The investment behavior of socially responsible individual investors

The investment behavior of socially responsible individual investors

Accepted Manuscript Title: The investment behavior of socially responsible individual investors Author: Nicha Lapanan PII: DOI: Reference: S1062-9769...

661KB Sizes 0 Downloads 5 Views

Accepted Manuscript Title: The investment behavior of socially responsible individual investors Author: Nicha Lapanan PII: DOI: Reference:

S1062-9769(17)30386-1 https://doi.org/doi:10.1016/j.qref.2018.05.014 QUAECO 1151

To appear in:

The

Received date: Revised date: Accepted date:

15-11-2017 7-4-2018 25-5-2018

Quarterly

Review

of

Economics

and

Finance

Please cite this article as: Nicha Lapanan, The investment behavior of socially responsible individual investors, (2018), https://doi.org/10.1016/j.qref.2018.05.014 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The Investment Behavior of Socially Responsible Individual

cr

NICHA LAPANAN∗

ip t

Investors

us

ABSTRACT

Using data on individual investors’ equity mutual fund portfolios from 2003 to 2007, this paper describes the behavior of investors in relation to socially responsible (SR) funds. The results suggest

an

that most SR investors rebalance their portfolios slightly more often than conventional investors do, hold potentially more diversified portfolios, and hold a mixed portfolio combining both conventional

M

and SR funds. Further, while SR investors’ buying decisions are similarly sensitive to the past returns on SR and conventional funds, SR investors’ selling decisions are more sensitive to the past

d

negative returns on SR than on conventional funds, which indicates that they are less likely to sell

te

SR than conventional funds as past negative returns decrease. Moreover, the aggregated flows of SR and conventional funds are similarly sensitive to past returns. However, the aggregated flows of

Ac ce p

SR funds for which the majority of their investors are sticky SR investors are less sensitive to past positive returns. Finally, sticky SR investors’ flows (at the individual level) are more sensitive to the past positive returns and are less sensitive to the past negative returns on SR than conventional funds. Although these results suggest that sticky SR investors do have values-driven motives for holding SR funds, their flows to SR funds are less persistent than their flows to conventional funds, which indicates that these investors are less likely to reinvest in SR than in conventional funds.



Ume˚ a School of Business, Economics and Statistics, Ume˚ a University. Financial support from the Wallander, Browald and Tom Hedelius Foundation is gratefully acknowledged. I would also like to thank Stefan Anchev, J¨ orgen Hellstr¨ om, Lisa Kramer, Lu Liu, Rickard Olsson, Oscar St˚ alnacke and participants at the 2015 European Responsible Investment & Institutions Conference for their helpful comments and suggestions. I am also grateful to ˚ Asa Skillius for providing the Folksam reports on ethical funds in Sweden and to Rickard Olsson for his effort in collecting and preparing the data on mutual funds.

Page 1 of 33

I.

Introduction

Since socially responsible (SR) funds impose restrictions against investing in companies that do not align with their responsibility measures, the existing literature generally agrees that SR investors have preference biases for assets. SR investors’ behavior thus appears to deviate from

ip t

traditional finance views in that non-financial attributes such as companies’ governance or their commitment to society or the environment seem to influence investment decisions, along with risk and return. Consistent with this view, prior studies find that, in comparison with conventional

cr

funds, SR fund flows are less sensitive to past negative returns (see Bollen, 2007; Renneboog, Ter Horst, and Zhang, 2011), suggesting that investors tend to be more loyal toward SR funds than conventional funds.

us

These studies, however, are based on analyses of aggregated fund flow data which may not correctly portray the behavior of SR individual investors because a fund’s clients may consist of both individual and institutional investors.1 Considering that institutional investors face higher

an

regulatory requirements2 and invest larger sums of money, it is probable that prior findings were largely influenced by the behavior of institutional investors. In this study, conversely, I use data from Sweden on individual investors’ complete equity mutual fund portfolios from 2003 to 2007.

M

A distinct characteristic of the data set is that it contains all the mutual fund positions of each individual (i.e., it includes individuals’ positions across different investment accounts). The data also contain detailed information on individuals’ socio-demographic characteristics, including their

d

gender, level of education, marital status, income and net wealth. These characteristics have been

te

found to be associated with the mutual fund trading behaviors of individuals (see, e.g., Barber and Odean, 2001; Bailey, Kumar, and Ng, 2011) and are, therefore, important control variables. The purpose of this paper is to document individual investors’ trading behavior in relation to

Ac ce p

SR equity mutual funds. To my knowledge, the question of how, if at all, individuals manage their portfolios to accommodate their preferences for more responsible investing is still not fully understood. Considering that there are a number of SR funds available on the market and that the evidence regarding the performance of these funds is inconclusive (see, e.g., Aktas, De Bodt, and Cousin, 2011; Kempf and Osthoff, 2007; Bauer, Koedijk, and Otten, 2005; Renneboog, Ter Horst, and Zhang, 2008a; Utz and Wimmer, 2014; Ghoul and Karoui, 2017),3 individuals who have preference biases for responsible investing might trade SR funds the same way they trade conventional funds.4 Moreover, the results from survey responses in Riedl and Smeets (2017) show that, while 1

Bollen (2007) states “Unfortunately, the Center for Research in Security Prices (CRSP) mutual fund database, described next, does not permit direct measurement of the level of institutional versus retail investment” (p. 691). 2 According to Eurosif, institutional investors are the primary driver of the growth in the market for SR assets, followed by the increase in regulatory requirements. Corresponding to this observation, the United Kingdom (UK), Sweden, Italy, and Germany have since 2000 required pension funds to report the extent to which social, environmental, and ethical factors have been incorporated into their investment policies (see, e.g.,Renneboog, Ter Horst, and Zhang (2008b); Bengtsson (2008)). 3 Bauer et al. (2005) and Utz and Wimmer (2014) find no significant difference in performance between SR and conventional funds, Bauer, Derwall, and Otten (2007) and Ghoul and Karoui (2017) find that SR funds underperform conventional funds and Gil-Bazo, Ruiz-Verd´ u, and Santos (2010) find that SR funds outperform conventional funds. 4 Prior studies have established that fund flows respond to past fund returns and that the relationship tends to

2 Page 2 of 33

48.7% of SR investors expect lower returns on SR funds, only 16.5% of them expect SR funds to provide higher returns. Interestingly, their results also show that investors (both conventional and SR) who expect lower returns tend to avoid SR investing. On the other hand, investors who expect higher returns are not more likely to hold SR funds. It thus remains an open question whether individuals indeed behave differently towards SR and conventional funds.

ip t

To gain a better understanding of individual investors and SR investing, it is useful to analyze their decisions to buy and sell SR funds, as well as the size of their trades in such funds. Although buying activities result in positive changes in fund flows, it is still important to study the economic

cr

significance of SR individual investors’ fund flows, particularly if investors tend to think of SR investing as a part of their mental accounts. According to Mackenzie and Lewis (1999), investors

us

may choose to allocate a smaller portion of their portfolios to SR assets in order to alleviate their guilty consciences because they may find the processing of information related to funds’ SR attributes to be costly. Moreover, the theoretical model of Levitt and List (2007) and the

an

empirical findings of Døskeland and Pedersen (2015) suggest that financial concerns increase in relative importance to moral concerns as the value of investments increases. In this regard, SR investors may behave more favorably towards SR funds, but only when their investments are small.

M

It is therefore important to investigate the behavior of SR and conventional fund flows both at the fund level and at the investor level.

I begin by describing the characteristics of individual investors’ equity mutual fund portfolios. I show that SR investors, defined as those who have a position in at least one SR fund in any given

d

year during the sample period,5 rebalance their portfolios slightly more often than conventional

te

investors do and they also hold potentially more diversified portfolios (i.e., portfolios with a larger number of funds). Furthermore, consistent with Riedl and Smeets (2017), the majority of SR

Ac ce p

investors hold a mixed portfolio combining both conventional and SR funds. Because of this, it is possible to study the trading behavior of the same individual towards SR and conventional funds. Moreover, the data also make it possible to study such behavior over time, which means that I can control for individuals’ unobservable time-invariant characteristics. A linear probability model (LPM) is then used to separately analyze the buying and selling decisions of individual investors in relation to SR and conventional funds’ past market-adjusted returns.6 For the buying decisions, I find that SR investors are similarly sensitive to the past be convex (Sirri and Tufano, 1998; Chevalier and Ellison, 1997): higher inflows to past positive returns and lower outflows to past negative returns. Ivkovi´c and Weisbenner (2009), to my knowledge, is the only study on investors’ mutual fund purchase and redeem decisions that uses data on individual investors’ mutual fund positions. This paper finds that fund inflows are related to relative performance (i.e., performance in comparison to other funds with the same objective) and fund outflows are related to absolute performance. 5 Prior studies usually define SR investors as those who hold at least one SR fund in a given period (McLachlan and Gardner, 2004; Riedl and Smeets, 2017). These studies, however, examine the behavior of SR investors in the cross-section, and thus the same investor might be an SR investor in one period and a conventional investor in the next. Unlike prior studies, my paper focuses on the behavior of individuals over time; therefore, in any given year, an investor is either an SR or a conventional investor. 6 It is worth noting that the study covers the period 2003-2007 when stock prices around the world were mostly increasing; therefore, it seems reasonable to use market-adjusted returns instead of raw returns. Otherwise, there are very few funds with negative returns during the period.

