Journal of Business Ethics (2007) 70:111–124 Ó Springer 2006

DOI 10.1007/s10551-006-9099-0

Rob Bauer
The Ethical Mutual Fund Performance Jeroen Derwall
Debate: New Evidence from Canada Rogér Otten

ABSTRACT. Although the academic interest in ethical investigate the aggregated performance and investment
mutual fund performance has developed steadily, the style of ethical and conventional mutual funds and allow
evidence to date is mainly sample-specific. To tackle this for time variation in the funds’ systematic risk.
critique, new research should extend to unexplored Our Canadian evidence supports the conjecture that
countries. Using this as a motivation, we examine the any performance differential between ethical mutual
performance and risk sensitivities of Canadian ethical funds and their conventional peers is statistically
mutual funds vis-à-vis their conventional peers. In order insignificant.
to overcome the methodological deficiencies most prior
papers suffered from, we use performance measurement KEY WORDS: ethical mutual funds, mutual fund
approaches in the spirit of Carhart (1997, Journal of performance, performance measurement, socially
Finance 52(1): 57–82) and Ferson and Schadt (1996, responsible investing (SRI), business ethics
Journal of Finance 51(2): 425–461). In doing so, we
JEL CLASSIFICATION: G12, G20, G23, M14
Rob Bauer is a Professor of Finance (chair: Institutional
Investors) at Maastricht University in the Netherlands, an
advisor to the board of ABP Investments, and co-founder of
the European Centre for Corporate Engagement (ECCE). In Introduction
addition, he is a senior researcher of Netspar and board
member of the International Centre for Pension Management The number of mutual funds investing in com-
at the Rotman Business School in Toronto. His present panies that meet a variety of ethical criteria, better
research focus is on socially responsible investments, corporate known as ethical mutual funds or socially
governance, asset liability management, risk budgeting, and
responsible mutual funds, is growing rapidly
stock selection. His SRI research has been awarded the
Moskowitz Prize (in 2002 and 2005) and the 2005
worldwide. Although the principles of socially
European Finance & Sustainability Research Award. responsible investing (SRI) have been known for
Jeroen Derwall is an assistant professor of Finance at RSM many decades, the need for ethical screening of
Erasmus University and at Maastricht University, and corporate behavior has become exceptionally
co-manager of the European Centre for Corporate Engage- fashionable since the recent reports of some seri-
ment (ECCE). He has research experience in the areas of ous corporate environmental and accounting
socially responsible investing, corporate social responsibility, scandals. In fact, the strongly growing interest in
equity and fixed-income portfolio management, and mutual the incorporation of social, moral, environmental
funds. His PhD research focused on the economic consequences or any other ethical criteria into the stock selec-
of corporate social responsibility and SRI policies. His work tion process may eventually affect corporate
on socially responsible investing has been published in widely behavior as companies may become out of favour
read international journals, and has been awarded the 2005
within the investment community when behaving
Moskowitz Prize and the 2005 European Finance &
Sustainability Research Award.
unethically.
Rogér Otten is an Assistant Professor of Finance at Maastricht For quite some time, however, it has been argued
University in the Netherlands. His research is primarily in the that imposing ethical constraints on the equity
areas socially responsible investing (SRI), corporate govern- investment process will come at the cost of inferior
ance and mutual funds. His work on SRI funds has been portfolio performance. Several oft-cited theories lie
awarded the 2002 Moskowitz Prize. at the core of this prediction. First, because an ethical
112 Rob Bauer et al.

investment opportunity set is by construction a risk-adjusted returns by comparing fund performance

subset of the entire security investment universe, to a broad Canadian equity benchmark, the TSE300.
ethical investors forgo some benefits to diversifica- Their approach builds on a 1-factor framework, in
tion. Second, developing ethical investment screens which returns on actively managed funds can be
and corporate social responsibility rankings can be an explained by only one passive market index, where a
expensive practice that may ultimately factor into net fund’s sensitivity to that index measures the fund’s
return. Third, it has been suggested that irresponsible systematic risk, plus a return component independent
activities are more lucrative and recession-proof than of risk that captures the contribution of the fund
responsible investments.1 Supporters of these views manager. This framework has been under strenuous
therefore claim that ethical mutual funds are likely to attacks in recent literatures, which suggest that single-
underperform their conventional peers. index specifications cannot fully explain the return-
In spite of these plausible opinions, prior related risk characteristics of equity investment portfolios.
literature has not been able to find a substantial per- For example, the single-index model does not
formance gap between ethical and mainstream port- account for risk associated with non-index holdings,
folios. Initially, ethical fund studies primarily focused such as small cap stocks. In addition, several cele-
on traditional performance methods, such as the brated articles by Fama and French (1992, 1993,
Sharpe or Jensen measure, to compare ethical mutual 1996) call for a risk factor that explains the anoma-
funds to either conventional mutual funds or an eth- lously large return spread between stocks with high
ical or standard index. Along these lines, the majority book-price ratios (‘‘value’’ stocks) and those with
of research provided empirical evidence from the low book-price ratios (‘‘growth’’ stocks). Fama and
United States, United Kingdom and Australia.2 French (1993) propose a multifactor model that in-
However, while the growing body of studies may cludes (i) a value-weighted market proxy, (ii) a factor
seem impressive at first glance, the evidence to date capturing the risk premium associated with small-
could be sample-specific. To tackle this critique, the versus-large company stocks, and (iii) a variable that
analysis should be carried over to other countries. accounts for the ‘‘value’’-versus-’’growth’’ stock
The objective of this study is twofold. First, with premium. Carhart (1997) augments the three-factor
the intention to provide significant complementary model by a momentum strategy that buys last year’s
evidence on ethical mutual fund performance, we return winners and sells short past losers. We argue
examine the Canadian ethical fund market. Al- that not using a multifactor model to evaluate ethical
though being the second-largest retail market in the funds can lead to an erroneous assessment of mutual
world (Social Investment Organization, 2000), the fund performance. Without multifactor models, we
Canadian ethical mutual fund industry has attracted cannot separate returns associated with social
little attention in the literature. To the best of our investment policies from the returns on common
knowledge, Asmundson and Foerster (2001) is the investment styles that do not incorporate those pol-
only published study that explored this area. They icies. Bauer et al. (2005) recently re-examined the
examined the performance of Canadian ethical performance of ethical mutual funds in the United
mutual funds with domestic equity orientation vis-à- States, the United Kingdom and Germany using
vis the performance of the TSE 300 Index and found multifactor models that control for size, book-to-
no statistically meaningful differences in return but market and momentum bias in equity portfolios.
some weak evidence suggesting that ethical investing Their evidence not only suggests that ethical funds
is less risky. However, their results are somewhat perform similar to conventional mutual funds, but
sensitive to their assumption of what represents an also indicate that multiple style factors are important
average conventional mutual fund, since they do not in explaining ethical mutual fund performance.
compare ethical mutual funds with conventional Therefore, we follow Bauer et al. (2005, hereafter
mutual funds directly. BKO) and evaluate ethical mutual fund performance
The second purpose of the paper is to address and investment style relative to conventional funds
potential benchmark problems when evaluating the using Carhart’s (1997) 4-factor model.
relative performance of ethical mutual funds in The remainder of this paper is organized as fol-
Canada. Asmundson and Foerster (2001) calculate lows: the next section gives an overview of the
 The Ethical Mutual Fund Performance Debate 113

