow analysis shows a significant flow-performance relation for self-assessed SRI mutual funds while the results for the rated SRI and conventional mutual funds are not significant.
The remainder of this paper proceeds as follows. I start with a literature review in section two, following with the research method in section three. I discuss the data in section four, the empirical results in section five and conduct robustness tests in section six. In section seven I discuss the limitation of this paper and end with the conclusion in section eight.
2. Literature review
While the popularity of socially responsible investing has undergone significant growth over the past few decades, the underlying ideology dates back to the 18th century in which John Wesley touched upon the importance of investing in ‘non-sinful’ firms in his sermon (The Use of Money, 1872). In stark contrast, Adam Smith (1776) argues that individuals only care about maximizing self-interest (i.e. wealth). Is it possible for both ideologies to converge and create a society in which self-interest is maximized without investing in sinful firms?
5 to give up profitable projects due to them not being socially responsible (e.g. weapons). The other viewpoint is that CSR and CFP complement each other. CSR can improve a firm’s reputation, due to higher employee loyalty (Moskowitz, 1972) which in turn leads to lower employee turnover and lower training costs, increasing financial performance. Dowell, Hart, and Yeung (2000) find a positive relation between environmental performance and market value, measured by Tobin’s Q. Guenster et al. (2010) find a positive relation between eco-efficiency and both operating performance and market value. However, after all the extensive research, there is still no clear evidence that allows for generalizable conclusions for either viewpoint. Wright and Ferris (1997) find a negative relationship, Aupperle, Carroll, and Hatfield (1985) find no significant relationship between CSR and CFP, Tsoutsoura (2004) finds a significant correlation and Griffin and Mahon (1988) find a positive significant relationship that leads to decreased firm risk. Thus the results on the relationship are inconclusive, as also argued by Jones and Wicks (1999) in their meta-analysis.
There are several methods to compare the performance of socially responsible securities to conventional securities. The first method focuses on the performance of social indexes. Both Sauer (1997) and Statman (2000) examine the performance of the Domini 400 Social Index compared to the S&P500 and find no significant performance differences. The same holds for Schröder (2007) in his performance analysis of twenty-nine social indexes.
The second method is to compare the performance of SRI mutual funds to CI mutual funds. Statman (2000) finds that SRI mutual funds and CI mutual funds perform equal under the CAPM model. Gregory, Matatko, and Luther (1997) include small firm effect (SMB) within the three-factor model of Fama and French (1993) and find a small significant positive effect. Bauer, Koedijk, and Otten (2005) find no significant performance differences based on the three-factor model and Carhart four-factor model (1997). Renneboog, Ter Horst, and Zhang (2008) also find no significant performance differences for the US, Europe and most Asian countries. Bauer, Derwall, and Otten (2007) use both single- and multifactor models to analyze the performance differences in the Canadian market, but also find no significant results.
6 negative screening method. As with the studies on CSR and CFP, there are no generalizable conclusions regarding SRI performance. Derwall, Koedijk, and Ter Horst (2011) argue that the mixed results are caused by the use of different screening methods and that the previous findings are not complements, but supplements. Revelli and Viviani (2015) perform a meta-analysis of 85 socially responsible performance studies and argue that the mixed results are due to the vast variety in the methodology of the studies.
One of the first studies that focusses on the theoretical background behind socially responsible investing is by Hamilton, Jo, and Statman (1993). They argue that there are three hypotheses when comparing price differences between SRI mutual funds and CI mutual funds. The return of SRI securities either underperform, outperform or perform equal to conventional securities. They argue that equal performance is a result of socially responsible factors not being related to risk. Under the MPT, only factors that are proxies for risk should be priced in the asset.
Underperformance is the result of a breach in MPT. As discussed before, risk-averse investors want to optimize their risk-return ratio. Since SRI mutual funds have a smaller investment universe, the unsystematic risk cannot be fully diversified away. An alternative explanation is the shunned-stock hypothesis (Statman and Glushkov, 2009). Shunned-stocks are defined as securities associated with alcohol, weapons, gambling, nuclear, tobacco and adult entertainment operations (also known as the sextet of sin). Under this hypothesis, the exclusion of shunned-stocks leads to a higher demand for socially responsible stocks and a lower demand for conventional stocks. The lower demand for conventional stocks leads to lower risk-sharing possibilities and thus requires a higher risk premium (Merton, 1987). In addition, Hong and Kacperczyk (2009) find evidence that the classical controversial stocks (alcohol, gambling and tobacco) outperform SRI stocks due to the increased litigation risk. This implies that both socially responsible and controversial characteristics are priced by the market. Socially responsible investors incur an opportunity cost when avoiding shunned-stocks.
7 tools to value intangible CSR assets into the free cash flow. Once again this shows that the markets are not as efficient as originally thought.
If markets are inefficient, arbitrage can be obtained. SRI mutual funds try to obtain arbitrage by screening for socially responsible securities. I previously discussed that mutual funds have three methods; positive screening, negative-screening and the best-in-class approach. Another variable is the number and intensity of screens. Barnett and Salomon (2006) find a curvilinear relationship between the number of screens used by the SRI fund and financial performance, which offsets the lack of diversification by increased performance. They conclude that only the funds that are ‘stuck-in-the-middle’, i.e. giving up diversification possibilities by only using a few screens, achieve suboptimal performance. High screening intensity offsets the loss of diversification possibilities, which can be linked to the active versus passive management debate. In contrast, Lee et al. (2010) find that the number of screens is negatively correlated with fund performance, while Guerard (1997) concludes that screened portfolios do not differ from unscreened portfolios. However, Trinks and Scholtens (2017) conclude that the risk-adjusted performance differential depends on the type of screens. The sextet of sin leads to significant underperformance and adult entertainment and stem cell screens leads to significant outperformance.
A lot of research has been done on the fund flow of CI mutual funds. Ippolito (1992) finds a positive relationship between fund performance and fund growth. Additionally, Sirri and Tufano (1998) find a positive flow-performance relationship, also known as the convex flow-performance relationship. This implies that investors tend to buy high performing mutual funds more than they sell low performing mutual funds. Sirri and Tufano (1998) investigate different measures of past performance and find that individual investors mostly use raw performance measures to select mutual funds. Clifford et al. (2011) confirm this, as they find no significant risk coefficients while using raw returns. Both Bollen (2007) and Benson and Humphrey (2008) find that fund flow of SRI mutual funds are less sensitive to historical fund performance, meaning that SRI investors have some non-pecuniary criteria in mind.
8 socially responsible. Nilsson (2008) confirms this phenomenon with his survey among 500 private investors.
Alternatively, Fisher’s Separation Theorem (1930) states that firms should invest solely in profit maximizing opportunities. Hirschleifer (1988) extends this and argues that investors should separate investment decisions from consumption decisions. If investors prefer social responsibility, they should invest to maximize investment profits and donate part of it to charity. Lantos (2002) concludes that altruistic publicly listed firms are immoral since they violate shareholder rights. This gives investors an irrational feeling of altruism due to their belief of investing in good. Renneboog et al. (2008) test whether SRI performance can be explained using an Ethics factor but find no significant results.