3 Page 3 of 33

positive and negative returns on SR and conventional funds. However, it seems that past returns are probably a poor determinant of investors’ selling decisions because the relationship between these decisions and the past returns of conventional funds is statistically insignificant when controlling for individual-specific fixed effects. Nonetheless, in the sample consisting of only SR investors, the results show that these investors’ selling decisions are more sensitive to the past negative returns on

ip t

SR than on conventional funds, which indicates that they are less likely to sell SR than conventional funds as past negative returns decrease.

Next, I investigate the relationship between aggregated fund flows (calculated using the flows

cr

of the investors in the sample) and the past performance of conventional and SR funds. The results on aggregated flows show that the flows of SR and conventional funds are similarly sensitive

us

to past positive and negative returns.7 Considering that individuals may invest in SR funds for profit-driven motives, values-driven motives, or both (see, e.g., Derwall, Koedijk, and Ter Horst, 2011; Døskeland and Pedersen, 2015), I further classify SR investors into subgroups of sticky and

an

non-sticky investors. Sticky SR investors are those who have fund positions in more than three years and at least one (not necessarily the same) SR fund in all of those years. I consider these investors to be a proxy for investors who are more likely to have values-driven motives towards SR

M

funds because, according to Fama and French (2007), individuals’ tastes for assets tend to persist over time and are likely to translate into consistent investment decisions. Interestingly, analyzing SR funds for which the majority of their investors are sticky SR investors

d

reveals that these funds’ aggregated flows are less sensitive to past positive returns than are those of conventional funds. While this can be interpreted as sticky SR investors caring less about returns

te

and potentially having values-driven motives toward SR funds (see, e.g., Benson and Humphrey, 2008; Renneboog et al., 2011), it is worth noting that the flows of SR funds with a majority of

Ac ce p

such investors and the flows of conventional funds are similarly sensitive to past negative returns. Moreover, the flows of such SR funds seem to experience smaller increases relative to those of conventional funds.

The analyses of aggregated fund flows could be biased by the activities of wealthier individuals because these individuals are likely to invest larger amounts of money. Therefore, I further investigate the flow-performance relationship at the individual investor level. Here, I find that the fund flows of sticky SR investors are more sensitive to the past positive returns and are less sensitive to the past negative returns on SR than on conventional funds. These observations are consistent with Bollen (2007) and indicate that these investors derive additional utility from the funds’ SR attributes, especially when SR funds perform as well as conventional funds. Interestingly, the results also show that sticky SR investors’ flows to SR funds are less persistent than their flows to conventional funds (i.e., larger flows to SR funds in the last period lead to lower flows in such funds in the current period). This result suggests that sticky SR investors are less likely to reinvest in 7

These results are consistent with those in Renneboog et al. (2011) on the relationship between fund flows and past returns of European funds. Renneboog et al. (2011) find that European SR funds’ inflows are more sensitive to past positive returns; however, the statistical significance of this result depends on the method used (i.e., matched samples versus regression analysis).

4 Page 4 of 33

SR than in conventional funds. The findings in this paper contribute to both theory and practice. The retail mutual fund market represents a sizable portion of the total assets under management (AUM). From 2003 to 2007, individual investors’ direct holdings of mutual funds are around 39% of the total AUM in the Swedish mutual fund market SIFA (2017). In my sample, individuals’ holdings in SR equity

ip t

mutual funds represent almost 4% of the total holdings in equity mutual funds. SR mutual funds thus account for a substantial portion of the total AUM. Fund managers can potentially benefit from a better understanding of the behavior of SR individuals. For example, Cooper, Gulen, and

cr

Rau (2005) find that fund managers can increase fund inflows by changing the name of the fund to reflect a “hot” current style. Moreover, Bollen (2007) finds that SR fund flows are less sensitive to

us

past negative returns and he argues that lower flow volatility makes it easier for fund managers to manage their funds.

Further, the results documented here suggest that SR investing may reflect prosocial biases in

an

investment decisions. Fehr and Fischbacher (2003) point out that a majority of selfish individuals can be forced to cooperate in an altruistic act by a minority of altruists through, for example, strong reciprocity (i.e, rewarding those who abide by the norms and punishing those who do not). In the

M

case of SR investing, Sparkes and Cowton (2004) argue that it becomes a mainstream investment strategy through the enforcement of legislation and an increase in pressure from actual and future beneficiaries. They also argue that SR investing plays a crucial role in influencing companies to

d

address issues related to their corporate social responsibility (CSR). This prosocial bias, moreover, is found to be transmitted over generations (Fehr and Fischbacher, 2003), something that also

te

seems to be true for investment behavior (Hellstr¨om, Lapanan, and Olsson, 2015). As such, asset management firms might be pressured to invest responsibly should demand for such investment

Ac ce p

strategies increase. Meanwhile, mutual fund investors might shun asset management firms that do not offer SR funds. In addition, the potential increase in the number of prospective SR investors would likely affect demand for sin stocks and their expected returns; Hong and Kacperczyk (2009) find that sin stocks have higher expected returns than comparable stocks because they are less held by norm-constrained institutions.

Finally, this study also contributes to the emerging literature on socially responsible investing among individual investors (see, e.g., Bollen, 2007; Berry and Junkus, 2013; Renneboog et al., 2011; Riedl and Smeets, 2017). It documents a new aspect of SR investors behavior, which might be consistent with behavioral portfolio theory (Shefrin and Statman, 2000). Investors in the study appear to choose SR funds as parts of larger fund portfolios which also include conventional funds and appear to behave differently toward SR and conventional funds. Consequently, how investors perceive their investments in SR funds may have an effect on fund flow volatility; for example, SR investors may be willing to accept lower returns on SR funds and they may be less likely to redeem their investments when such funds yield poor returns. The rest of this paper proceeds as follows: Section II describes the data, Section III describes the methodology and presents the results from the empirical analyses, and Section IV provides a

5 Page 5 of 33

conclusion.

II.

Data

The primary data set for this study contains information at the end of each year from 2003 to

ip t

2007 on individual investors who reside in Sweden and who were born in 1963 or 1973. The data on these individuals’ complete equity mutual fund portfolios and their socio-demographic characteristics are obtained from Statistics Sweden. Since 1990, Statistics Sweden has been collecting

cr

data on all individuals who are 16 years of age and older at the end of each December. The information recorded for each individual includes detailed information such as gender, marital status,

us

education, place of residence and income. Before the abolition of the wealth tax in 2008, highly detailed information on the value of individuals’ disaggregated wealth, including their cash in bank accounts, complete stock and mutual fund portfolios, real estate and their portfolios of derivatives,

an

was also collected. Moreover, the data contain individuals’ separate positions in certain securities (i.e., the number of shares for a stock position or the number of units for a position in a fund or a derivative), where the securities are identified by their International Security Identification

M

Numbers (ISINs). Due to the highly detailed information on all individuals, similar data, but for different samples and periods, have been used extensively in the area of household finance; some examples are Campbell (2006), Calvet, Campbell, and Sodini (2007), and Calvet, Campbell, and

d

Sodini (2009).

The data on the characteristics of mutual funds are from the Swedish Investment Fund Associa-

te

tion (SIFA). These data include information on mutual fund names, types of mutual funds (equity, fixed income, balanced, or others) and daily net asset values (NAVs). Although SIFA does not

Ac ce p

provide information on all mutual funds that are held by the individuals in the sample,8 over 3,400 funds were registered with this organization as of 2010. To check the quality of the data, I manually calculate the value of individuals’ complete mutual fund portfolios (using the number of units held in each fund and the funds’ NAVs) and I compare these values to the reported value of individuals’ mutual fund wealth from Statistics Sweden. This calculation shows that these values match for more than 95% of the individuals in the sample. I use individuals’ portfolios of only equity mutual funds in this paper, however, for two main reasons. First, most SR funds are equity funds. Second, investors might have different investment horizons with respect to fixed income and balanced funds, as these fund categories tend to be less risky (Dahlquist, Martinez, and S¨oderlind, 2016). Since my primary interest is to investigate differences in individual investors’ trading behavior towards conventional and SR funds, limiting the study to include only equity mutual funds allows me to compare behaviors towards the same asset class. To classify funds into conventional and SR funds, I first manually screen their names for the following terms: sustainable, ethical, socially responsible, SRI, social and green. This method, however, suffers from look-back bias. To minimize this problem, I also check the reports on SR 8

Some individuals hold mutual funds that are marketed outside Sweden or mutual funds that are not registered with SIFA.