Canadian retail market for ethical investments and an probably not significantly affect the performance
outline of the data set. In Section ‘‘Empirical results’’ statistics on ethical funds because, as far as we know,
we discuss the empirical analysis and the results. no domestic ethical fund disappeared during the
Finally, Section ‘‘Concluding remarks’’ provides a sample period. On the other hand, the absence of
summary and conclusion. dead funds in the sample is likely to bias conventional
mutual fund performance in this study upwards.5
Whenever necessary, the potential effects of survi-
Data
vorship bias will be taken into consideration.
Table I reports summary statistics of an equally
General market overview
weighted portfolio of all ethical mutual funds in the
sample and those of an equally weighted conven-
In December 2000 the Social Investment Organiza-
tional fund portfolio. In addition, we present
tion (SIO), Canada’s leading non-profit organization
annualized performance data on the S&P/TSX
dedicated to the progression of socially responsible
Composite Index. As a first indication of risk-ad-
investing, released its first results of an extensive
justed performance, we present the Sharpe ratio.
survey on the condition of ethical investing in
This ratio divides the average excess fund return
Canada. Key findings of this report showed that the
over the sample period by the standard deviation of
total amount of assets in ethical investments in
returns over that period. The numerator is the return
Canada equalled approximately $50 billion at the end
earned by the fund relative to the return on an
of the previous millennium, about 3% of the industry
investment in a risk-free asset. The denominator
as a whole. Moreover, according to SIO estimates the
penalizes the achieved excess return for associated
SRI-retail market grew at a higher than average rate
volatility. Therefore, the ratio measures the return to
of 75%, from $5.9 billion in 1998 to $10.35 billion in
risk trade-off, where a higher Sharpe ratio is better.
2000. Although the total number of Canadian ethical
Over the entire sample period, the average ethical
equity mutual funds is still rather low relative to other
fund earned a lower average annualized return than
countries, the estimates suggest that the Canadian
its conventional counterpart: 5.12% versus 5.48%.
SRI-retail industry is the second largest in the world
Corresponding standard deviations (14.21% and
(Social Investment Organization, 2000).3
14.05%) suggest that ethical funds were also more
risky. Unsurprisingly, the Sharpe ratios suggest that
Mutual fund data conventional funds outperformed ethical funds on a
return-to-risk basis. The average return on the S&P/
Our mutual fund sample consists of Canadian ethical TSX Composite was higher for this particular sample
and conventional mutual funds with domestic equity period (6.40%) compared to the average ethical and
orientation only. Hence, we exclude foreign, bal- conventional fund return, but the return variability
anced and guaranteed funds. We furthermore ignore was substantially higher as well (17.49%). The cor-
funds less than 12-months of age. The resulting data responding Sharpe ratio suggests that the index
set comprises adjusted Net Asset Values on mutual outperformed ethical funds as well as conventional
funds that focus on long-term capital appreciation by funds. Furthermore, a comparison of the expense
investing primarily in Canadian stocks, although a ratios presented in the table reveals that ethical
small fraction may occasionally be invested in bonds, mutual funds charge some additional compensation.
cash positions or non-domestic securities. All data are
obtained from Globefund.com. Monthly logarithmic
returns are calculated using funds’ net asset values Factor benchmarks
adjusted for distributions.4 All fund returns are net of
expenses. The regression approaches we employ to assess fund
Unfortunately, our sample does not include data performance required us to collect various data. In
on disappearing mutual funds. We expect the impact order to estimate multifactor models in the spirit of
of survivor bias on the empirical results to be rather Carhart (1997), we construct factor portfolios using all
asymmetrical. That is, survivor bias in the data set will Canadian stocks in the Worldscope database. The
114 Rob Bauer et al.