From previous literature we learn that additional work in this field is necessary. As most studies focus on the performance of partnered mutual funds, bias is induced. During the next sections of this paper I analyze the performance and flow of SRI mutual funds and try to eliminate the biases by creating my own definition of socially responsible mutual funds.
3. Research methodology
This section discusses the performance and mutual fund flow measures of socially responsible mutual funds. I conduct two different time-series analyses. I start out by individually analyzing the mutual funds under the Carhart four-factor model (1997), which is the main model from which this paper draws conclusions. After individually analyzing the mutual funds, I follow the calendar-time portfolio method of Jaffe (1974) and Mandelker (1974) and construct equally weighted calendar-time portfolios in the following manner
Ri,t = 1 Nm,t ∑ Dz,t mR m,t Nt z=1 (1)
where Nm,t is the amount of mutual funds belonging to fund type m in at time t; Dz,tm is
a dummy that takes on a value of one if the mutual fund belongs to fund type m and zero otherwise; and Rm,t is the return of mutual fund type m at time t. The risk-adjusted
returns are estimated with three different empirical asset pricing models.
9 returns. The Fama-French three-factor model (1993) is estimated in the following manner
Ri,t – Rf,t = αi + βi(RMt – RFt) + SiSMBt + ViHMLt + εi,t (2) where Ri,t is the return on portfolio i in period t; Rf,t is the risk-free rate in period t; RMt –
RFt is the excess return of the market portfolio over the risk-free rate in period t; SMBt
is the size effect for period t; HMLt is the value effect for period t; and εi,t is the residual
term for portfolio i in period t. The third model is based on Carhart (1997), which extends the three-factor model with a momentum factor UMD, which is defined as the difference between the return of a portfolio of high return stocks and a portfolio of low return stocks. The Carhart four-factor model (1997) is estimated in the following manner
Ri,t – Rf,t = αi + βi(RMt – RFt) + SiSMBt + ViHMLt + MiUMDt+ εi,t (3) where UMDt is the Momentum factor for period t. The third model is the five-factor
model by Fama and French (2015). This also builds upon the three-factor model and includes a profitability factor RMW and an investment factor CMA. The profitability factor is defined as the difference between the return of a portfolio of robust returns and a portfolio of weak returns. The investment factor is the difference in returns between a portfolio of conservative assets and a portfolio of aggressive assets. The Fama-French five-factor model (2015) is estimated in the following manner
Ri,t – Rf,t = αi + βi(RMt – RFt) + SiSMBt + ViHMLt + PiRMWt+ IiCMAt + εi,t (4) where RMWt is the profitability factor for period t; and CMAt is the investment factor for
period t. In addition of testing the returns on the SRI and CI portfolios, differentials are calculated by subtracting the return of the conventional portfolio from the SRI portfolio and regressing the same factor model.
The market portfolio is defined as a value-weighted index of all available stocks in the Center for Research in Security Prices (CRSP) database and the risk-free rate is defined as the one-month US T-bill rate. αi estimates are obtained using Ordinary
Least Squares (OLS) regressions and resemble the return beyond what the factor models predict.
I use the Gibbens, Ross, and Shanken (1989) test (GRS-test) for testing the joint hypothesis that all mutual fund alphas are equal to zero (H0 : ai = 0 for all i). The
10 and residuals for each individual mutual fund. The test statistic is estimated in the following manner
FGRS= [N(T-L - 1)T(T-N-L) ] (
̂p'∑̂-1̂p
1+ θ̂p2 ) ~ F(N,T - N - L) (5)
where T is the number of time-series observations on returns; N is number of test assets (N < T); L is the number of factors; ̂p is a vector consisting of (̂1, ̂2,…, ̂N);
∑̂-1 is the inverse covariance matrix of the residual returns; and θ ̂ =(r̅m-r̅f )
sm is the Sharpe
ratio, defined as dividing r̅m-r̅f , the average excess market return by Sm, the standard
deviation of excess market return. The GRS-test has a central F-distribution with T-N-L degrees of freedom.
After analyzing the performance of SRI and CI mutual funds, I continue with the aggregated time series analysis of mutual fund flows. The analysis follows the method of Chevalier and Ellison (1997) and Sirri and Tufano (1998). As Sirri and Tufano (1998) note that individual investors mostly take raw performance into account, fund flow is defined in the following manner
fi,t =
TNAi,t - (1+Ri,t)TNAi,t-1
TNAi,t (6)
where TNAi,t is the total net assets of fund i at time t; and Ri,t is the return of fund i at
time t. When analyzing time series data, fi,t is displayed as the flow percentage over
the total net assets of the fund in t-1. As with the performance analysis under Eq. (1), I aggregate both the mutual fund return and the total net assets to create four equally weighted portfolios.
I start out with modelling the fund flow as a function of its past returns and fund flow and control for AUM and risk. This leads to the following equation
fi,t = αi+ β ln(TNAi,t-1) + β Riski,[t-1,t-12] + β Flowi, t-1 + β Flowi, [t-1,t-3] + β Returni,t+ β Returni,t-1+ β Returni,t-2 + β Returni,t-3 + εit (7)
where fi,t is the net flow of portfolio i in period t as defined in (8); ln(TNAi,t-1) is an added
control variable as a dollar flow has larger impact on smaller mutual funds; (Risk i,[t-1,t-12) is the risk measure defined as the standard deviation over the past twelve months;
Flowi, t-1 is the fund flow of portfolio i at time t-1; Flowi, [t-1, t-3] is the three month average
fund flow of portfolio i; Returni,t to Returni,t-3 are the returns of portfolio i from month t
11 Seeing that previous literature found evidence indicating a stronger fund flow for better performing mutual funds compared to poorer ones, fund flow is estimated according to performance ranks. All funds are ranked between zero and one according to monthly raw performance, where zero is the poorest performing mutual fund and one is the best performing mutual fund. Three groups are formed, leading to the following equation
fi,t = αi+ β ln(TNAi,t-1) + β Riski,[t-1,t-12] + β Flowi, t-1 + β Flowi, [t-1,t-3] + β LOWPERFi,t+
β MIDPERFi,t+ β HIGHPERFi,t + εit (8)
where LOWPERFi,t is the return of the bottom ranked quartile performing mutual funds;
MIDPERFi,t is defined as the two middle ranked quartile performing mutual funds; and
HIGHPERFi,t is defined as the top ranked quartile performing mutual funds, capturing
the sensitivity of fund flows to the previous monthly average performance.
4. Data
This paper uses the Thomson Reuters (TR) Lipper Database to extract mutual fund performance and characteristics and the TR ESG Database to extract company ESG ratings. I analyze a sample of 63 Dutch and 172 German mutual funds and include the entire population as opposed to a matched-pair analysis, since the quality of the matched-pair analysis worsens over time (Kreander et al., 2005). This selection is the outcome of a process starting with the full sample of mutual funds available in Lipper and subsequently filtering on AUM’s below $ 10,000,000, monthly returns over the period 1990 to 2018, monthly availability of TNA and the presence of an TR ESG rating. One limitation of this paper is that the data set is prone to survivorship bias. Thomson Reuters does not provide historical ESG ratings meaning that mutual funds that have either merged or stopped existing cannot be given an ESG score.