6 Page 6 of 33

funds compiled by Folksam (a Swedish insurance company); these reports cover all Swedish ethical mutual funds operating from 2002 to 2008. In total, 218 of the available 3,400 mutual funds are classified as SR funds. Most of these funds (i.e., 183) employ negative screening during the period when selecting stocks; that is, they do not invest in sin stocks. Of the 218 SR funds, 75 are held by the individuals in the sample, who also hold 687 unique conventional funds.

ip t

Overall, the sample consists of 101,912 individuals who invested in mutual funds during the sample period. While about 65% of the investors have fund positions in all five years, the distribution of investors with such positions in one, two, three or four years is almost equal. 8,984

cr

investors invested in at least one SR fund in any given year during the sample period. Of these, 3,958 investors have fund positions in three years or more and have at least one (not necessarily

us

the same) SR fund in all of those years; these investors are classified as sticky SR investors. All other SR investors are classified as non-sticky SR investors. It may be intuitive to think that sticky SR investors are less likely to sell their SR funds, but it should be kept in mind that a sticky SR

an

investor (as defined in this paper) may switch from one SR fund to another during the sample period. Moreover, note that investors who retain a position in SR funds, but partially reduce its

III.

M

size, are still classified as sticky SR investors.

Methodology and Empirical Results

d

In this section, I begin by examining SR and conventional individual investors’ trading activities. Specifically, I examine their decisions to buy and sell SR and conventional mutual funds, for the

te

full sample of individuals and for the subsample that includes only SR investors. I also examine how sticky and non-sticky SR investors affect fund flows, using aggregated flows at the fund level

A.

Ac ce p

and flows at the individual investor level.

Trading Behavior of SR vs. Conventional Investors Most prior studies on investors in mutual funds use net or gross fund flows to examine investors’

decisions to buy and sell mutual funds (see, e.g., Chevalier and Ellison, 1997; Sirri and Tufano, 1998; Keswani and Stolin, 2008; Barber, Huang, and Odean, 2016). To my knowledge, little is known about mutual fund investors’ heterogeneity in terms of their mutual fund portfolio characteristics and their trading behavior, such as the frequency of their trades (see, e.g., Ivkovi´c and Weisbenner, 2009; Bailey et al., 2011; Dahlquist et al., 2016, for studies on mutual funds’ individual investors). Hence, I begin by describing SR and conventional investors’ trading activities and their portfolio characteristics. The results are presented in Table I and Table II. In this section, the focus is on Panels A, B and C in each of these tables (Panels D and E will be discussed in section III.B). [Table I about here] Panel C in Table I shows that SR investors, on average, hold a larger number of funds each period, with an average of 4 funds for SR investors in comparison with an average of 2.37 funds 7 Page 7 of 33

for conventional investors (Panel B). Furthermore, 24.17% of SR investors hold only one fund, which is almost two times lower than the corresponding figure of 45.96% among conventional investors. These results suggest that SR investors hold potentially more diversified portfolios than conventional investors. Moreover, on average, SR investors allocate about 41% of their equity mutual fund portfolios to SR funds. Investors’ portfolio weights on SR funds are, however, not

ip t

normally distributed, as about 24.1% of SR investors hold only one fund (i.e., 100% portfolio weight), suggesting that the remaining majority of them have portfolio weights on SR funds of less than 41%. SR investors holding a larger number of funds and, on average, a lower percentage of

cr

SR relative to conventional funds seems to be in line with the view that SR investors allocate a smaller portion of their portfolios to SR funds and that they likely consider these funds as a part

us

of their mental accounts (Mackenzie and Lewis, 1999). However, the average fund wealth among SR investors is almost twice as high as that among conventional investors (i.e., about SEK 96,621 versus SEK 57,726). According to prior studies, the higher average fund wealth and the higher

an

number of funds held among SR investors may indicate a higher level of investor sophistication (see, e.g., Alexander, Jones, and Nigro, 1997; O’neal, 1997; Wilcox, 2003). Therefore, the portfolio characteristics of SR investors discussed here may not be an accurate indicator of their social

M

preferences towards SR funds simply because, for example, these investors may be able to pick SR funds that perform better than conventional funds. Moreover, Riedl and Smeets (2017) find that the number of transactions that investors make during a year increases the probability of holding

te

chance.

d

SR funds. It is therefore possible that more diversified investors have holdings in SR funds by mere

[Table II about here]

Ac ce p

Table II reports the trading activities of individual investors. If an investor’s end-of-year position (i.e., the number of units held in a fund) is the same as her position in the prior year, then I consider her as having no trading activity during the year. If at the end of a given year, there is an increase (a decrease) in the number of units in an individual’s fund position, then the individual is considered to have engaged in buying (selling) activities.9 Concerning the frequency of individuals’ trades, Dahlquist et al. (2016) find that active investors (i.e., those who invest in funds other than the default funds in the Swedish Premium Pension System (PPS) and those who made at least one fund change in their pension accounts over the period from 2000 to 2010), on average make about 0.85 changes in their portfolios per year. Individuals in this study hold mutual funds in their private accounts, and thus their behavior should be more in line with that of active traders. As a result, the annual observations of individuals’ portfolios should not represent a major limitation in this paper.10 On average, 32.35% of individuals have no trading activities during the year, while 9 Some mutual funds reinvest their dividends and investors may therefore appear to have more fund units after dividend reinvestments. Since I am trying to study buying decisions, I do not consider increases in the number of fund units of less than 5% as buy decisions. 10 Dahlquist et al. (2016) state that the Swedish PPS functions like a national 401(k) plan, receiving an overall contribution of 18.5% of the gross income of individuals. Of this, 16% is paid to the national defined contribution

8 Page 8 of 33

55.87% engage in buying and 23.47% engage in selling. In relation to conventional investors, SR investors are more engaged in buying (63.15% versus 55.15%) and selling (28.45% versus 22.98%). It seems that individuals tend to buy and hold funds because there are considerably more buy than sell transactions. Moreover, the positive within-individual correlation (i.e., intraclass correlation) between the buying and selling activities is strongest among conventional investors (0.51) and

ip t

weakest among non-sticky SR investors (0.23), suggesting that conventional investors are most likely to be engaged in buying and selling funds within the same year, while non-sticky SR investors are least likely.11

cr

Tables I and II show that SR investors’ portfolio characteristics and trading activities are suitable for studying the dynamics of individuals’ behavior towards conventional and SR funds.

us

Indeed, given that the prior literature (e.g., Bollen, 2007; Benson and Humphrey, 2008; Renneboog et al., 2011) has found that SR fund flows are less sensitive to past performance, one may infer that SR investors are less active. However, the descriptive statistics reveal that the behavior of SR

an

investors is, on average, similar to that of conventional investors, which suggests that SR investors do rebalance their portfolios and trade as often as conventional investors, if not more. The results in Table II indicate that the following analyses, which rely largely on comparing individuals’ propensity

M

to buy and sell SR versus conventional funds, are less likely to suffer from selection problems that may arise if, for example, most SR investors are not active traders. That is, if the results show that SR investors are less sensitive to past returns, my analyses are less likely to be driven by their

d

passive behavior.

To test whether the social responsibility of mutual funds is related to individuals’ propensity to

te

buy and sell funds, I estimate linear probability models (LPMs) in which the dependent variable is an indicator variable for buying and selling decisions, respectively. The advantage of LPMs over

Ac ce p

non-linear models, such as logit or probit models, is that they allow for controlling for individualspecific fixed effects12 and for a straightforward interpretation of interaction terms (see Ai and Norton, 2003, for a discussion of interaction terms in non-linear models). Moreover, the use of categorical variables offers additional benefits. First, it is possible that the amount of money allocated to an SR fund is less related to the fund’s performance; indeed, if investors view SR investments as one of their mental accounts, they might allocate a smaller amount to SR funds and try to optimize within this account or they might be willing to allocate a smaller amount to SR funds regardless of their performance. Second, since I analyze investors’ trading behavior at an individual investor level, the size of their investments can be dependent on many factors segment and 2.5% is used to fund individuals’ accounts in the PPS. Participation in the PPS is mandatory and individuals are offered a number of mutual funds. They can adjust their mutual fund choices on a daily basis; those who do not make an investment choice have their money invested in a default fund. 11 The within-individual correlation is calculated using the R2 from a regression model with individual-specific fixed effects and it is always positive. Years in which individuals have no trading activities are excluded, and there are thus no observations without buy and sell transactions. The regression model is given by Buyi,t = β(Selli,t ) + αi + εi,t , where Buyi,t (Selli,t ) is equal to 1 if individual i buys (sells) at least one fund during year t and 0 otherwise. 12 Although individual-specific fixed effects can be used in non-lineal models as well, in this paper, their use becomes challenging because individuals are observed over a short period of time, which means that the estimated coefficients are likely to be biased (Greene, 2004).