TABLE I
Summary statistics on Canadian mutual funds

Fund Return St.Dev. Sharpe Exp.ratio #Funds

Ethical portfolio 5.12% 14.21% 0.05 2.71% 8

Conventional portfolio 5.48% 14.05% 0.08 2.41% 267
S&P/TSX composite 6.40% 17.49% 0.12

This table reports summary statistics on ethical and conventional mutual funds in the sample and on the S&P/TSX
Composite Index. Ethical and conventional fund returns are calculated based on an equally weighted portfolio of all funds.
Mean return, corresponding standard deviation and Sharpe ratio are presented on an annualized basis. The Sharpe ratio is
defined as the ratio of the excess return on the fund portfolio or index to the standard deviation of return.Sample period:
1994:01–2003:01.

main benefit of the Worldscope stock universe is that Traditionally, performance has been measured
it covers over 98% of total market capitalization. To using unconditional expected returns under the
calculate the first determinant in the multifactor assumption that investors make no use of informa-
model, the excess market return, we deduct the tion about the state of the economy to form
monthly T-Bill rate from the monthly return on the expectations and to engage in dynamic trading
value-weighted market proxy. We use the Canadian strategies. However, if money managers trade on
30-day T-bill rate as a measure of the risk-free rate. publicly available information, unconditional models
End-of-month T-Bill rates are from the Bank of that assume a time-invariant beta portray a biased
Canada. The second determinant is the return spread picture of fund performance. For this reason, Ferson
between a small cap portfolio and a large cap portfolio. and Schadt (1996) advocate conditional performance
The small cap portfolio covers the bottom 20% of total measurement.
market capitalization after ranking all stock according Our study follows Ferson and Schadt (1996) and
to size. The remaining part is assigned to the large cap uses a conditional performance evaluation approach
portfolio. The third regressor is the difference in re- that allows for time variation in funds’ betas. For this
turn between a value stock portfolio and a growth purpose, we constructed various Canadian economic
stock portfolio. Following Fama and French (1993), information variables. The variables are: (a) the
we first rank all stocks according to their book-to- 3-month Treasury-Bill rate, (b) the term spread, cal-
market ratio and then assign the top 30% of market culated as the long-term government bond rate
capitalization to the high book-to-market (i.e. value) minus the T-Bill rate, (c) the quality spread, defined
portfolio and the bottom 30% to the low book-to- as the yield spread between the long-term corporate
market (i.e. growth) portfolio. Finally, we consider a bond and the long government bond, and (d) the
momentum factor which is defined as the monthly dividend yield of the S&P/TSX Composite Index.
return spread between a past 12-month winner
portfolio and a past 12-month loser portfolio. After
having ranked all stocks on their prior 12-month re- Empirical results
turn, we classify the top 30% as ‘‘winners’’ and the
bottom 30% as ‘‘losers’’. All factor portfolios are re- Jensen’s alpha
balanced annually at the end of each year.
We additionally consider the explanatory power The most widely employed benchmark model in
of a leading Canadian ethical equity index, launched mutual fund performance studies is the CAPM-
recently. The Jantzi Social Index is a value-weighted based single-factor model. In this scenario, a fund’s
index consisting of 60 companies that pass a variety outperformance, known as Jensen’s alpha, is mea-
of social and environmental screens.6 Monthly sured as the difference between the return on the
returns on the index are provided by Michael Jantzi mutual fund and the return on the single-factor
Research Associates Inc. benchmark according to an estimated CAPM.
 The Ethical Mutual Fund Performance Debate 115

Hence, when computing the Jensen measure, it is Regression results show that both ethical mutual
implicitly assumed that the single-index model is funds and their conventional peers underperform the
sufficiently capable of explaining the cross-section of value-weighted market proxy. The observed alphas
stock returns. () 2.93% and ) 2.56%, respectively) are statistically
In this section, we discuss the results of applying significant at the 10% level. The performance results
the single-factor regression on our data to estimate for the ‘difference’ portfolio show that neither alpha
the Jensen measure. Formally, we estimate the fol- nor beta is statically significant. Hence, based on these
lowing 1-factor model for both the ethical fund outcomes we cannot reject the notion that no statis-
portfolio and its non-ethical counterpart: tically significant difference exists in the performance
of ethical funds and their conventional peers.7
Rit
Rft ¼ ai þ bi ðRm
Rft Þ þ eit ; ð1Þ Since the investment universe of ethical mutual
fund managers is determined by ethical screens, the
where, Rit is the return on mutual fund i in month
single-factor regression of excess ethical fund
t, Rft the risk-free rate at t measured by the 30-day
returns on a standard equity index may lead to biased
T-Bill rate, Rm the return on the market proxy in
estimates of mutual fund performance. Therefore, it
month t, e it an error term.
is useful to consider a relevant ethical equity index to
Consequently, bi (beta) measures the market risk measure the performance of ethical mutual funds.
exposure of the fund and ai represents Jensen’s alpha. We repeat our computations but now using the
The market proxy is measured by a value-weighted return on the Jantzi Social Index as the determinant
market portfolio, constructed using the Worldscope in the single-factor model. Although the Jantzi
database. Social Index was introduced only recently, its per-
Empirical results are presented in Table II. For formance has been back-tested for the period
each coefficient, (we report the corresponding 1995–1999. Consequently, we are only able to
t-statistic) derived from Newey and West (1987) estimate Jensen’s alpha using monthly ethical index
standard errors. As the primary focus of our research returns for the period January 1995–January 2003.
is the performance and style differential between For comparison purposes, we also run a similar
ethical mutual funds and their conventional coun- single-index regression using 8-year data on the
terparts, we also investigate the returns on a ‘differ- Worldscope market proxy.
ence’ portfolio, which is constructed by subtracting Regression results are provided in Table III. With
the conventional mutual funds returns from the regard to the ethical fund portfolio, the estimated
return on the ethical mutual fund portfolio. alpha is negative and not statistically significant at the

TABLE II
Empirical results 1-factor regressions

Fund a b Adj. R2

Ethical portfolio )2.93()1.74)* 0.84(35.56)  0.91

Conventional portfolio )2.56()1.67)* 0.84()23.31)  0.92
Difference )0.37()0.31) 0.00()0.07) )0.01

This table reports the results from CAPM-based regressions. To measure ethical and conventional mutual fund perfor-
mance, we estimated the model formally defined by Equation (1):Rit
Rft ¼ ai þ bi ðRm
Rft Þ þ eit ; ð1Þ where
Rit ) Rft denotes the local return on a fund portfolio in excess of the risk-free rate and Rm ) Rft is the excess return on
the market portfolio constructed using the Worldscope database. The ‘‘difference’’ portfolio is constructed by subtracting
conventional mutual fund returns from the returns on the ethical mutual fund portfolio. T-statistics (in brack-
ets) are derived from Newey–West heteroskedasticity and autocorrelation consistent standard errors. Sample period:
1994:01–2003:01.
*Coefficient is statistically significant at 10% level.
 Coefficient is statistically significant at 1% level.
116 Rob Bauer et al.