I use two different classifications whether or not a mutual fund engages in SRI. The first classification is based on a self-assessment made by the mutual fund. A mutual fund is labeled as SRI if either the fund name or fund fact sheet suggests SRI. If not, the mutual fund is included in the conventional portfolio.
12 Classification Industry benchmark, as Microsoft’s greenhouse gas emission cannot be compared to that of Royal Dutch Shell. These screens lead to a quantifiable percentile ranked score for over 7,000 international firms. Mutual fund ESG scores are constructed in the following manner
ESGj = ∑ ωi,j Nj
i=1
× ESGi (9)
where ωi,j is the capitalization weight of stock i in fund j; Nj is the number of stocks held
by fund j; and ESGi is the ESG score of stock i. I calculate a coverage ratio, defined
as the AUM in a mutual fund with an ESG score. Overall, high coverage ratios are observed. However, there are some outliers due to large holdings in small capitalization stocks. Although the TR ESG Database covers over 7,000 stocks worldwide, some remain unscreened. As a result, mutual funds with a coverage ratio below 80% have been omitted. I maintain higher coverage ratios than previous literature (Cremers and Petajisto (2009) maintain a ratio of 67% and Borgers et al. (2015) maintain a ratio of 75%) to increase the mutual fund ESG score accuracy. Coverage ratios between 80% and 100% are mostly due to cash positions or government bond holdings.
All funds are ranked between zero and one according to their ESG score and I define the top quartile ranked mutual funds as socially responsible (SRIRated). For the
first analysis I maintain inclusion screening, resulting in a conventional portfolio of 176 mutual funds (CONVRated).
Additionally, I replicate the paper of Diltz (1995) and create portfolios based on best- and worst-in-class screening. Calendar-time portfolios are created consisting of the top quartile and decile ranked mutual funds that are compared to the bottom quartile and decile ranked portfolios. The top and bottom quartile ranked portfolios consist of 59 mutual funds and the top and bottom decile ranked portfolios consist of 24 mutual funds.
International factor data to analyze fund risk-adjusted performance is obtained from the Kenneth R. French Data Library1. It offers extensive data on all factors and
the market risk premium. It creates factor portfolios based on all available equity data in the CRSP database. As the mutual fund holdings best resemble international and American stocks, all return and factor data is denoted in US dollars. For consistency, the fund characteristics are also denoted in US dollars.
13
Table I
Summary Statistics mutual fund performance
This table shows the summary statistics of the constructed SRI and conventional portfolios obtained from the Thomson Reuters Lipper database. The data set consists of returns from November 1st, 1990 and continued until August 1st, 2018. The SRI mutual fund subsample
consists of 25 Dutch and 22 German mutual funds, totaling 47 SRI mutual funds. The CI mutual fund subsample consists of 38 Dutch and 150 German mutual funds, totaling 188 mutual funds. The top quartile ranked SRI subsample consists of 18 Dutch and 41 German mutual funds, totaling 59 mutual funds. The conventional subsample consists of 45 Dutch and 131 German mutual funds, totaling 176 mutual funds. The monthly return, standard deviation, and ratios are displayed in percentages. The age is displayed in months and the assets under management are displayed in $millions.
Sample Mean
SRISelf Convself SRIRated ConvRated
Panel A. Returns
Monthly Return 0,35% 0,43% 0,42% 0,43% Monthly Standard Deviation 4,37% 4,50% 4,59% 4,46%
Skewness -0.73 -0.67 -0.66 -0.69 Kurtosis 5.08 4.91 4.81 4.98 Jarque-Bera test 89.43*** 75.79*** 69.63*** 80.86*** Panel B. Characteristics N 47 188 59 176 Age 126 185 168 175
Assets under Management 327.94 540.57 460.62 512.46 Total Expense Ratio 0.94% 1.30% 1.11% 1.27%
ESG Score 74,20% 71,93% 78,02% 70,49%
ESG Coverage Ratio 94,17% 93,12% 95,18% 92,73% 10 Year Sharpe Ratio 19.53% 18.51% 16.82% 19.26% 10 Year Treynor Ratio 7.55% 7.42% 6.67% 7.70% 10 Year Tracking Error 16.77% 16.56% 15.25% 17.03%
14 compared to their conventional peers. In line with Statman (2000) I find a lower expense ratio for the self-assessed SRI sample and note this effect crosses over to the top quartile mutual funds sample. Both SRI classifications score higher on their ESG score than the conventional portfolio and the coverage ratio is high amongst all samples. The self-assessed mutual funds and the conventional mutual funds show an equal Sharpe ratio, Treynor ratio and tracking error, whereas the top quartile ranked mutual funds show lower ratios and a lower tracking error. This is the first sign that the top quartile ranked mutual funds have a lower risk-return ratio, while the self-assessed SRI mutual funds do not seem to compromise.
Table II
Summary statistics mutual fund flow
This table shows the mutual fund flow summary statistics of the constructed SRI and conventional portfolios obtained from the Thomson Reuters Lipper database. The data set consists of net asset values from September 1st, 2008 and continued until August 1st, 2018.
The SRI mutual fund subsample consists of eight Dutch and 11 German mutual funds, totaling 19 SRI mutual funds. The CI mutual fund subsample consists of 15 Dutch and 104 German mutual funds, totaling 119 mutual funds. The top quartile ranked subsample consists of six Dutch and 26 German mutual funds, totaling 32 mutual funds. The conventional subsample consists of 17 Dutch and 89 German mutual funds, totaling 106 mutual funds. The mean flow, standard deviation, ESG score and coverage ratio are displayed in percentages.
Sample Mean
SRISelf ConvSelf SRIrated Convrated
Panel A. Summary Statistics
Monthly Mean Flow 0.008% 0.005% 0.008% 0.005% Monthly St. Dev. 0.045% 0.043% 0.051% 0.042% Skewness 0.19 0.03 0.71 0.06 Kurtosis 3.39 3.68 5.56 3.62 Jarque-Bera test 1.35 2.11 38.52*** 1.79 Panel B. Characteristics N 19 119 32 106 ESG Score 72,05% 72,23% 77,73% 70,57%
ESG Coverage ratio 93,03% 96,50% 92,96% 94,22%
15 monthly flows exceeding -90%< and >1,000% have also been omitted. This leads to the following descriptive statistics.
Compared to Table I, I see a large drop in mutual funds across all subsamples. This is the result of the additional selection process and lacking monthly AUM data in the TR Lipper Database. For both classifications, SRI mutual funds show a higher mean fund flow compared to their conventional counterparts. Only the rated SRI classification rejects normality under the Jarque-Bera test (1980), as its fund flows are more peaked. The self-assessed SRI mutual funds seem to score equal on their TR ESG score compared to their conventional counterparts, as opposed to the higher score in Table I.