9 Page 9 of 33

that may not be observable, even after controlling for investors’ portfolio size and net wealth. For example, individuals may have different liquidity needs from one period to another (e.g., they may allocate SEK 10,000 for investment in one year, but only SEK 2,000 in the next year). Although the indicator variables do not capture differences in the size of individuals’ investments, they do capture their decisions to buy and sell funds.

ip t

Regarding the determinants of investors’ decisions with respect to mutual funds, prior studies have established that there is a correlation between fund flows and past returns (see, e.g. Ippolito, 1992; Chevalier and Ellison, 1997; Sirri and Tufano, 1998). The relationship between fund flows

cr

and returns, however, tends to be convex: there are larger inflows in response to positive returns and smaller outflows in response to negative returns (Chevalier and Ellison, 1997; Sirri and Tufano,

us

1998; Bollen, 2007). Moreover, there are many ways to measure fund performance, such as using unadjusted (raw) returns (Bollen, 2007; Ivkovi´c and Weisbenner, 2009; Renneboog et al., 2011), market-adjusted returns (Chevalier and Ellison, 1997; Barber, Odean, and Zheng, 2005) or risk-

an

adjusted returns (Del Guercio and Tkac, 2008; Barber et al., 2016). The high correlation between these performance measures makes it difficult to isolate the factors to which investors attend and to estimate their impact (Del Guercio and Tkac, 2008; Barber et al., 2016). Nonetheless, Del Guercio

M

and Tkac (2002), and Barber et al. (2016) find that more sophisticated investors respond to riskadjusted returns and less sophisticated investors respond to market-adjusted returns.13 Considering that this paper focuses on individual investors (who are presumably less sophisticated (see, e.g.,

d

Grinblatt and Keloharju, 2000; Barber and Odean, 2013))14 , market-adjusted returns may be the more relevant measure of fund performance.

te

Table III presents descriptive statistics about the raw and the market-adjusted returns on the mutual funds held by individual investors in the sample (note that these funds represent about a

Ac ce p

third of all the available equity mutual funds in Sweden during the sample period). [Table III about here]

To examine whether and how the past performance of SR and conventional funds is related to individual investors’ buying and selling decisions, I estimate the following LPM: Buyi,j,t or Selli,j,t = α + β1 (P ositivej,t × Rj,t ) + β2 (N egativej,t × Rj,t ) + β3 SRj + β4 (P ositivej,t × Rj,t × SRj ) + β5 (N egativej,t × Rj,t × SRj ) + F und charateristicsj,t × Ψ

(1)

+ Socio-demographic characteristicsi,t × Ω + εi,j,t

where Buyi,j,t (Selli,j,t ) is equal to 1 if there is an increase (a decrease) in the number of units in individual i ’s position in fund j at the end of year t relative to year t − 1 and 0 if there is no change 13

Del Guercio and Tkac (2002) find that retail mutual fund flows are strongly related to unadjusted returns and weakly related to risk-adjusted returns. However, Del Guercio and Tkac (2002) refer to unadjusted returns as returns in excess of those on the S&P 500, but they interpret the findings as revealing a relationship between unadjusted returns and fund flows. 14 Related to this, Reid and Rea (2003) show that most households buy mutual funds after receiving assistance from financial advisors.

10 Page 10 of 33

in the number of units, Rj,t denotes the market-adjusted return on fund j during year t (where the return on the MSCI World Index is used as a benchmark), P ositivej,t (N egativej,t ) is equal to 1 if Rj,t is positive (negative) and 0 otherwise, and SRj is equal to 1 if fund j is an SR fund and 0 otherwise.15 The fund characteristics included are the volatility of fund returns and the estimated coefficients

ip t

from the Carhart (1997) four-factor model, estimated with daily global risk factors and the returns on the 30-day U.S. Treasury bill obtained from the Applied Quantitative Research (AQR) Capital Management website. The estimated coefficients from the four-factor model are included to account

cr

for investors’ style preferences; according to Bollen (2007), SR funds have investment styles that are significantly different than those of conventional funds, in the sense that they tend to have a

us

smaller exposure to stocks with high momentum and stocks with smaller market capitalizations. Given that prior studies find that individuals’ socio-demographic characteristics are related to the propensity to trade (e.g., male, better educated and wealthier individuals tend to trade more

an

often (Barber and Odean, 2001; Dahlquist et al., 2016)), control variables for individuals’ gender, parenthood, marital status, cohort, education level, completion of an economics degree, annual income and net wealth are also included in the model above.

M

[Table IV about here]

Table IV presents the results from estimating Equation (1). Models 1, 2, 5 and 6 are based on

d

the sample of all investors and Models 3, 4, 7 and 8 are based on the sample consisting of only SR

te

investors. Given that this study focuses on the within variation in the behaviors of SR investors towards SR and conventional funds, I focus on the results obtained in Models 4 and 8. In general, the results in Models 1 to 4 indicate that individuals are more likely to buy better performing funds.

Ac ce p

These results are consistent with prior studies on fund flows, (e.g., Sirri and Tufano, 1998; Chevalier and Ellison, 1997; Del Guercio and Tkac, 2002), which find that fund flows chase past returns. The results from Models 1 to 4 further indicate that investors’ buying decisions are similarly sensitive to the past positive returns on SR and conventional funds. Moreover, while Models 1 to 3 suggest that investors’ buying decisions are more sensitive to the past negative returns on SR funds and that investors are less likely to buy SR than conventional funds as past negative returns decrease, Model 4 (which includes individual-specific fixed effects) indicates that SR investors’ buying decisions are similarly sensitive to the past negative returns on SR and conventional funds. Furthermore, Models 5 to 8 show that past fund returns are not a strong predictor of individuals’ selling decisions because the statistical significance disappears after controlling for individual-specific fixed effects. Nonetheless, the results in Model 8 indicate that investors are more sensitive to the past negative returns on SR than conventional funds and that they are less likely to sell SR funds as their past negative returns decrease. Given that prior studies, such as Odean (1999) and Grinblatt and Keloharju (2001), find that investors’ trades are more sensitive to extreme returns, I also show, in 15

In this analysis, P ositivej,t (N egativej,t ) is winsorized at the 99th (1st) percentile.

11 Page 11 of 33

Table A2, additional tests including dummy variables for funds that perform in the top and the bottom deciles in a given year. The findings, however, remain similar. For ease of interpretation, in Figure 1, I also plot the marginal effects from Models 4 and 8, which confirm the relationships established above. Although one of the drawbacks of using a LPM is that the predicted probability can be less than 0 and more than 1, Figure 1 shows that this is

ip t

not the case in this paper.

The result that SR investors are less likely to sell poor performing SR than conventional funds indicates that they derive utility from investing in funds with SR attributes. These findings are

cr

consistent with the findings in Bollen (2007) and Renneboog et al. (2011) that poor performing SR funds seem to have smaller decreases in flows than poor performing conventional funds. Moreover,

us

the relations between past returns and the buying or selling decisions of conventional investors documented in this study are in line with the already established results in the literature on the relation between mutual fund flows and past returns, that poor performance does not lead to fund

an

outflows to the same extent that superior performance leads to fund inflows (see, e.g., Ippolito, 1992; Chevalier and Ellison, 1997; Sirri and Tufano, 1998).

SR Investors and Fund Flows

M

B.

In the previous section, I treated all SR investors as a homogeneous group. However, it seems more realistic to assume that they buy SR mutual funds for different reasons. According to Fama

d

and French (2007), investment choices that are consistent over time are more likely to be influenced by personal preferences because individuals’ preference biases are not expected to diminish over

te

time. Thus, categorizing investors into types by looking at their behavior over time could be useful. Indeed, if we look at investors’ behaviors at a particular point in time, it is possible that

Ac ce p

some investors hold a specific SR fund just because that fund happens to have other characteristics that investors are looking for; for example, an SR fund may have a good performance record or have an investment style (i.e., market, growth, value or momentum) that matches the preferences of some investors. I therefore classify SR investors into two groups, sticky and non-sticky SR investors, as defined earlier. B.1.

Sticky SR Investors’ Portfolio Characteristics

Panels D and E in Table I present summary statistics of sticky and non-sticky SR investors’ portfolio characteristics. Non-sticky SR investors appear to hold a greater number of funds (4.21) than sticky SR investors (3.81) and to have a slightly lower fund wealth. Moreover, only 20.75% of non-sticky SR investors hold one fund, while this is true of 27.19% of sticky SR investors. Furthermore, the frequency of trading activities, as shown in Panels D and E in Table 2, reveals that non-sticky SR investors are more likely to trade: they buy and sell mutual funds more frequently, but they are less likely to engage in both buying and selling within the same year (the withinindividual correlation between the buying and selling activities for non-sticky SR investors is 0.23, compared to 0.32 for sticky SR investors. 12 Page 12 of 33

[Table V about here] To obtain a better overview of sticky SR investors’ portfolio characteristics, the composition of their portfolios is presented in Table V. This table provides detailed summary statistics for sticky SR investors, grouped by deciles of their portfolio weights in SR funds.16 Perhaps the most

ip t

interesting statistic from the table is that across all groups of sticky SR investors, the median number of SR funds in their portfolios is one. There are at least two plausible explanations for this. First, it is possible that there is a limited number of SR funds, so there are less SR funds that

cr

satisfy the investment criteria of these investors. In relation to this, Benson and Humphrey (2008), for example, find that SR investors are more likely to reinvest in funds they already own. They

us

hypothesize that, because non-financial criteria differ significantly across funds, SR investors may not be able to find an alternative fund easily and are thus less likely to switch between SR funds. Second, it is possible that sticky SR investors avoid incurring extra costs in their search for SR

an

investments, in the sense that these investors may find learning about SR attributes overwhelming (Mackenzie and Lewis, 1999). Sethi-Iyengar, Huberman, and Jiang (2004), for example, find that the participation rate in 401(k) plans falls as the number of fund options increases. It is, however, difficult to draw any conclusion with respect to these explanations. Nevertheless,

M

the data confirms the finding in Mackenzie and Lewis (1999) that investors may choose to be somewhat responsible, as about 70% of sticky SR investors hold both conventional and SR funds. In addition, the portfolio weights on SR funds seem to be negatively correlated with the number

d

of holding funds and the value of equity fund wealth. This indicates that the amounts of money

te

invested in SR funds are not largely different across individuals in terms of monetary value; that is, investors with larger portfolios are not investing more in SR funds (in monetary terms) than

B.2.