TABLE III
Empirical results 1-factor regression using ethical equity index

Fund Worldscope market proxy Jantzi social index

a b Adj. R2 a b Adj. R2

Ethical portfolio )2.44()1.35) 0.84(32.59)  0.90 )0.58()0.24) 0.68(10.31)  0.74

Conventional portfolio )2.18()1.27) 0.84(21.55)  0.92 )0.30()0.11) 0.67(7.82)  0.75
Difference )0.27()0.21) 0.00(0.02) )0.01 )0.28()0.23) 0.01(0.20) )0.01

This table reports the results from CAPM-based regressions to estimate mutual fund performance using a conventional
value-weighted market proxy based on the Worldscope stock universe, and the Jantzi Social Index. Columns 2, 3 and 4
display results from a single-factor regression of excess aggregate fund returns on the excess return on the value-weighted
market proxy. Columns 5, 6 and 7 display results from similar single-factor regressions using the Jantzi social index serving
as an alternative market proxy. The ‘‘difference’’ portfolio is constructed by subtracting conventional mutual fund returns
from the returns on the ethical mutual fund portfolio. All alphas are annualized. T-statistics (in brackets) are derived from
Newey–West heteroskedasticity and autocorrelation consistent standard errors. Sample period: 1995:01–2003:01.
 Coefficient is statistically significant at 1% level.

usual cut-off levels. The results therefore point out if funds are considerably involved in style investment
that ethical mutual funds are not able to outperform strategies (e.g. small caps or value strategies), their
their ethical index. Striking is that the adjusted R2 returns cannot be fully explained by the single-factor
from the model with the ethical index is lower model. We therefore expect to shed more light on
(0.74) than the R2 from the standard single-index ethical fund performance and investment behaviour
model (0.90), indicating that the ethical index is less throughout the remainder of this paper by means of
capable of explaining ethical mutual fund perfor- enhanced performance analyses in a multivariate
mance than a standard equity index. Equally setting.
remarkable are the fund betas, which reveal that the
ethical fund portfolio is more exposed to a stan-
dard market index (b = 0.84) than to the ethical Multifactor models
index (b = 0.68). It is questionable whether these
findings can be fully explained by the fact that the It has been repeatedly argued that the 1-factor asset
ethical index comprises only 60 stocks, whereas the pricing model is insufficiently able to explain the
conventional index represents a much larger stock cross-section of expected stock returns. Fama and
universe. The results are in line with BKO (2005), French (1993) demonstrated that the CAPM is
who examined the explanatory power of domestic inefficient and introduced a 3-factor model that
ethical indexes in the United Kingdom and the includes the factors SMB and HML in addition to
United States. Provided the ethical index sufficiently the excess market return. In their study SMB
represents an ethical stock portfolio, our findings raise corresponds to the return difference between a small
concern over whether Canadian ethical funds are and a large stock portfolio and HML represents the
truly distinguishing themselves by investing in ethical return difference between a value stock portfolio and
securities only. Similar observations are put forward a growth stock portfolio using the book-to-market
by Haigh and Hazelton (2004) in their investigation ratio as a discriminating factor. Fama and French’s
of Australian ethical mutual funds. They even go as far findings imply that the 3-factor model will be
as suggesting: ‘‘The distinction between SRI funds and incrementally useful in explaining mutual fund
conventional funds may be largely in name’’. returns if fund managers significantly engage in style
Recently the traditional performance measure- investment strategies.
ment routines discussed thus far have been criticized While the benefits of the 3-factor model are
for providing biased estimates of mutual fund per- nowadays acknowledged, the model is subject to
formance. There is substantial evidence showing that further improvement. In response to evidence
 The Ethical Mutual Fund Performance Debate 117

showing that the 3-factor model is insufficiently month t, HMLt the return difference between a
capable of explaining the Jegadeesh and Titman value (high B/M) portfolio and a growth (low B/
(1993)-momentum strategy, Carhart (1997) sug- M) portfolio in month t, MOMt the return differ-
gested the addition of a momentum factor to existing ence between a portfolio of past 12-month win-
models to capture persistence in fund performance. ners and a portfolio of past 12-month losers in
The resulting 4-factor model is expected to provide month t, USt the excess Canadian dollar return on
reliable information on a fund’s relative performance the S&P500 index at t.
and allows us to further estimate to what degree a
Panel A and B in Table IV report the results from
fund engages in various accepted equity investment
the estimation of the 4-factor model and 5-factor
strategies. Recent studies by Berkowitz and Qiu
model respectively. In the multivariate framework
(2001; 3-factor model) and by L’Her et al. (2001;
the adjusted R2 is higher than in the CAPM-envi-
4-factor model) confirm the notion that multifactor
ronment, which indicates that the multifactor models
models are able to explain the cross-sectional varia-
are more capable of explaining aggregate mutual fund
tion in Canadian equity returns.
returns. The alphas presented in both panels provide
In addition to this, there now is evidence con-
evidence of strong and significant underperformance
firming that ethical mutual fund performance is in-
for ethical mutual funds (a = ) 3.18%) as well as for
deed attributable to style tilts, which cannot be
their conventional peers (a = ) 2.90%). The coef-
accounted for in a single-index environment. For
ficients and t-statistics on the factors MKT and SMB
example, Gregory et al. (1997) found that the small
imply that both ethical and conventional mutual
firm effect is significant in explaining U.K. ethical
funds are significantly exposed to the market factor
trust performance. BKO (2005) found evidence
and also significantly biased towards small capitali-
suggesting that ethical mutual funds are less exposed
zation stocks. Loadings on HML and MOM, how-
to the market portfolio compared to conventional
ever, are insignificant. Furthermore, empirical results
funds, but more small cap- and growth stock-ori-
in panel B show that the addition of the foreign
ented. Estimates of a mutual fund’s factor loadings
component does not affect the alpha estimates sub-
and alpha are therefore likely to be more reliable in a
stantially. However, it appears that a small but sta-
multivariate framework.
tistically significant portion of the exposure to MKT
Although most funds in our sample primarily
is now captured by the U.S. factor.
invest in the Canadian equity market, some occa-
The magnitude of the observed underperfor-
sional exposure to foreign markets may be allowed
mance of Canadian mutual funds as a whole seems
up to pre-specified limits. For that reason, we
rather extreme at first glance. Our results are
additionally measure fund performance by means of
nonetheless similar to those reported by Berkowitz
the domestic 4-factor model described earlier aug-
and Qiu (2003), who measured the performance of
mented by a U.S. market index.
mutual funds managed by publicly traded companies
Using the average ethical and conventional mu-
relative to the performance of funds managed by
tual fund data for further analysis, we estimate the
private management companies. A potential source
following equations:
of the strong overall underperformance is the high
Rit
Rft ¼ ai þ b0i ðRm
Rft Þ þ b1i SMBt expense ratio associated with the Canadian mutual
ð2Þ fund industry.8
þ b2i HMLt þ b3i MOMt þ eit ;
The performance evaluation results for the ‘dif-
and ference’ portfolio point out that the performance gap
between ethical and conventional mutual funds is
Rit
Rft ¼ ai þ b0i ðRm
Rft Þ þ b1i SMBt þ statistically insignificant. While the absence of a
b2i HMLt þ b3i MOMt þ b4i USt þ eit performance difference is not surprising and consis-
ð3Þ tent with the results of prior multifactor regression
results, i.e. BKO (2005), our results are puzzling in
where, SMBt is the return difference between a the sense that none of the differences in factor
small cap portfolio and a large cap portfolio in loadings between ethical mutual funds and their
 118