5. Empirical Results
In this section I analyze whether investors can increase their performance by investing in SRI mutual funds. I begin by presenting the time-series analysis of the individual mutual funds under the main Carhart four-factor model (1997). Next, I form four aggregated calendar-time portfolios as defined in Eq. (1) and analyze the monthly returns of the self-assessed and rated SRI portfolio. Furthermore, I compare both SRI classifications to the conventional counterparts. Table III presents the results of the Fama and French three-factor model (1993), the Carhart four-factor model (1997) and the Fama and French five-factor model (2015). Table IV continues with the formation of top and bottom quartile and decile ranked portfolios and analyzes the monthly performance under the three multi-factor models. Table V presents the monthly results of the three multi-factor models in consecutive periods of five years. Table VI and VII concludes this section with the monthly results of the mutual fund flow analysis.
16 funds and insignificant results for the remaining 29 mutual funds. Their conventional counterparts in Appendix E present six significantly outperforming mutual funds, 58 significantly underperforming mutual funds and insignificant results for the remaining 112 mutual funds. Of all mutual funds, 3.4% significantly outperform the market, 36.6% significantly underperform the market and the remaining 60% shows insignificant results. Overall, this shows little evidence that active management consistently outperforms the market.
To generalize results and compare both SRI classifications to its conventional counterparts, I form four portfolios. Table III presents the factor estimates for the three-, four- and five-factor model respectively. To adequately estimate the test-statistics and probability values, I test for heteroskedasticity and autocorrelation within the four portfolios under the four-factor model. All portfolios significantly reject the null-hypothesis of no heteroskedasticity and/or no autocorrelation. Robustness testing in section six looks further into the present heteroskedasticity. For now I conclude that the standard errors need to be adjusted and thus I use Newey-West HAC standard errors to calculate test statistics and probability values. Differentials are calculated by subtracting the return of the conventional portfolio from the return of the SRI portfolio and running the same regression.
The individual time-series analysis shows no evidence that mutual funds are able to outperform the market. This crosses over to the portfolio analysis. Panel A in Table III shows the alpha and factor estimates and differentials of the self-assessed SRI portfolio. It shows significantly negative alphas under the three- and five-factor model, while showing an insignificant negative alpha under the four-factor model. Alpha differentials are negative under all factor models, but not significant. As in Gregory, Matatko, and Luther (1997), I find a significant size differential of 0.061% for the self-assessed SRI portfolio. The value factor shows no significant estimates and differential, whereas the momentum factor shows a highly significant momentum differential. This is due to self-assessed SRI mutual funds being more long-term focused. It engages with the firms as opposed to simply dropping them and trailing high performing firms. These results show that self-assessed SRI mutual funds are mostly able to offset the limited diversification possibilities by screening for social responsibility. From the insignificant differentials I conclude that self-assessed SRI mutual funds perform equal to the conventional portfolio and therefore find no evidence to reject the null-hypothesis of equal risk-adjusted performance.
TABLEIII
FACTOR MODELS SINCE 1990
17 observations. The SRISelf portfolio consists of 47 mutual funds and the ConvSelf portfolio consists of 188 mutual funds. The ESGrated portfolio consists of 59 mutual funds and the
Convrated portfolio consists of 176 mutual funds. It shows the coefficients for all factor models per subsample. All returns are displayed in monthly percentages. Alphas and factors are calculated using calendar-time portfolios. Newey-West HAC standard errors are used to obtain probability values due to the presence of heteroskedasticity. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
FF Three-Factor model CH Four-Factor model FF Five-Factor model
Panel A. Self-assessed SRI
SRISelf ConvSelf Differential SRISelf ConvSelf Differential SRISelf ConvSelf Differential
Alpha -0.176* -0.086 -0.090 -0.146 -0.109 -0.037 -0.215** -0.107 -0.108 Mkt – rf 0.992*** 1.023*** -0.035** 0.984*** 1.023*** -0.047*** 0.990*** 1.013*** -0.023 SMB 0.086 0.047 0.036 0.100* 0.039 0.061* 0.104 0.051 0.053* HML -0.005 -0.084 0.078 -0.040 -0.079 0.039 0.084 0.019 0.064 MOM -0.041 0.025 -0.069*** RMW 0.138 0.105 0.034 CMA -0.202** -0.211** 0.009 Adj. R2 0.883 0.898 0.052 0.885 0.898 0.106 0.889 0.902 0.055
Panel B. Ranked SRI Inclusion screening
SRIrated Convrated Differential SRIrated Convrated Differential SRIrated Convrated Differential
Alpha -0.139 -0.078 -0.061* -0.145 -0.100 -0.045 -0.174* -0.098 -0.076* Mkt – rf 1.042*** 1.013*** 0.029*** 1.044*** 1.020*** 0.024*** 1.038*** 1.002*** 0.036*** SMB -0.021 0.070 -0.090*** -0.025 0.063 -0.088*** -0.009 0.073* -0.082*** HML 0.023 -0.109* 0.132*** 0.031 -0.109* 0.140*** 0.130 -0.007 0.137*** MOM 0.008 0.023 -0.015* RMW 0.141* 0.099 0.042 CMA -0.202** -0.214** 0.012 Adj. R2 0.879 0.904 0.284 0.879 0.904 0.288 0.884 0.908 0.288
18 Table IV to see whether these results persist when the conventional portfolio is defined as the bottom quartile and decile ranked portfolio.
TABLEIV
FACTOR MODELS SINCE 1990
This table reports the results of the factor models for the top and bottom TR ESG ranked quartiles and deciles. Data is extracted from the Thomson Reuters Lipper Database. The sample period is November 1st, 1990 until August 1st, 2018, leading to 333 monthly
observations. The top and bottom quartile ranked portfolios consist of 59 mutual funds and the top and bottom decile ranked portfolios consist of 24 mutual funds. It shows the coefficients for all factor models per subsample. All returns are displayed in monthly percentages. Alphas and factors are calculated using calendar-time portfolios. Newey-West HAC standard errors are used to obtain probability values due to the presence of heteroskedasticity. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
FF Three-Factor model CH Four-Factor model FF Five-Factor model
Panel A. Top and bottom ranked quartile
Topquartile Bottomquartile Differential Topquartile Bottomquartile Differential Topquartile Bottomquartile Differential
Alpha -0.139 0.042 -0.187*** -0.145 -0.008 -0.137** -0.174* 0.089 -0.263*** Mkt – rf 1.042*** 1.014*** 0.028* 1.044*** 1.029*** 0.015 1.038*** 0.974*** 0.064*** SMB -0.021 0.165* -0.186*** -0.025 0.148* -0.173*** -0.009 0.131 -0.140*** HML 0.023 -0.315*** 0.338*** 0.031 -0.316*** 0.347*** 0.130 -0.194** 0.324*** MOM 0.008 0.052 -0.044** RMW 0.140 -0.055 0.195*** CMA -0.202** -0.257*** 0.055 Adj. R2 0.879 0.911 0.413 0.879 0.9120 0.420 0.882 0.915 0.437
Panel B. Top and bottom ranked decile
Topdecile Bottomdecile Differential Topdecile Bottomdecile Differential Topdecile Bottomdecile Differential
Alpha -0.299*** 0.048 -0.347*** -0.294*** -0.056 -0.238** -0.375*** 0.159 -0.534*** Mkt – rf 0.996*** 1.031*** -0.035 0.995*** 1.064*** -0.069** 1.015*** 0.978*** 0.037 SMB 0.061 0.402*** -0.341*** 0.066 0.371*** -0.305*** 0.098 0.333*** -0.235*** HML 0.090 -0.362*** 0.452*** 0.076 -0.375*** 0.451*** 0.122 -0.329*** 0.451*** MOM -0.008 0.105** -0.113*** RMW 0.200* -0.259** 0.459*** CMA -0.101 -0.142 0.041 Adj. R2 0.841 0.850 0.328 0.841 0.853 0.349 0.845 0.853 0.376
19 outperformance. The differential alphas do exhibit highly significant underperformance under all factor models. The bottom quartile ranked portfolio exhibits a highly significant negative size effect, meaning that the portfolio favors large capitalization stocks. This results in a highly significant differential, as the top ranked portfolio exhibits no significant size effect. The bottom quartile ranked portfolio also shows a highly significant negative value factor, meaning the portfolio favors growth stocks. As with the self-assessed SRI portfolio, I find a significant negative momentum differential, meaning that the top quartile ranked portfolio is more long-term focused.