Ac ce p

investors with smaller portfolios.

Sticky SR Investors and Fund Flows

To investigate the impact of SR investor types on fund flows, two approaches are being used. First, I examine the relationship between aggregated money flows (calculated using the flows of the investors in the sample) and past market-adjusted returns. This test also serves as a basis for comparison with earlier closely related studies that use a similar methodology (Bollen, 2007; Benson and Humphrey, 2008; Renneboog et al., 2011). Second, I investigate the flow-performance relationship at the individual investor level, while also controlling for individuals’ unobserved timeinvariant characteristics (i.e., individual-specific fixed effects). To examine the flow-performance relationship, I employ a similar methodology as in Bollen (2007) and Renneboog et al. (2011), with the distinction that I compute aggregated fund flows from individuals’ mutual fund positions instead of using funds’ total net asset values. Although this measure of fund flows is incomplete (since it does not consider funds’ total net asset values), the 16 Note that deciles 8-10 are presented together because all the portfolios in these deciles consist of only SR funds (i.e., 100% portfolio weight in SR funds).

13 Page 13 of 33

advantage of this approach is that the data consist of money flows contributed by only individual investors. I first estimate the relative changes in funds’ aggregated fund flows as: n X

n X

(2) F und valuei,j,t−1

i=1

ip t

F und f lowj,t =

(F und valuei,j,t − F und valuei,j,t−1 (1 + Rj,t ))

i=1

where F und V aluei,j,t denotes the value of the position of individual i in fund j at the end of year

cr

t and n is the number of individuals in the sample who hold fund j at the end of year t. Note that F und F lowj,t has several extreme positive values and, therefore, I winsorize it at the 95th

us

percentile. The following regression model is then estimated:

F und F lowj,t = α + β1 (P ositivej,t × Rj,t ) + β2 (N egativej,t × Rj,t ) + β3 SRj

(3)

an

+ β4 (P ositivej,t × Rj,t × SRj ) + β5 (N egativej,t × Rj,t × SRj ) + F und characteristicsj,t × Ψ + εj,t

where the control variables for the characteristics of the funds include the volatility of fund returns

M

and the natural logarithm of the sum of F und V aluei,j,t−1 across all individuals in the sample who hold fund j at the end of year t − 1.17

d

[Table VI about here]

te

Model 1 in Table VI shows that the aggregated fund flow-performance relationship documented here is consistent with Renneboog et al. (2011) for European funds. That is, the flows of SR and

Ac ce p

conventional funds are similarly sensitive to past positive and negative returns. Next, I investigate how sticky SR investors affect fund flows by selecting SR funds with high percentage of sticky SR investors. I estimate a regression model similar to that in Equation (3), Sticky where SRj is replaced with an indicator variable, SRj,t , which takes a value of 1 if the majority

(i.e., more than 50%) of SR investors in SR fund j at the end of year t are sticky SR investors and Sticky 0 otherwise. Note that SRj,t remains almost unchanged over time for a given fund.

Model 2 in Table VI shows that the flows of SR funds with a higher percentage of sticky SR investors seem to be less sensitive to past positive returns than those of conventional funds (β4 = −0.378, statistically significant at the 10% level). However, the flows of such SR and conventional funds seem to be similarly sensitive to past negative returns. Moreover, these SR funds seem to experience smaller increases in their flows compared to conventional funds (β3 = −0.123, statistically significant at the 1% level). Although these results can be interpreted as sticky SR investors caring less about returns, the result that the flows of SR funds with a higher percentage of sticky SR investors and of conventional 17 Note that, in Equation (3), F und F lowj,t = ln(−min(F und F lowj,t ) + 0.00001 + F und F lowj,t ). Moreover, some funds are held by a small number of individuals in the sample and I therefore exclude funds that have 100 or fewer investors at the end of year t − 1.

14 Page 14 of 33

funds are similarly sensitive to past negative returns weakens the evidence in favor of the valuesdriven motive hypothesis. These results, however, should be interpreted with caution because important control variables (e.g., fund age18 ) are omitted in the regressions due to data limitations. Given that the measure of fund flows in this paper is calculated using the flows of only the individual investors in the sample, it may be a considerably imprecise measure. Thus, to show

ip t

that the flow-performance relationship found here is consistent with those found in prior studies on mutual funds, I also test this relationship using the decile ranks of the funds’ past returns (deciles 5 and 6 are used as the base level). The results in Model 3 show that the flows of the top performing

cr

funds (those in deciles 8-10) are more sensitive to past performance, which is consistent with prior studies (e.g., Ivkovi´c and Weisbenner, 2009; Sirri and Tufano, 1998; Chevalier and Ellison, 1997).

us

Finally, I also examine how individuals’ fund flows respond to past fund performance. This way, I can investigate the behaviors of individuals while also being able to control for the unobserved heterogeneity among them. These analyses complement the analyses on aggregated fund flows,

an

which may be influenced by large changes in the positions of very wealthy individuals. The relative changes in individuals’ fund flows are calculated as:

F und valuei,j,t − F und valuei,j,t−1 (1 + Rj,t ) F und valuei,j,t−1

(4)

M

Investor fund flow i,j,t =

Table VII presents the results from estimating a regression model similar to that in Equation (1), where the dependent variable is Investor fund flowi,j,t . Based on the full sample, investors’ flows

d

respond to past positive returns (significant at the 1% level), but not to past negative returns.

te

Moreover, Table VII provides strong support for the notion that individuals do have values-driven motives towards investing in SR funds. In Models 3 to 5, there is a strong evidence that SR

Ac ce p

investors (both sticky and non-sticky) are less sensitive to the past negative returns on SR than conventional funds, in line with prior studies which find that investors tend to be more loyal to SR funds (Bollen, 2007; Renneboog et al., 2011). While Models 3 to 5 also indicate that SR investors are more sensitive to the past positive returns on SR than conventional funds, only the results in Models 3 and 5 are statistically significant. [Table VII about here]

In addition, Benson and Humphrey (2008) find that SR fund flows are more persistent than conventional fund flows and that SR investors are more likely to invest in the SR funds they already own. In relation to this, I also include lagged investor fund flows (Investor fund flowi,j,t−1 ) in the models presented in Table VII. Surprisingly, the coefficients on the lagged flows for SR funds in Models 3 and 5 of Table VIII are significantly negative (at the 1% level). This indicates that sticky SR investors’ flows to SR funds are less persistent than their flows to conventional funds. This result is somewhat puzzling given that there is also evidence in the same model that sticky SR investors are more (less) sensitive to past positive (negative) returns. On the other hand, the 18

Chevalier and Ellison (1997) find that the flows of younger funds are more sensitive to past performance.

15 Page 15 of 33

results in Model 4 (sample of only non-sticky SR investors) indicate a higher persistence of SR fund flows relative to conventional fund flows. Moreover, the results in Model 4 of Table VIII differ dramatically from the those in Model 4 of Table VII (i.e., after controlling for the lagged flows). In this model, non-sticky SR investors’ flows appear to be less sensitive to the past positive returns on SR than conventional funds and similarly sensitive to the past negative returns on SR

ip t

and conventional funds.

cr

[Table VIII about here]

One potential interpretation of these results is that sticky SR investors may be willing to allocate some portion of their portfolios to SR funds and that they are more likely to hold on to these funds,

us

but are less willing to increase their investments in them in subsequent periods, unless the funds perform extremely well. This is possible considering that the interview resuls in Mackenzie and Lewis (1999) suggest that investors may choose to allocate a small amount to SR funds so that they

an

feel less guilty and avoid incurring extra search costs. On the other hand, the observed positive relation between the lagged and the current flows to SR funds in the sample of only non-sticky SR investors could be driven by new or profit-driven investors or both in SR funds. The increased

M

interest in SR funds, especially after 2000 (see Bengtsson, 2008; Renneboog et al., 2008b) may lead to the observations in prior studies that SR fund flows are more sensitive to lagged flows than are conventional fund flows. If this is the case, the consecutively large increases in SR fund flows could

d

be driven by new investors in SR funds rather than reinvestment from existing fund clients. This

te

could potentially explain the observed lower aggregated flow sensitivity of SR funds with a higher percentage of sticky SR investors (Model 2 in Table VI).

Ac ce p

IV.