TABLE IV
Multifactor regression results

Fund a Rm ) Rf SMB HML MOM Adj. R2

Panel A:4-factor model estimation results
Ethical portfolio ) 3.18()2.37)** 0.89(41.50)  0.28(6.87)  0.02(1.10) 0.00(0.15) 0.94
   
Conventional portfolio ) 2.90()3.27) 0.89(51.78) 0.29(9.90)  0.03(1.34) 0.01(0.64) 0.97
Difference ) 0.27()0.22) 0.00()0.21) ) 0.02()0.47) ) 0.01()0.39) ) 0.01()0.55) ) 0.02
Panel B:4-factor model + US factor
Fund a Rm ) Rf SMB HML MOM US Adj. R2
     
Ethical Portfolio ) 3.26()2.68)  0.84(30.45) 0.30(7.17 ) 0.01(0.89) 0.01(0.65) 0.07(2.95) 0.95
Conventional portfolio ) 2.94()3.30)  0.86(37.25)  0.31(11.64 ) 0.02(1.18) 0.02(0.91) 0.05(2.02)** 0.97
Difference ) 0.31()0.28) 0.02()0.72) ) 0.01()0.21) ) 0.01()0.47) ) 0.01()0.40) 0.02(0.88) ) 0.03

This table reports empirical results corresponding to the multifactor regressions formulated by Equation (2) and (3) respectively:
Rob Bauer et al.

Rit
Rft ¼ ai þ b0i ðRm
Rft Þ þ b1i SMBt þ b2i HMLt þ b3i MOMt þ eit ; ð2Þ
Rit
Rft ¼ ai þ b0i ðRm
Rft Þ þ b1i SMBt þ b2i HMLt þ b3i MOMt þ b4i USt þ eit ; ð3Þ
Rm ) Rft represents the returns on the market proxy in excess of the risk-free rate, SMB denotes the difference in return between a small cap portfolio and a large cap
portfolio, HML denotes the return spread between a value portfolio and a growth portfolio and MOM is the return difference between a prior 12-month winner
portfolio and a prior 12-month loser portfolio. US indicates the excess return on the S&P500 index. The ‘‘difference’’ portfolio is constructed by subtracting
conventional mutual fund returns from the returns on the ethical mutual fund portfolio. All alphas are annualized. T-statistics (in brackets) are derived from Newey–
West heteroskedasticity and autocorrelation consistent standard errors. We also present the adjusted R2 from each model. Sample period: 1994:01–2003:01.
**Coefficient is statistically significant at 5% level.
 
Coefficient is statistically significant at 1% level.
 The Ethical Mutual Fund Performance Debate 119