Panel B shows the top and bottom decile ranked portfolio. While the top ranked portfolio significantly underperforms the market in all factor models, the bottom ranked portfolio fails to show any significance. This results in highly significant negative alpha differentials. Furthermore, the factors exhibit the same results as in Table III Panel B and Panel A of Table IV, only more persistent and more significant. Under the four-factor model we see a negative alpha differential of 0.239%, a negative size effect of 0.306%, a positive value effect of 0.451% and a negative momentum effect of 0.114% which are all highly significant. These results suggest that the top decile ranked mutual funds significantly underperform the bottom ranked mutual funds, meaning we reject the null-hypothesis of equal risk-adjusted performance.
Following Bauer, Koedijk, and, Otten (2005), Table V presents the results in consecutive periods of five years. It shows the alpha estimates under the Carhart four-factor model (1997) over four different periods and provides more insight in the performance differences over time. The periods prior to 1998 are excluded, as the self-assessed mutual fund portfolio drops to merely six mutual funds.
Highly significant negative alpha differentials are observed for all top ranked SRI portfolios during the 1998-2003 period. This period captures the Dot-Com bubble in 2001. This is due to the highly significant negative size factor differential, as literature finds evidence that small capitalization firms significantly outperform the market after financial turmoil (Switzer, 2010). The consecutive two periods show no significant alpha differentials and the self-assessed SRI portfolio manages to outperform its conventional peers in the 2013-2018 period. These results are in line with Bauer, Koedijk, and Otten (2005), who argue that socially responsible mutual funds underwent a catching-up phase due to learning, as it is a relatively new industry.
TABLEV
FACTOR MODELS THROUGH TIME
This table presents the results of the alpha estimates for all SRI and CI portfolios. Data is extracted from the Thomson Reuters Lipper database. It shows the factor alphas per subsample for consecutive periods of five years, starting on August 1st, 1998 and ending on
20 and in percentages. Newey-West HAC standard errors are used to obtain probability values due to the presence of heteroskedasticity. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
1998-2003 2003-2008 2008-2013 2013-2018
Panel A. Four-factor model alphas over time for self-assessed SRI
Carhart Four-Factor alpha estimates
SRISelf -0.440** -0.082 -0.285** -0.211**
ConvSelf -0.117 -0.078 -0.304** -0.281***
Differential -0.323** -0.004 0.019 0.070**
Panel B. Alpha estimates for top quartile and bottom 0.75
Carhart Four-Factor alpha estimates
SRIrated -0.320 -0.083 -0.332* -0.298**
Convrated -0.099 -0.077 -0.293** -0.257***
Differential -0.221** -0.006 -0.039 -0.041
Panel C. Alpha estimates for top and bottom quartile
Carhart Four-Factor alpha estimates
Topquartile -0.320 -0.083 -0.332* -0.298**
Bottomquartile 0.207 -0.101 -0.272*** -0.265
Differential -0.527** 0.018 -0.060 -0.033
Panel D. Alpha estimates for top and bottom decile
Carhart Four-Factor alpha estimates
Topdecile -0.547* -0.121 -0.350** -0.302**
Bottomdecile 0.558 -0.013 -0.376*** 0.291***
Differential -1.105** -0.108 0.026 -0.011
I conclude this section with the time-series analysis of the mutual fund flow. I start by estimating the relation between fund flow and current and lagged performance in Table VI. Table VII estimates the relation between fund flow and lagged ranked performance, according to performance. As the analysis includes a risk factor, defined as the standard deviation of the previous 12 months, the observations drop from 120 monthly observations to 109 monthly observations.
21 confirmed by Nilsson (2008) in his study amongst 500 private investors, who finds that socially responsible investors are also motivated by altruism and the feeling of having a positive impact on society. The top quartile ranked portfolio and its conventional counterpart fail to show a significant flow-performance relation, which is in line with the arguments above. The top ranked quartile mutual funds do not advertise as being socially responsible and thus are not part of the new SRI industry.
Table VI
Aggregated lagged flow-performance analysis
This table presents the results of the mutual fund flow analysis for all SRI and CI portfolios extracted from the Thomson Reuters Lipper database. The data set consists of net asset values and raw performance from September 1st, 2008 continued until August 1st, 2018. It
shows the coefficients per calendar-time portfolio. The mean flow, return and standard deviation are displayed in percentages. Newey-West HAC standard errors are used due to the presence of heteroskedasticity. Test-statistics are presented in parentheses. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
SRISelf ConvSelf SRIrated Convrated
Panel A. Mutual fund flow
Intercept 0.096** (0.252) 0.201 (1.301) 0.065 (0.164) 0.259 (1.405) Log lag TNA -0.005
(-0.273) -0.01 (-1.370) -0.004 (-0.199) -0.133 (-1.449) St. Devt-1, t-12 0.231** (2.544) 0.190** (2.450) 0.311 (1.592) 0.149** (2.447) Flowt-1 -0.053 (-0.486) -0.140** (-2.250) -0.428* (-1.982) -0.079 (-0.952) Flowt-1, t-3 -1.481** (2.307) -0.059 (-0.031) -2.653 (-0.855) 0.343 (0.666) Returnt 0.974*** (34.472) 0.986*** (29.898) 1.101*** (11.201) 0.903*** (46.339) Returnt-1 0.514** (2.258) 0.158 (0.242) 1.320 (1.092) -0.025 (-0.157) Returnt-2 0.482** (2.328) 0.027 (0.043) 0.904 (0.846) -0.091 (-0.585) Returnt-3 0.493** (2.395) 0.024 (0.037) 0.842 (0.820) -0.096 (-0.582) N 19 119 32 106 Adj. R2 0.863 0.951 0.680 0.952 Obs. 109 109 109 109
22 The self-assessed SRI portfolio shows results in line with Bollen (2007). The fund flow of the top performing self-assessed mutual funds is smaller compared to the conventional mutual funds, although I find no significance. The standard deviation is not significant, indicating that investors indeed only take raw performance into account. The top performing SRI mutual funds experience a 0.327% increase in fund flow for every 1% increase in monthly return. The middle performing mutual funds experience a 0.544% increase and the bottom performing mutual funds experience a 0.601% decrease for every 1% flow increase.