Conclusion

This study describes the behavior of individual investors in relation to investment in SR equity mutual funds. Investors who hold at least one SR fund in any given year (SR investors) appear to hold both conventional and SR funds (about 70% of all SR investors) and to hold a larger number of funds. The findings also indicate that they have lower portfolio weights in SR funds, possess greater portfolio wealth and engage in slightly more trading activities. Furthermore, I examine investors’ behavior towards SR and conventional funds in relation to past returns using two approaches. Because prior studies, such as (Mackenzie and Lewis, 1999), suggest that individuals may think of SR investment as a part of their mental accounts (which likely results in SR investors limiting their SR holdings to a certain amount), I examine their propensity to buy and sell SR and conventional funds separately and I also examine the flows in and out of these funds at an individual investor level and at a fund level. The results show that SR investors’ buying decisions are similarly sensitive to the past positive and negative returns on SR and conventional funds. Their selling decisions, however, are more sensitive to the past negative returns on SR funds and SR investors are less likely to sell SR than 16 Page 16 of 33

conventional funds as past negative returns decrease. In aggregate, the flows of SR and conventional funds are similarly sensitive to past returns. Nevertheless, when considering only SR funds with a majority of their investors being sticky SR investors, the results show that the flows of these funds are less sensitive to past positive returns than the flows of conventional funds. On the other hand, at an individual investor level, the results show that sticky SR investors’ flows are more sensitive

ip t

to past positive returns and are less sensitive to past negative returns on SR than conventional funds. Although these results support the hypothesis that sticky SR investors are likely to have values-driven motives for holding SR funds and care less about the returns on such funds, the

cr

analysis of the persistence of their fund flows indicates that they are more likely to reinvest in their conventional funds than in their SR funds.

us

For the mutual fund industry, these findings suggest that SR mutual funds may have less volatile flows when their performance is poor and that they may find it difficult to raise new funds from their existing clients. This indicates that the managers of such funds may face less uncertainty with

an

respect to their funding, but they may not be able to benefit from economies of scale. My findings provide interesting avenues for future research. I believe that it may be worthwhile to investigate the dynamics of SR investors’ trades in relation to the SR scores and attributes

M

of funds, such as their screening strategies in relation to the environment, sin companies, and corporate governance. It would also be of interest to re-examine investors’ trading behavior with respect to SR funds by using more detailed data on all of their transactions, as well as other mutual

Ac ce p

te

d

fund characteristics such as fees, expenses, age, and size.

17 Page 17 of 33

REFERENCES Ai, Chunrong, and Edward C Norton, 2003, Interaction terms in logit and probit models, Economics Letters 80, 123–129. Aktas, Nihat, Eric De Bodt, and Jean-Gabriel Cousin, 2011, Do financial markets care about SRI?

ip t

evidence from mergers and acquisitions, Journal of Banking & Finance 35, 1753–1761.

Alexander, Gordon J, Jonathan D Jones, and Peter J Nigro, 1997, Investor self-selection: evidence

cr

from a mutual fund survey, Managerial and Decision Economics 719–729.

Bailey, Warren, Alok Kumar, and David Ng, 2011, Behavioral biases of mutual fund investors,

us

Journal of Financial Economics 102, 1–27.

Barber, Brad M, Xing Huang, and Terrance Odean, 2016, Which factors matter to investors?

an

evidence from mutual fund flows, Review of Financial Studies 29, 2600–2642. Barber, Brad M, and Terrance Odean, 2001, Boys will be boys: Gender, overconfidence, and

M

common stock investment, The Quarterly Journal of Economics 116, 261–292. Barber, Brad M, and Terrance Odean, 2013, The behavior of individual investors, in Handbook of the Economics of Finance, volume 2, 1533–1570 (Elsevier).

d

Barber, Brad M, Terrance Odean, and Lu Zheng, 2005, Out of sight, out of mind: The effects of

te

expenses on mutual fund flows, The Journal of Business 78, 2095–2120. Bauer, Rob, Jeroen Derwall, and Rog´er Otten, 2007, The ethical mutual fund performance debate:

Ac ce p

New evidence from canada, Journal of Business Ethics 70, 111–124. Bauer, Rob, Kees Koedijk, and Roger Otten, 2005, International evidence on ethical mutual fund performance and investment style, Journal of Banking & Finance 29, 1751–1767. Bengtsson, Elias, 2008, A history of scandinavian socially responsible investing, Journal of Business Ethics 82, 969–983.

Benson, Karen L, and Jacquelyn E Humphrey, 2008, Socially responsible investment funds: Investor reaction to current and past returns, Journal of Banking & Finance 32, 1850–1859. Berry, Thomas C, and Joan C Junkus, 2013, Socially responsible investing: An investor perspective, Journal of Business Ethics 112, 707–720. Bollen, Nicolas PB, 2007, Mutual fund attributes and investor behavior, Journal of Financial and Quantitative Analysis 42, 683–708. Calvet, Laurent E, John Y Campbell, and Paolo Sodini, 2007, Down or out: Assessing the welfare costs of household investment mistakes, Journal of Political Economy 115, 707–747.

18 Page 18 of 33

Calvet, Laurent E, John Y Campbell, and Paolo Sodini, 2009, Fight or flight? portfolio rebalancing by individual investors, The Quarterly Journal of Economics 124, 301–348. Campbell, John Y, 2006, Household finance, Journal of Finance 61, 1553–1604.

ip t

Carhart, Mark M, 1997, On persistence in mutual fund performance, Journal of Finance 52, 57–82. Chevalier, Judith A, and Glenn D Ellison, 1997, Risk taking by mutual funds as a response to incentives, The Journal of Political Economy 105, 1167–1200.

cr

Cooper, Michael J, Huseyin Gulen, and P Raghavendra Rau, 2005, Changing names with style:

us

Mutual fund name changes and their effects on fund flows, Journal of Finance 60, 2825–2858. Dahlquist, Magnus, Jos´e Vicente Martinez, and Paul S¨oderlind, 2016, Individual investor activity and performance, Review of Financial Studies 30, 866–899.

an

Del Guercio, Diane, and Paula A Tkac, 2002, The determinants of the flow of funds of managed portfolios: Mutual funds vs. pension funds, Journal of Financial and Quantitative Analysis 37,

M

523–557.

Del Guercio, Diane, and Paula A Tkac, 2008, Star power: The effect of monrningstar ratings on mutual fund flow, Journal of Financial and Quantitative Analysis 43, 907–936.

d

Derwall, Jeroen, Kees Koedijk, and Jenke Ter Horst, 2011, A tale of values-driven and profit-seeking

te

social investors, Journal of Banking & Finance 35, 2137–2147. Døskeland, Trond, and Lars Jacob Tynes Pedersen, 2015, Investing with brain or heart? a field

Ac ce p

experiment on responsible investment, Management Science 62, 1632–1644. Fama, Eugene F, and Kenneth R French, 2007, Disagreement, tastes, and asset prices, Journal of Financial Economics 83, 667–689.

Fehr, Ernst, and Urs Fischbacher, 2003, The nature of human altruism, Nature 425, 785–791. Ghoul, Sadok El, and Aymen Karoui, 2017, Does corporate social responsibility affect mutual fund performance and flows?, Journal of Banking & Finance 77, 53 – 63. Gil-Bazo, Javier, Pablo Ruiz-Verd´ u, and Andr´e AP Santos, 2010, The performance of socially responsible mutual funds: the role of fees and management companies, Journal of Business Ethics 94, 243–263. Greene, William, 2004, The behaviour of the maximum likelihood estimator of limited dependent variable models in the presence of fixed effects, The Econometrics Journal 7, 98–119. Grinblatt, Mark, and Matti Keloharju, 2000, The investment behavior and performance of various investor types: a study of finland’s unique data set, Journal of financial economics 55, 43–67.

19 Page 19 of 33

Grinblatt, Mark, and Matti Keloharju, 2001, What makes investors trade?, Journal of Finance 56, 589–616. Hellstr¨om, Jorgen, Nicha Lapanan, and Rickard Olsson, 2015, Intergenerational transmission of pro-social values: Socially responsible investment among parents and adult children, Unpublished

ip t

working paper . Hong, Harrison, and Marcin Kacperczyk, 2009, The price of sin: The effects of social norms on

cr

markets, Journal of Financial Economics 93, 15–36.

Ippolito, Richard A, 1992, Consumer reaction to measures of poor quality: Evidence from the

us

mutual fund industry, Journal of Law and Economics 35, 45–70.

Ivkovi´c, Zoran, and Scott Weisbenner, 2009, Individual investor mutual fund flows, Journal of

an

Financial Economics 92, 223–237.

Kempf, Alexander, and Peer Osthoff, 2007, The effect of socially responsible investing on portfolio performance, European Financial Management 13, 908–922.

M

Keswani, Aneel, and David Stolin, 2008, Which money is smart? mutual fund buys and sells of individual and institutional investors, Journal of Finance 63, 85–118. Levitt, Steven D, and John A List, 2007, What do laboratory experiments measuring social pref-

d

erences reveal about the real world?, Journal of Economic Perspectives 21, 153–174.

te

Mackenzie, Craig, and Alan Lewis, 1999, Morals and markets: the case of ethical investing, Business

Ac ce p

Ethics Quarterly 9, 439–452.

McLachlan, Jonathan, and John Gardner, 2004, A comparison of socially responsible and conventional investors, Journal of Business Ethics 52, 11–25. Odean, Terrance, 1999, Do investors trade too much?, American Economic Review 89, 1279–1298. O’neal, Edward S, 1997, How many mutual funds constitute a diversified mutual fund portfolio?(digest summary), Financial Analysts Journal 53, 37–46. Reid, Brian K, and John D Rea, 2003, Mutual fund distribution channels and distribution costs, Perspective 9, 1–20.