conventional peers is significant. Thus, at the HML and MOM existed up to the end of 2000. Thus,
aggregate level, it seems like Canadian ethical and ethical funds were more value oriented than their
conventional mutual funds exhibit virtually identical conventional peers and exhibited a more positive
sensitivities. loading on the momentum factor. As it appears,
This result can be explained along several lines. however, these differences in loading disappear
One potential source of this observation is the so quickly due to a ‘‘shock’’ in the data at the beginning
called ‘‘best-of-sector’’ approach that is currently of 2001. This shock is almost entirely attributable to
very popular in the ethical investment industry. The one company in the data set, namely Nortel.
oldest generation of ethical funds primarily filtered The boom and bust of the Nortel stock is well
out companies that operate in sectors considered known. Through a large number of acquisitions
controversial from an ethical perspective (e.g. alco- during the nineties Nortel attempted to meet
hol, gambling, military equipment, nuclear energy growing Internet demands. Prior to the burst of the
and tobacco). Companies that violate environmental technology bubble the company had grown tre-
standards or certain human rights are usually mendously, making the Nortel stocks jointly rep-
excluded from the investment universe as well. It is resent over 35% of the entire Canadian market.
therefore no surprise that ethical stocks may be However, the collapse of the TMT market initiated
sector-specific; see for example diBartolomeo and a sharp decrease in Nortel’s stock prices, causing an
Kurtz (1999). As a result of this approach, known incredible decline in market value in less than
as negative screening, ethical portfolios may suf- 1 year.
fer from a lack of diversification. To reduce Unsurprisingly, Nortel’s impact on our factor
extreme sector tilts, many ethical investors now portfolios is severe. At the end of June 2000,
adopt so-called positive screening methods to approximately 70% of our large cap portfolio and of
identify firms that are ‘‘best-in-sector’’ improvers of the growth portfolio consisted of Nortel stocks. At
their social or environmental performance. Yet, the next rebalancing date Nortel’s weight in the
critics believe that improper use of best-of-sector large stock and growth portfolio became 22% and
analysis easily results into a portfolio that insuffi- 11%, respectively. The impact on the momentum
ciently represents social awareness. In other words, factor is even more dramatic. At the end of June
if ethical funds support investing in ‘‘the least 2000, Nortel’s shares made up 73% of the winner
controversial’’ companies from each sector rather portfolio’s market value. One year later, the com-
than only in ‘‘good’’ ones, the distinction between pany’s stocks disappeared into the loser segment.
ethical and conventional mutual funds could It would be interesting to investigate the perfor-
become too vague. However, testing such a mance of ethical funds versus conventional funds
hypothesis is beyond the scope of this paper. after adjusting for Nortel’s effects. We thus consid-
A second possibility is that the outcomes for the ered the addition of a ‘‘Nortel factor’’ to our mul-
‘‘difference’’ portfolio are not robust to the choice of tifactor performance models. The results of adding
observation window. In order to investigate this the excess return on Nortel stocks, not reported due
possibility, we perform a sensitivity analysis with to space limitations, were similar to previous out-
respect to the coefficient computations by estimating comes. For all the portfolios under consideration, we
the coefficients in the four-factor model recursively. did not observe a significant loading on the Nortel
Figure 1 displays the time-variation in alpha and factor. Furthermore, the model adj. R2 decreased as
factor loading differences as well as corresponding a result of adding the Nortel factor, suggesting that
95% confidence intervals. this regressor does not have any incremental
Consistent with prior observations, the alpha for explanatory power.
the ‘‘difference’’ portfolio remained insignificant
throughout the entire expanding window analysis.
Contrary to the results in Table IV, which suggested Conditional performance evaluation
the absence of any difference in factor loadings
between ethical and conventional funds, Figure 1 In this section, we extend conventional multifactor
indicates that significant differences in loading on modeling techniques and consider the conditional
120 Rob Bauer et al.

0.004 0.15

0.10
0.002
0.05
0.000
0.00
-0.002
-0.05
-0.004 -0.10

-0.006 -0.15
96 97 98 99 00 01 02 03 96 97 98 99 00 01 02 03
Difference Alpha ± 2 S.E. Coefficient on Rm-Rf ± 2 S.E.

0.4 0.4

0.2 0.3

0.2
0.0
0.1
-0.2
0.0
-0.4 -0.1

-0.6 -0.2
96 97 98 99 00 01 02 03 96 97 98 99 00 01 02 03
Coefficient on SMB ± 2 S.E. Coefficient on HML ± 2 S.E.

0.3

0.2

0.1

0.0

-0.1

-0.2
96 97 98 99 00 01 02 03
Coefficient on MOM ± 2 S.E.

Figure 1. Recursive coefficient estimates: the ‘difference’ portfolio.

modeling approach suggested by Ferson and Schadt Rit
Rft ¼ ai þ bi0 ðRmt
Rft Þ
(1996) and Chen and Knez (1996) to assess ethical ð4Þ
þ B0i Zt
1 ðRmt
Rft Þ þ eit ;
mutual fund performance. Unconditional factor
models may deliver heavily biased performance where, Zt-1 = a vector of lagged information
measures if managers employ dynamic trading strat- variables. While Ferson and Schadt (1996) con-
egies, e.g., using publicly available information about sider a conditional CAPM, their methodology is
the economy. Conditional modeling solves this also applicable in a multivariate framework. In
problem by incorporating publicly available eco- order to obtain the most accurate estimates of al-
nomic instruments into performance estimation to pha, we condition all coefficients in the uncon-
account for the possibility of time variation in betas ditional 4-factor model on a predetermined set
and expected return. of economic information variables. In other
Formally, the conditional single-index model is of words, in addition to including the determinants
the form: in the unconditional model, we also include
 The Ethical Mutual Fund Performance Debate 121

their interactions with a vector of lagged infor- the conditional 4-factor model. Wald test results
mation variables. The end t is a conditional (p-values) are presented in the last column to
4-factor model that allows for time variation in examine the incremental explanatory power of
factor loadings. conditioning betas on a vector of lagged economic
Largely following prior studies in the area of variables.
conditional performance evaluation, we incorporate Wald test results show that in every case the
the following information variables into the condi- hypothesis of time-invariant betas can be rejected at
tional multifactor regression procedure: (a) a lagged the 5% level, confirming the importance of a
T-bill rate, (b) the lagged term spread defined as the framework that considers time variation in factor
yield spread between a long term government bond exposures. Furthermore, conditional performance
and the T-bill rate, (c) the lagged quality spread de- measurement has provided a more optimistic judg-
fined as the yield spread between a 10-year corporate ment of ethical and conventional fund performance
bond and a 10-year government bond and (d) the as the alphas () 2.50% and ) 2.11%) are slightly
lagged dividend yield of the S&P/TSX Composite higher compared to the results from the uncondi-
Index. A large body of literature has discussed the tional approach () 3.18% and ) 2.90%). Nonethe-
economic relevance of these instruments; see for less, the results pertaining to the difference portfolio
example Chen et al. (1986) and Pesaran and suggest that any performance difference between
Timermann (1995). ethical and non-ethical mutual funds is negligible
Regression results are reported in Table V. In from a statistical point of view, thereby confirming
order to compare the results from the unconditional previous outcomes.
model with those from the conditional model, we
import some of the results presented in Table IV.
Columns 2 and 3 display the alphas, t-statistics and Concluding remarks
adjusted R2 values calculated in the previous
section using the unconditional 4-factor model. This study presents new evidence on the perfor-
Columns 4 and 5 show the estimation results from mance and investment style of ethical mutual