Table VII
Aggregated ranked flow-performance analysis
This table presents the results of the mutual fund flow analysis for all SRI and CI portfolios. Data is extracted from the Thomson Reuters Lipper Database. The data set consists of net asset values and raw performance from September 1st, 2008 continued until August 1st, 2018.
It shows the coefficients per calendar-time portfolio. The mean flow, return and standard deviation are displayed in percentages. Newey-West HAC standard errors are used due to the presence of heteroskedasticity. Test-statistics are presented in parentheses. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
SRISelf ConvSelf SRIrated Convrated
Panel A. Mutual fund flow
Intercept 2.145* (1.955) 2.147** (2.126) 1.945* (1.667) 2.205** (2.227) Log lag TNA -0.108**
(-1.989) -0.108** (-2.173) -0.098* (-1.700) -0.111** (-2.275) St. Devt-1, t-12 0.069 (0.249) 0.116 (0.455) 0.272 (0.760) 0.061 (0.241) Flowt-1 -0.399 (-1.102) -0.426 (-1.280) -0.670* (1.731) -0.349 (1.071) Flowt-1, t-3 0.313 (1.334) 0.298 (1.241) 0.179 (0.679) 0.336 (1.415) LOWt-1 -0.601 (-0.574) 0.351 (0.382) 0.418 (0.358) 0.155 (0.171) MIDt-1 0.544 (0.282) -1.320 (-0.783) -1.006 (0.507) -1.078 (-0.649) HIGHt-1 0.327 (0.330) 1.272 (1.412) 1.149 (1.202) 1.146 (1.269) N 19 119 32 106 Adj. R2 0.106 0.129 0.095 0.129 Obs. 109 109 109 109
23 and Ellison (1997) further investigate this convex flow-performance relation and argue that this is due to the call option mutual fund managers experience, as their fees depend on performance. This introduces agency costs, which incentivizes them to take on more risk to achieve the highest return and thus management fees. It is logical that the self-assessed SRI portfolio is less prone to this effect, as their investment strategy is more limited than conventional mutual funds. Additionally, socially responsible management companies should eliminate the agency costs and maintain the same ESG criteria for themselves as they do for the stocks that they pick.
6. Robustness Tests
In order to test the robustness of the risk-adjusted performance analysis, I leave the aggregated calendar-time portfolio and introduce the GRS-test, which collectively tests whether the alphas are equal to zero. The main restriction of this test is T>N with matched observations amongst all mutual funds, which is not feasible under the initial risk-adjusted performance analysis since November 1st, 1990. This leads to a new 15
year monthly data set.
Table VIII presents the results of the GRS-test. All factor alphas are significantly negative, although the self-assessed SRI mutual funds show higher alpha estimates compared to the conventional mutual funds. This is due to this data set not capturing the 1998-2003 period, which shows the largest underperformance. The rated SRI mutual funds show lower alpha estimates compared to its conventional portfolio. All factor estimates are in line with the results presented earlier.
The higher alpha estimates for the self-assessed SRI mutual funds leads to a F-statistic of 1.486, meaning it does not reject the null-hypothesis of ai = 0 under the
Carhart four-factor model (1997). The conventional mutual funds do reject the null-hypothesis with a F-statistic of 3.011 at the 1% significance level. The rated SRI portfolio and its conventional portfolio both reject the null-hypothesis at the 1% significance level with F-statistics of 2.398 and 3.025 respectively.
TABLEVIII GRSTEST
This table contains the results of the factor models and the GRS-test for all SRI and CI mutual funds. Data is extracted from the TR Lipper database. Panel A consists of 14 self-assessed mutual funds and 112 conventional funds. Panel B consists of 31 top quartile ranked mutual funds and 91 conventional mutual funds. It shows the coefficients for all factor models per subsample. All returns are monthly and in percentages. Calculations are based on data since July 1st, 2003 continued until August 1st, 2018, totaling 180 monthly observations. Newey-West
24 FF Three-Factor CH Four-Factor FF Five-Factor
Panel A. Self-assessed SRI
SRISelf ConvSelf SRISelf ConvSelf SRISelf ConvSelf
Alpha -0.156** -0.197*** -0.155** -0.204*** -0.147** -0.189*** Mkt – rf 1.083*** 1.096*** 1.083*** 1.099*** 1.074*** 1.087*** SMB -0.062 -0.066 -0.069 -0.077 -0.075 -0.079 HML -0.085** -0.135*** -0.076 -0.117** -0.044 -0.092 MOM 0.003 0.019 RMW 0.014 -0.020 CMA -0.071 -0.069 Adj. R2 0.964 0.965 0.964 0.965 0.965 0.965 FGRS 1.676 2.944*** 1.486 3.011*** 1.880* 3.076***
Panel B. Rated SRI
SRIrated Convrated SRIrated Convrated SRIrated Convrated
Alpha -0.244*** -0.186*** -0.226*** -0.199** -0.236*** -0.179*** Mkt – rf 1.146*** 1.085*** 1.141*** 1.090*** 1.141*** 1.076*** SMB -0.174*** -0.035 -0.185*** -0.045 -0.187*** -0.048 HML -0.016 -0.157*** -0.008 -0.137** 0.022 -0.114* MOM -0.017 0.026* RMW 0.014 -0.020* CMA -0.036 -0.077 Adj. R2 0.949 0.966 0.949 0.966 0.949 0.966 FGRS 2.493*** 3.005*** 2.398*** 3.025*** 2.725*** 3.027***
Financial time-series data is often prone to volatility clustering. This means that periods of high (low) volatility are often followed by periods of high (low) volatility. I conclude the robustness testing by modelling volatility clustering. The first approach is the autoregressive conditional heteroskedasticity model (ARCH) by Engle (1982). It allows the conditional variance to depend on the squared error term in the previous period, leading to the following equation
σt2 = γ
0 + γ1εt-12 + γ2εt-22 + ... + γ3εt-q2 (10)
25 Under the GARCH model the conditional variance is dependent on both the lagged squared errors and the lagged conditional variance, resulting in the following equation
σt2 = ω + α1εt-12 + β1σt-12 (11)
where ω is a constant term; εt-12 is the squared error term in period t-1 as under Eq. (10); and σt-12 is the variance at time t-1. Table VIII shows the results of the ARCH(1) and GARCH(1,1) model.