Renneboog, Luc, Jenke Ter Horst, and Chendi Zhang, 2008a, The price of ethics and stakeholder governance: The performance of socially responsible mutual funds, Journal of Corporate Finance 14, 302–322. Renneboog, Luc, Jenke Ter Horst, and Chendi Zhang, 2008b, Socially responsible investments: Institutional aspects, performance, and investor behavior, Journal of Banking & Finance 32, 1723–1742. 20 Page 20 of 33

Renneboog, Luc, Jenke Ter Horst, and Chendi Zhang, 2011, Is ethical money financially smart? nonfinancial attributes and money flows of socially responsible investment funds, Journal of Financial Intermediation 20, 562–588. Riedl, Arno, and Paul Smeets, 2017, Why do investors hold socially responsible mutual funds?,

ip t

Journal of Finance 72, 2505–2550. Sethi-Iyengar, Sheena, Gur Huberman, and Wei Jiang, 2004, How much choice is too much? contributions to 401 (k) retirement plans, in Pension Design and Structure: New Lessons From

cr

Behavioral Finance, volume 83, 84–87 (Oxford University Press, New York).

Shefrin, Hersh, and Meir Statman, 2000, Behavioral portfolio theory, Journal of Financial and

us

Quantitative Analysis 35, 127–151.

SIFA, 2017, Swedish investment fund association, http://fondbolagen.se/en/Statistics/

an

Fund-saving-by-category/Archive/, Accessed: 2017-01-30.

Sirri, Erik R, and Peter Tufano, 1998, Costly search and mutual fund flows, Journal of Finance 53,

M

1589–1622.

Sparkes, Russell, and Christopher J Cowton, 2004, The maturing of socially responsible investment: A review of the developing link with corporate social responsibility, Journal of Business Ethics

d

52, 45–57.

te

Utz, Sabastian, and Maximillian Wimmer, 2014, Are they any good at all? a financial and ethical analysis of socially responsible mutual funds, Journal of Asset Management 15, 72–82.

Ac ce p

Wilcox, Ronald T, 2003, Bargain hunting or star gazing? investors preferences for stock mutual funds, The Journal of Business 76, 645–663.

21 Page 21 of 33

Table I Individual investors’ equity mutual fund portfolios This table presents summary statistics for individual investors’ equity mutual fund portfolios. Panel A reports summary statistics for all investors, Panel B for conventional investors, Panel C for SR investors, Panel D for sticky SR investors and Panel E for non-sticky SR investors. The fund portfolio value and the SR fund portfolio value are measured in SEK. M denotes millions. SD

Min.

Max.

2.51 0.09 61,261 2,187 0.04

2.23 0.38 318,222 32,633 0.17

1 0 200 0 0

125 9 104M 9M 1

402, 523 402, 523 402, 523 402, 523 402, 523 43.98

2.37 57,726

1.97 299,607

35 104M

365, 946 365, 946 45.96

4.00 1.01 96,621 24,069 0.41

3.69 0.82 463,635 105,797 0.40

1 0 200 0 0

125 13 43M 9M 1

36, 577 36, 577 36, 577 36, 577 36, 577 24.17

3.81 1.33 99,455 34,629 .57

3.55 0.76 0.43M 0.14M .38

1 1 200 200 .001

66 9 43M 9M 1

19, 428 19, 428 19, 428 19, 428 19, 428 27.19

4.21 0.65 93,411 12,105 0.22

3.84 0.72 499,667 44,282 0.33

1 0 200 0 0

125 13 37M 1.8M 1

17, 149 17, 149 17, 149 17, 149 17, 149 20.75

Number of funds Fund portfolio value % of investors holding 1 fund

d

Ac ce p

Number of funds Number of SR funds Fund portfolio value SR fund portfolio value SR fund portfolio weight % of investors holding 1 fund

te

Panel D: Sticky SR investors

M

Panel C: SR investors Number of funds Number of SR funds Fund portfolio value SR fund portfolio value SR fund portfolio weight % of investors holding 1 fund

1 200

an

Panel B: Conventional investors

cr

Number of funds Number of SR funds Fund portfolio value SR fund portfolio value SR fund portfolio weight % of investors holding 1 fund

us

Panel A: All investors

Obs.

ip t

Mean

Panel E: Non-Sticky SR investors Number of funds Number of SR funds Fund portfolio value SR fund portfolio value SR fund portfolio weight % of investors holding 1 fund

22 Page 22 of 33

Table II Individual investors’ trading activities

2003

2004

2005

2006

2007

All years

32.34 14.51 59.31 82,751

35.47 20.68 52.78 80,352

33.26 24.12 54.34 79,939

31.71 28.7 54.87 79,428

28.98 29.71 57.96 80,053

32.35 23.47 55.87 402,523 0.49

32.44 14.40 58.97 75,495

36.20 20.39 51.69 72,979

33.75 23.62 53.54 72,606

32.20 28.02 54.04 72,135

29.38 28.83 57.34 72,731

32.79 22.98 55.15 365,946 0.51

31.30 15.67 62.82 7,256

28.22 23.61 63.54 7,373

28.31 29.14 62.24 7,333

26.86 35.40 63.09 7,293

24.98 38.36 64.08 7,322

27.93 28.45 63.15 36,577 0.28

32.85 15.74 60.2 3,887

35.02 19.86 56.42 3,958

36.26 24.89 54.07 3,958

36.51 28.78 52.43 3,958

34.55 20.38 57.13 19,428 0.32

23.06 32.39 67.27 3,486

20.44 40.03 69.07 3,375

15.71 47.89 73.79 3,335

11.41 49.64 77.79 3,364

20.43 37.59 69.97 17,149 0.23

d te

Panel D: Sticky SR investors 31.91 11.92 63.05 3,667

Ac ce p

No activity Sell Buy Number of investors Correlation Buy&Sell

M

Panel C: SR investors No activity Sell Buy Number of investors Correlation Buy&Sell

us

No activity Sell Buy Number of investors Correlation Buy&Sell

an

Panel B: Conventional investors

cr

Panel A: All investors No activity Sell Buy Number of investors Correlation Buy&Sell

ip t

This table presents summary statistics for individual investors’ trading activities over time. Investors are classified as having buying (selling) activity if they have bought (sold) at least one fund in a given year, otherwise they are classified as having no activity. Panel A reports summary statistics for all investors, Panel B for conventional investors, Panel C for SR investors, Panel D for sticky SR investors and Panel E for non-sticky SR investors. M denotes millions.

Panel E: Non-sticky SR investors No activity Sell Buy Number of investors Correlation Buy&Sell

30.68 19.50 62.58 3,589

23 Page 23 of 33

Table III Returns on conventional and SR equity mutual funds This table presents summary statistics for the raw and the market-adjusted returns on the conventional and the SR equity mutual funds held by the individual investors in the sample over time. 2003

2004

2005

2006

0.261 0.228 0.198

0.119 0.111 0.127

0.337 0.315 0.189

0.170 0.170 0.172

−0.085 −0.118 0.198

−0.035 −0.043 0.127

0.221 0.199 0.189

−0.043 −0.043 0.172

2007

All years

Observations

381

420

0.176 0.159 0.109

0.087 0.103 0.062

−0.170 −0.187 0.109

−0.067 −0.051 0.062

497

62

661

2,528

0.142 0.177 0.104

−0.003 0.026 0.097

0.140 0.130 0.139

−0.071 −0.036 0.104

−0.125 −0.148 0.097

−0.046 −0.069 0.139

0.197 0.184 0.066

63

68

71

75

339

Ac ce p

te

Observations

0.011 −0.026 0.207

M

Market-adjusted return Mean Median SD

−0.015 −0.074 0.220

0.314 0.300 0.066

d

Raw return Mean Median SD

569

an

Panel B: SR funds

0.191 0.165 0.207

cr

Market-adjusted return Mean Median SD

0.107 0.048 0.220

us

Raw return Mean Median SD

ip t

Panel A: Conventional funds

24 Page 24 of 33

25

Page 25 of 33

Controls Year FE Individual FE R2 Number of investors Observations

N egativej,t × Rj,t × SRj

P ositivej,t × Rj,t × SRj

SRj

N egativej,t × Rj,t

Fund performance P ositivej,t × Rj,t

764,540

0.021

Yes Yes

0.203∗∗∗ (0.008) 0.031∗∗∗ (0.008) 0.006 (0.007) −0.026 (0.034) 0.191∗∗∗ (0.038)

(1)

(2)

Yes Yes Yes 0.008 88,419 764,540

0.149∗∗∗ (0.006) 0.013∗∗ (0.006) 0.016∗∗∗ (0.004) 0.004 (0.031) 0.097∗∗∗ (0.028)

All investors

108,970

0.022

Yes Yes

0.228∗∗∗ (0.021) 0.098∗∗∗ (0.021) 0.053∗∗∗ (0.008) −0.017 (0.040) 0.145∗∗∗ (0.045) Yes Yes Yes 0.006 7,858 108,970

676,107

0.044

101,995

0.069

ip t

cr Yes Yes Yes 0.035 82,300 676,107

Yes Yes

us Yes Yes

0.128∗∗∗ (0.022) −0.035∗ (0.021) −0.058∗∗∗ (0.007) 0.091∗∗ (0.046) 0.068∗ (0.040)