TABLE V
Unconditional versus conditional performance evaluation

Unconditional model Conditional model

4-Factor a Adj.R2 4-Factor a Adj. R2 Wald (p-value)

Ethical portfolio )3.18()2.37)** 0.94 )2.50()1.64)* 0.94 0.015

Conventional portfolio )2.90()3.27)  0.97 )2.11()2.04)** 0.97 0.000
Difference )0.27()0.22) )0.03 )0.39()0.33) 0.12 0.000

This table reports the results from unconditional (columns 2 and 3) and conditional (columns 4 and 5) multifactor
regressions. Empirical results correspond to the conditional alternative to the unconditional 4-factor model formulated by
Equation (2). In the conditional modeling framework, we allow the 4 factor exposures to vary over time as a function of
(a) a lagged T-bill rate, (b) the lagged term spread, (c) the lagged quality spread and (d) the lagged dividend yield of the
S&P/TSX Composite Index. The ‘‘difference’’ portfolio is constructed by subtracting conventional mutual fund returns
from the returns on the ethical mutual fund portfolio. All alphas are annualized. T-statistics (in brackets) are derived from
Newey–West heteroskedasticity and autocorrelation consistent standard errors. We also present the adjusted R2 from each
model. Wald test results (p-values) are presented to examine the incremental explanatory power of conditioning betas on a
vector of lagged economic variables. Sample period: 1994:01–2003:01.
*Coefficient is statistically significant at 10% level.
**Coefficient is statistically significant at 5% level.
 
Coefficient is statistically significant at 1% level.
122 Rob Bauer et al.