TABLEIX
ARCH(1) AND GARCH(1,1)
This table shows the results of the ARCH(1) and GARCH(1,1) model for all mutual Dutch and German funds extracted from the TR Lipper database. It shows the ARCH and GARCH coefficients per calendar-time portfolio. Calculations are based on data since November 1st,
1990 until August 1st, 2018, totaling 333 monthly observations. Test-statistics are presented in
parentheses in panel A and Z-statistics are presented in parentheses in panel B. *, **, and *** show the significance of the variables at the 10%, 5% and 1% level.
SRISelf ConvSelf SRIrated Convrated
Panel A. ARCH(1) ARCH(1) 0.168*** (3.119) 0.246*** (4.630) 0.140** (2.587) 0.253*** (4.771) Chi-Square probability 0.002*** 0.000*** 0.010*** 0.000*** Panel B. GARCH(1,1)
SRISelf ConvSelf SRIrated Convrated
ω (1.457) 0.022 0.022 (1.623) 0.040 (1.751) 0.015 (1.426) α1 (Ut-12) 0.094*** (2.597) 0.066*** (2.699) 0.053** (2.160) 0.071*** (2.911) β1 (σt-12) 0.887*** (22.554) 0.913*** (36.248) 0.922*** (32.002) 0.911*** (38.248)
Table IX presents the results of the ARCH(1) estimate in panel A and the ARCH(1) and GARCH(1,1) estimates in panel B, both under the main Carhart four-factor model (1997). The ARCH(1) model rejects the null-hypothesis of no ARCH effects at the highest significance level, meaning that volatility can, to a certain degree, be predicted by lagged volatility.
The GARCH(1,1) model in panel B also significantly rejects the null-hypothesis of no GARCH effects at the 1% significance level, meaning that volatility can, to a certain degree, be predicted by lagged volatility. The GARCH(1,1) model is sufficient as it follows the ω > 0, α1 ≥ 0 and 0 ≤ β1 < 1 conditions. As α1 + β1 ≈ 1 , the volatility
26 highly significant presence of persistent volatility clustering. This should be taken into account when interpreting the results.
7. Limitations
Multiple problems arise when dealing with this data set. The first is survivorship bias. Due to the inability to determine whether funds went bankrupt or merged with other funds, this paper might show an inflated effect. Many studies have been done on the effect survivorship bias on performance measurement and find that it leads to significantly overestimated alphas (see, e.g., Grinblatt and Titman, 1989; Brown, Goetzmann, and Ibbotson 1992; Elton, Gruber, and Blake, 1996). This paper uses a data set that is not corrected for survivorship bias which should be considered when interpreting the results.
Another issue is that I assume that the screening factors have remained constant over the years. This paper differentiates by creating rated socially responsible portfolio, based on TR ESG scores. However, ESG scores are extracted in August 2018 while raw performance data is extracted since November 1st, 1990. TR
continuously screens stocks, leading to time varying ESG. While publicly traded stocks are not able to become socially responsible overnight, there might be differences since 1990. However, historical mutual fund ESG scores are not available within the TR ESG database.
A common issue with mutual fund performance analysis is putting higher weight on older mutual funds. As SRI is quite new, there are more observations in recent years (48 mutual funds) as compared to back in 1990 (three mutual funds). I controlled for this by analyzing risk-adjusted performance through time in Table V and analyzing a 15 year monthly data set in Table VIII.
A final issue is the ex post calculation of the beta. Fama and Macbeth (1973) were the first to conclude that the beta is not constant, but changes overtime. I controlled for this by including alpha estimates of consecutive five year periods.
8. Conclusions
This paper investigates the performance of Dutch and German domiciled self-assessed SRI mutual funds compared to conventional mutual funds. I extend this by ranking all mutual funds according to their TR ESG score and deeming the top quartile and decile socially responsible.
27 The calendar-time portfolios present similar results. The self-assessed mutual fund portfolio underperforms the market under the three- and five-factor model, while the four-factor model shows no significant results. Its conventional portfolio shows no significant risk-adjusted returns among any of the factor models. The differentials show no significance and thus I conclude that self-assessed SRI mutual funds perform equal to conventional mutual funds.
The risk-adjusted return of the rated SRI portfolios are analyzed in two different ways. First, the top quartile ranked portfolio is compared to the conventional portfolio based on the inclusion method, that deems the remaining 75% as conventional. This results in a significant negative alpha differential under the three- and five-factor model at the 10% level. The second method analyzes the risk-adjusted returns according to the best- and worst-in-class method. It provides alpha estimates of the top quartile and decile ranked mutual funds and compares it to the bottom quartile and decile ranked mutual funds. The results show that the alpha differentials become more significantly negative as the difference between social responsible and conventional grows larger. The fund flow analysis shows results in line with theory. As a result of the socially responsible label, it shows a positive significant lagged performance-flow relation. The relation is more pronounced as socially responsible investors take both financial and non-pecuniary motivations into account. It is not surprising that the top ranked mutual funds exhibit the same relation as the conventional portfolios, as these mutual funds do not advertise itself as socially responsible. This is a result of investor irrationality. Investors need to assess whether mutual funds actually are socially responsible, instead of taking it at face value.
29 Reference list
Andreoni, J., 1990. Impure altruism and donations to public goods: a theory of warm-glow giving. The Economic Journal 100, 464-477.
Aupperle, K. E., Carroll, A. B., Hatfield, J. D., 1985. An empirical examination of the relationship between corporate social responsibility and profitability. Academy of Management Journal 28, 446-463.
Ball, R., 2009. The global financial crisis and the efficient market hypothesis: what have we learned? Journal of Applied Corporate Finance 21, 8-16.
Barnett, M. L., Salomon, R. M., 2006. Beyond dichotomy: the curvilinear relationship between social responsibility and financial performance. Strategic
Management Journal 27, 1101-1122.
Bauer, R., Derwall, J., Otten, R., 2007. The ethical mutual fund performance debate: new evidence from Canada. Journal of Business Ethics 70, 111-124.
Bauer, R., Koedijk, K., Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. Journal of Banking & Finance 29, 1751-1767.
Benson, K. L., Humphrey, J. E., 2008. Socially responsible investment funds:
investor reaction to current and past returns. Journal of Banking & Finance 32, 1850-1859.
Black, F., 1986. Noise. The Journal of Finance 41, 528-543. Bollerslev, T., 1986. Generalized autoregressive conditional
heteroskedasticity. Journal of Econometrics 31, 307-327.
Borgers, A., Derwall, J., Koedijk, K., ter Horst, J., 2015. Do social factors influence investment behavior and performance? Evidence from mutual fund
holdings. Journal of Banking & Finance 60, 112-126.
Bollen, N. P., 2007. Mutual fund attributes and investor behavior. Journal of Financial and Quantitative Analysis 42, 683-708.
Breusch, T. S., 1978. Testing for autocorrelation in dynamic linear models. Australian Economic Papers 17, 334-355.
Brown, S. J., Goetzmann, W., Ibbotson, R. G., Ross, S. A., 1992. Survivorship bias in performance studies. The Review of Financial Studies 5, 553-580.
Capelle‐Blancard, G., Monjon, S., 2014. The performance of socially responsible funds: does the screening process matter? European Financial Management 20, 494-520.
Carhart, M., 1997. On persistence in mutual fund performance. The Journal of Finance 52, 57-82.
30 Chevalier, J., Ellison, G., 1997. Risk taking by mutual funds as a response to
incentives. Journal of Political Economy 105, 1167-1200.
Clifford, C., Fulkerson, J., Jordan, B., Waldman, S., 2013. Risk and fund flows. Unpublished working paper. University of Kentucky, Lexington.
Cohen, M. A., Fenn, S., Naimon, J. S., 1995. Environmental and financial performance: are they related? Unpublished working paper. Vanderbilt University, Nashville.
Cremers, K. M., Petajisto, A., 2009. How active is your fund manager? A new measure that predicts performance. The Review of Financial Studies 22, 3329-3365.
Diltz, D.J., 1995. The private cost of socially responsible investing. Applied Financial Economics 5, 69-77.
Derwall, J., Guenster, N., Bauer, R., Koedijk, K., 2005. The eco-efficiency premium puzzle. Financial Analysts Journal 61, 51-63.
Derwall, J., Koedijk, K., Ter Horst, J., 2011. A tale of values-driven and profit-seeking social investors. Journal of Banking & Finance 35, 2137-2147.
Dowell, G., Hart, S., Yeung, B., 2000. Do corporate global environmental standards create or destroy market value? Management Science 46, 1059-1074. Elton, E. J., Gruber, M. J., Blake, C. R., 1996. Survivor bias and mutual fund
performance. The Review of Financial Studies 9, 1097-1120.
Engle, R. F., 1982. Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica: Journal of the Econometric Society 50, 987-1007.
Fama, E.F., French K., 1993. Common risk factors in the returns on stocks and bonds. The Journal of Financial Economics 33, 3-56.
Fama, E. F., French, K. R., 2015. A five-factor asset pricing model. Journal of Financial Economics 116, 1-22.
Fama, E. F., MacBeth, J. D., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political Economy 81, 607-636.
Fisher, I., 1930. The Theory of Interest. The Macmillan Company, New York.
Frazzini, A., 2006. The disposition effect and underreaction to news. The Journal of Finance 61, 2017-2046.
Freeman, R. E., 1984. Strategic management: a stakeholder approach. Cambridge University Press, Cambridge.
Friedman, M., 1970. The social responsibility of business is to increase its profits. New York Times Magazine 13, 32-33.
31 Godfrey, L. G., 1978. Testing against general autoregressive and moving average
error models when the regressors include lagged dependent
variables. Econometrica: Journal of the Econometric Society 48, 1293-1301. Gregory, A., Matatko, J., Luther, R., 1997. Ethical unit trust financial performance:
small company effects and fund size effects. Journal of Business Finance & Accounting 24, 705-725.
Griffin, J. J., Mahon, J. F., 1997. The corporate social performance and corporate financial performance debate: twenty-five years of incomparable
research. Business & Society 36, 5-31.
Grinblatt, M., Titman, S., 1989. Mutual fund performance: an analysis of quarterly portfolio holdings. The Journal of Business 62, 393-416.
Gruber, M. J., 1996. Another puzzle: the growth in actively managed mutual funds. The Journal of Finance 51, 783-810.
Guenster, N., Derwall, J., Bauer, R., Koedijk, K., 2010. The economic value of corporate eco-efficiency. European Financial Management 17, 1-26.
Guerard, J.B., 1997. Additional evidence on the cost of being socially responsible in investing. The Journal of Investing 6, 31–36.
Hamilton, S., Jo, H., Statman, M., 1993. Doing well while doing good? The investment performance of socially responsible mutual funds. Financial Analysts Journal 49, 62-66.
Hirshleifer, J., 1988. Price Theory and Applications. Prentice Hall, Englewood Cliffs, New Jersey.
Hong, H., Kacperczyk, M., 2009. The price of sin: the effects of social norms on markets. Journal of Financial Economics 93, 15-36.
Humphrey, J. E., Tan, D. T., 2014. Does it really hurt to be responsible? Journal of Business Ethics 122, 375-386.
Ippolito, R. A., 1992. Consumer reaction to measures of poor quality: evidence from the mutual fund industry. The Journal of Law and Economics 35, 45-70. Jaffe, J.F., 1974. Special information and insider trading. Journal of Business 47,
410-428.
Jarque, C. M., Bera, A. K., 1980. Efficient tests for normality, homoscedasticity and serial independence of regression residuals. Economics Letters 6, 255-259. Jensen, M. C., 1968. The performance of mutual funds in the period 1945–1964. The
Journal of Finance 23, 389-416.
Jensen, M.C., 1978. Some anomalous evidence regarding market efficiency. Journal of Financial Economics 6, 95-101.
32 Klassen, R. D., McLaughlin, C. P., 1996. The impact of environmental management
on firm performance. Management Science 42, 1199-1214.
Knoll, M.S.(2002). Ethical screening in modern financial markets: the conflicting claims underlying socially responsible investment. Business Lawyer 57, 681-726.
Kreander, N., Gray, R. H., Power, D. M., Sinclair, C. D., 2005. Evaluating the
performance of ethical and non‐ethical funds: a matched pair analysis. Journal of Business Finance & Accounting 32, 1465-1493.
Lantos, G. P., 2002. The ethicality of altruistic corporate social responsibility. Journal of Consumer Marketing 19, 205-232.
Lee, D. D., Humphrey, J. E., Benson, K. L., Ahn, J. Y., 2010. Socially responsible investment fund performance: the impact of screening intensity. Accounting & Finance 50, 351-370.
Malkiel, B. G., 1995. Returns from investing in equity mutual funds 1971 to 1991. The Journal of Finance 50, 549-572.
Malkiel, B. G., Fama, E. F., 1970. Efficient capital markets: a review of theory and empirical work. The Journal of Finance 25, 383-417.
Mandelker, G., 1974. Risk and Return: the Case of Merging Firms. Journal of Financial Economics 1, 303-335.
Markowitz, H., (1952). Portfolio selection. The Journal of Finance, 7(1), 77-91. Merton, R.C., 1987. A simple model of capital market equilibrium with incomplete
information. Journal of Finance 42, 483–510.
Moskowitz, M., 1972. Choosing socially responsible stocks. Business and Society Review 1, 71-75.
Nilsson, J., 2008. Investment with a conscience: examining the impact of pro-social attitudes and perceived financial performance on socially responsible
investment behavior. Journal of Business Ethics 83, 307-325.
Renneboog, L., Ter Horst, J., Zhang, C., 2008. Socially responsible investments: institutional aspects, performance, and investor behavior. Journal of Banking & Finance 32, 1723-1742.
Revelli, C., Viviani, J. L., 2015. Financial performance of socially responsible
investing (SRI): what have we learned? A meta‐analysis. Business Ethics: A European Review 24, 158-185.
Sauer, D. A., 1997. The impact of social‐responsibility screens on investment performance: evidence from the Domini 400 social index and Domini Equity Mutual Fund. Review of Financial Economics 6, 137-149.