funds. Studying the Canadian market yields out-of ces. These counterintuitive observations should be
sample evidence for a quickly growing retail of interest to regulators who oversee the transpar-
market. Using a wide range of statistical models, ency and information dissemination of mutual fund
we compare the risk-adjusted performance of products to investors.
Canadian ethical mutual funds relative to their
conventional peers. While most previous work on
ethical mutual fund performance has limited its Acknowledgements
attention to single-factor benchmark models, we
utilize multifactor performance evaluation models. We would like to thank Michael Jantzi for supplying
Multifactor specifications not only improve per- performance data on the Jantzi Social Index. Further
formance measurement but also enable us to data support by Robert Schwob and Style Research
investigate ethical mutual fund investment styles in Co. Ltd. is gratefully acknowledged. We appreciate the
greater detail. valuable comments of an anonymous referee, Brent
Our empirical results are fourfold. First, using a Sutton and Kees Koedijk. All remaining errors are
single-factor model, we find no significant perfor- the sole responsibility of the authors. The views
mance difference between ethical and conventional expressed in this paper are not necessarily shared by
ABP Investments.
mutual funds. Second, we find that a single-factor
model containing a standard market proxy has more
explanatory power than an ethical equity index. This Notes
surprising result casts doubt on the distinctive ethical
component of ethical mutual funds. Third, after 1
See for instance the evidence on investing in so-
applying a multifactor model that controls for returns called ‘‘Vice’’sectors (gambling, tobacco etc.) in Luck
associated with several common investment styles and Tigrani (1994) and Ali and Gold (2002).
2
(i.e. based on size, book-to-market, and stock price Studies for the United States include Hamilton
momentum), we find no significant difference in et al. (1993), Statman (2000) and Goldreyer et al.
performance between ethical mutual funds and their (1999), amongst others. Evidence from the United
conventional peers. Fourth, on average, we find no Kingdom includes Luther et al. (1992), Luther and
evidence that the investment style exposures of Matatko (1994), Mallin et al. (1995) and Gregory et al.
ethical mutual funds differ significantly from those of (1997). Australian evidence has been documented by
Cummings (2000), Tippit (2001) and Bauer et al.
conventional mutual funds.
(2006).
The results of our study largely corroborate 3
SIO’s report, entitled ‘‘Canadian Social Investment
previous research on ethical mutual fund perfor- Review 2000. A comprehensive survey of socially
mance. Investing in ethical mutual funds does not responsible investment in Canada’’, can be downloaded
lead to returns that are significantly different from from the SIO website: http://www.socialinvestment.ca
those delivered by conventional mutual funds. 4
We also obtained an alternative data set from Funda-
Therefore, our study does not support the claim ta Canada Inc, which included unadjusted Net Asset
that imposing ethical constraints leads to weaker Values: although the overall mean fund returns were
investment performance. The investment implica- biased downwards since distributions were not corrected
tions are clear: Canadian investors can allocate their for, results were very similar and would not affect any
money to ethical mutual funds without experi- of the conclusions in this paper.
5
encing a financial penalty vis-à-vis conventional For example, in a study on conventional Canadian
mutual fund performance, Athanassakos et al. (2002)
mutual funds.
find that excluding dead funds from the sample poten-
However, the degree to which ethical mutual
tially biases results on fund performance upwards.
funds distinguish themselves from conventional 6
Information regarding the index construction meth-
funds is not very evident. Although ethical funds odology and the list of companies included can be
claim to screen in several ways, we cannot find found at the official website: http://www.mjra-jsi.com
significant differences in investment styles. More- 7
We additionally repeated our analyses using returns
over, ethical fund returns correlate more with on S&P/TSX Composite Index which yielded simi-
conventional market indices than with ethical indi- lar results.
 The Ethical Mutual Fund Performance Debate 123
8
For more on Canadian mutual fund expense ratios, Ferson, W. E. and R. W. Schadt: 1996, ÔMeasuring Fund
see Ruckman (2003). Strategy and Performance in Changing Economic
ConditionsÕ, Journal of Finance 51(2), 425–461.
Goldreyer, E. F., P. Ahmed and J. D. Diltz: 1999, ÔThe
References Performance of Socially Responsible Mutual Funds:
Incorporating Sociopolitical Information in Portfolio
Ali P. and M. Gold: 2002, ‘Analysing the Cost of Ethical SelectionÕ, Managerial Finance 25(1), 23–36.
Investments’, JASSA 9–14. Gregory, A., J. Matatko and R. Luther: 1997, ÔEthical
Asmundson, P. and S. R. Foerster: 2001, ‘Socially Unit Trust Financial Performance: Small Company
Responsible Investing: Better for Your Soul or Your Effects and Fund Size EffectsÕ, Journal of Business Finance
Bottom Line?’, Canadian Investment Review. & Accounting 24(5), 705–725.
Athanassakos, G., P. Carayannopolous and M. Racine: Haigh, M. and J. Hazelton: 2004, ÔFinancial Markets: A
2002, ‘How Effective is Aggressive Portfolio Man- Tool for Social Responsibility?Õ, Journal of Business
agement? Mutual Fund Performance in Canada, 1985– Ethics 52(1), 59–71.
1996’, Canadian Investment Review (Fall) 39–49. Hamilton, S., H. Jo and M. Statman: 1993, ÔDoing Well
Bauer, R., K. Koedijk and R. Otten: 2005, ÔInternational While Doing Good? The Investment Performance of
Evidence on Ethical Mutual Fund Performance and Socially Responsible Mutual FundsÕ, Financial Analysts
Investment StyleÕ, Journal of Banking and Finance 29, Journal 49(6), 62–66.
1751–1767. Jegadeesh, N. and S. Titman: 1993, ÔReturns to
Bauer, R., R. Otten and A. Tourani Rad: 2006, ÔEthical Buying Winners and Selling Losers: Implications for
Investing in Australia, is there a Financial Penalty?Õ, Stock Market EfficiencyÕ, Journal of Finance 48(1), 65–
Pacific-Basin Finance Journal 14(1), 33–48. 91.
Berkowitz, M. K. and J. Qiu: 2001, ‘Common Risk L’Her, J.-F., T. Masmoudi and J.-M. Suret: 2001,
Factors in Explaining Canadian Equity Returns’, ÔEvidence to support the four-factor pricing model
Working Paper, University of Toronto. from the Canadian stock marketÕ, Journal of Interna-
Berkowitz, M. K. and J. Qiu: 2003, ÔOwnership, Risk tional Financial Markets, Institutions and Money 14(4),
and Performance of Mutual Fund Management 313–328.
CompaniesÕ, Journal of Economics and Business 55(2), Luck, C. and V. Tigrani: 1994, ‘Ethical Investing and the
109–134. Returns to Sinful Industries’, BARRA Newsletter,
Carhart, M. M.: 1997, ÔOn the Persistence in Mutual Spring.
Fund PerformanceÕ, Journal of Finance 52(1), 57–82. Luther, R. G., J. Matatko and D. Corner: 1992, ÔThe
Chen, Z. and P. J. Knez: 1996, ÔPortfolio Performance Investment Performance of UK Ethical Unit TrustsÕ,
Measurement: Theory and ApplicationsÕ, Review of Accounting, Auditing and Accountability Journal 5(4),
Financial Studies 9(2), 511–556. 57–70.
Chen, N., R. Roll and S. A. Ross: 1986, ÔEconomic Luther, R. G. and J. Matatko: 1994, ÔThe Performance
Forces and the Stock MarketÕ, Journal of Business 59(3), of Ethical Unit Trusts: Choosing an Appropriate
383–403. BenchmarkÕ, British Accounting Review 26(1), 77–
Cummings, L.: 2000, ÔThe Financial Performance of 89.
Ethical Investment Trusts: An Australian PerspectiveÕ, Mallin, C. A., B. Saadouni and R. J. Briston: 1995, ÔThe
Journal of Business Ethics 25, 79–92. Financial Performance of Ethical Investment FundsÕ,
diBartolomeo, D. and L. Kurtz: 1999, ‘Managing Risk Journal of Business Finance & Accounting 22(4), 483–
Exposures of Socially Screened Portfolios’, Research 496.
Paper, Northfield Information Services. Newey, W. K. and K. D. West: 1987, ÔA Simple, Positive
Fama, E. and K. R. French: 1992, ÔThe Cross-Section of Semi-Definite Heteroskedatiscity and Autocorrelation
Expected Stock ReturnsÕ, Journal of Finance 47, 427– Consistent Variance Covariance MatrixÕ, Econometrica
465. 55(3), 703–708.
Fama, E. and K. R. French: 1993, ÔCommon Risk Fac- Pesaran, M. and A. Timmermann: 1995, ÔPredictability of
tors in the Returns on Stocks and BondsÕ, Journal of Stock Returns: Robustness and Economic Signifi-
Financial Economics 33, 3–53. canceÕ, Journal of Finance 50(4), 1201–1228.
Fama, E. and K. R. French: 1996, ÔMultifactor Expla- Ruckman, K.: 2003, ÔExpense Ratios of North American
nations of Asset Pricing AnomaliesÕ, Journal of Finance Mutual FundsÕ, Canadian Journal of Economics 36(1),
51, 55–84. 192–223.
124 Rob Bauer et al.

Social Investment Organization: 2000, ‘Canadian Social Jeroen Derwall

Investment Review 2000: A Comprehensive Survey RSM Erasmus University,
of Socially Responsible Investment in Canada’, http:// Burgemeester Oudlaan 50, P.O. Box 1738,
www.socialinvestment.ca. 3000, DR, Rotterdam, The Netherlands
Statman, M.: 2000, ÔSocially Responsible Mutual FundsÕ, E-mail: [email protected]
Financial Analysts Journal 57(5–6), 30–39.
Tippet, J.: 2001, ÔPerformance of Australia’s Ethical
Rob Bauer
FundsÕ, The Australian Economic Review 34(2), 170–178.
ABP Investments,
Rob Bauer and Rogér Otten Heerlen, The Netherlands
Department of Finance, E-mail: [email protected]
Maastricht University,
Rob Bauer and Jeroen Derwall
Maastricht, The Netherlands
European Centre for Corporate Engagement,
E-mail: [email protected]
Maastricht, The Netherlands
E-mail: [email protected]

Jeroen Derwall
Maastricht University,
Maastricht, The Netherlands