(7)

(8)

Yes Yes Yes 0.048 7,576 101,995

0.011 (0.016) −0.016 (0.016) −0.023∗∗∗ (0.005) 0.026 (0.038) 0.066∗ (0.035)

SR investors

0.006 (0.006) 0.001 (0.006) −0.023∗∗∗ (0.005) 0.013 (0.033) 0.023 (0.030)

( 6)

All investors

0.145∗∗∗ (0.009) −0.012 (0.008) −0.047∗∗∗ (0.006) 0.085∗∗ (0.040) 0.008 (0.036)

(5)

an

M

(4) 0.168∗∗∗ (0.016) 0.068∗∗∗ (0.017) 0.015∗∗∗ (0.005) 0.026 (0.037) 0.050 (0.034)

SR investors

d

(3)

te

Buyi,j,t

Selli,j,t

This table presents the results from the analyses of the relation between individual investors’ buying and selling decisions and the past positive and negative returns on conventional and SR funds. Models 1-8 represent the results from estimating Equation (1); these models differ with respect to the sample of investors and the individual-specific fixed effects. The standard errors (reported in parentheses) are clustered by individuals. *, **, *** indicate statistical significance at the 10%, 5% and 1%, respectively. See Table A1 for the variable definitions.

Table IV Individual investors’ buying and selling decisions and past returns on conventional and SR funds

Ac ce p

26

Page 26 of 33

Number of funds Mean Median 90th percentile Number of SR funds Mean Median 90th percentile Total investment in equity mutual funds Mean Median 90th percentile Portfolio weight on SR funds Mean Median 90th percentile Observations

te

6.46 6 11 1.19 1 2 146,258 78,452 338,884 0.13 0.12 0.18 1,945

1.16 1 2 228,926 108,481 525,164 0.04 0.04 0.08 1,945

2

8.11 7 14

1

d

0.22 0.22 0.29 1,939

126,367 65,437 300,943

1.27 1 2

5.08 4 8

3

0.33 0.32 0.43 1,946

102,830 53,753 222,513

5

0.47 0.47 0.61 1,943

135,512 40,072 194,552

1.40 1 2

6 3.37 3 6

1.59 1 3

2.78 2 5

7

0.65 0.65 0.93 1,939

81,310 37,092 176,039

0.85 0.88 1 1,946

1 1 1 5,825

0.57 0.54 1 19,428

99,455 37,480 222,990

1.33 1 2

3.81 3 8

All

ip t

33,060 15,491 75,952

1.28 1 2

1.32 1 2

8-10

cr

73,904 29,544 147,723

us

1.46 1 3

an

3.90 3 7

M

1.36 1 2

4.40 4 8

4

Deciles of sticky SR investors’ average SR fund portfolio weight

168,178

24,308 8,573 55,725

1 1 1

One fund

Conventional investors

This table presents summary statistics for sticky SR individual investors’ equity mutual fund portfolio characteristics by deciles of their average SR fund portfolio weight. Since a little more than 30% of sticky SR investors hold only SR funds (i.e., with SR fund portfolio weight of 100%), deciles 8-10 are presented in the same column.

Table V Sticky SR individual investors’ equity mutual fund portfolios

Ac ce p

Table VI Aggregated flows of conventional and SR funds and past returns

(1)

(2)

(0.137) (0.103) (0.055) (0.366) (0.262)

(0.276) (0.011)

M

−0.683∗∗ −0.096∗∗∗

SE

te

Ac ce p

Year FE Fund FE R2 Observations

2.201∗∗∗

(0.192)

All funds

Coef.

SE

cr

1.375∗∗∗ 0.211∗∗ −0.086 −0.210 0.143

Coef. 1.353∗∗∗ 0.226∗∗

(0.138) (0.104)

−0.123∗∗∗ −0.378∗ −0.197

(0.034) (0.194) (0.226)

−0.627∗∗ −0.095∗∗∗

(0.277) (0.011)

1.531∗∗∗ −0.331∗∗∗

(0.279) (0.011)

(0.190)

0.048∗ 0.030 0.037 0.027 0.029 0.055∗∗ 0.124∗∗∗ 0.270∗∗∗ 5.271∗∗∗

(0.029) (0.029) (0.028) (0.027) (0.025) (0.026) (0.027) (0.032) (0.162)

an

SE

vs. C

d

Fund performance P ositivej,t × Rj,t N egativej,t × Rj,t SRj P ositivej,t × Rj,t−1 × SRj N egativej,t × Rj,t−1 × SRj SR funds with more sticky SR investors Sticky SRj,t Sticky SRj,t × P ositivej,t × Rj,t Sticky SRj,t × N egativej,t × Rj,t Other fund characteristics Fund return volatility j,t ln(Net asset valuej,t−1 ) Fund return rankj,t Decile 1 Decile 2 Decile 3 Decile 4 Decile 7 Decile 8 Decile 9 Decile 10 Constant

Coef.

SR

(3)

us

All funds

Sticky

ip t

This table presents the results from the analyses of the relation between aggregated flows and past positive and negative returns on conventional and SR funds. Models 1 and 2 represent the results from estimating Equation (3); these models differ with respect to the sample of funds. In Model 3, the independent variables are the deciles of fund returns. The standard errors (reported in parentheses) are clustered by funds. *, **, *** indicate statistical significance at the 10%, 5% and 1%, respectively. See Table A1 for the variable definitions.

2.171∗∗∗

Yes

Yes

0.431 1,026

0.682 983

Yes Yes 0.431 1,026

27 Page 27 of 33

28

Page 28 of 33

Year FE Individual FE R2 Number of investors Observations

Fund performance P ositivej,t × Rj,t N egativej,t × Rj,t SRj P ositivej,t × Rt−1 × SRj N egativej,t × Rt−1 × SRj Other fund characteristics Fund return volatility j,t M arket Betaj,t SM Bj,t HM Lj,t M omentumj,t Socio-demographic characteristics Children Single Economics degree High school Bachelor or Master Ph.D. ln(Annual salary) ln(Net wealth) Female Born 1973 Constant (0.033) (0.007) (0.004) (0.002) (0.004) (0.003) (0.003) (0.004) (0.005) (0.006) (0.010) (0.001) (0.014) (0.002) (0.003) (0.037)

0.071∗∗ 0.015∗∗ 0.012∗∗∗ 0.049∗∗∗ 0.049∗∗∗ −0.018∗∗∗ 0.003 0.001 −0.028∗∗∗ −0.047∗∗∗ −0.035∗∗∗ 0.010∗∗∗ −0.065∗∗∗ 0.024∗∗∗ −0.009∗∗∗ 0.314∗∗∗

947,862

0.010

Yes

(0.018) (0.012) (0.007) (0.064) (0.047)

SE

0.257∗∗∗ 0.013 0.002 0.067 0.035

Coef.

All investors

(1)

d

(0.076)

−0.764∗∗∗ Yes Yes 0.021 91,601 947,862

(0.004) (0.005) (0.019) (0.041) (0.042) (0.046) (0.001) (0.025)

Coef.

Yes Yes 0.019 3,958 71,090

0.030

(0.301)

(4)

Yes Yes 0.022 4,474 70,267

−0.434

cr

−0.026 0.014 0.006 −0.372∗∗ −0.266 −0.286 0.026∗∗∗ 0.325∗∗∗

0.071 0.038∗∗ 0.028∗∗∗ 0.068∗∗∗ 0.065∗∗∗

0.331∗∗∗ 0.122∗∗∗ −0.106∗∗∗ 0.179 −0.268∗∗

Coef.

(0.297)

ip t

(0.017) (0.020) (0.073) (0.181) (0.186) (0.200) (0.005) (0.090)

(0.107) (0.019) (0.011) (0.007) (0.011)

(0.042) (0.045) (0.018) (0.138) (0.125)

SE

Non-sticky SR inv.

us

(0.011) (0.015) (0.052) (0.235) (0.238) (0.243) (0.003) (0.074)

(0.086) (0.015) (0.009) (0.006) (0.009)

an −0.028∗∗ 0.006 0.073 −0.102 −0.137 −0.114 0.010∗∗∗ 0.046

0.833∗∗∗ −0.061∗∗∗ 0.044∗∗∗ 0.037∗∗∗ 0.046∗∗∗

(0.035) (0.035) (0.009) (0.068) (0.064)

SE

Sticky SR inv.

(3)

0.151∗∗∗ 0.150∗∗∗ 0.032∗∗∗ 0.218∗∗∗ −0.122∗

M

(0.028) (0.005) (0.003) (0.002) (0.003)

(0.011) (0.011) (0.008) (0.058) (0.053)

SE

−0.027∗∗∗ 0.030∗∗∗ −0.022 −0.007 −0.003 0.039 0.015∗∗∗ 0.335∗∗∗

0.262∗∗∗ 0.029∗∗∗ 0.039∗∗∗ 0.057∗∗∗ 0.070∗∗∗

te

0.333∗∗∗ 0.012 −0.008 0.153∗∗∗ −0.036

Coef.

All investors

(2)

Yes Yes 0.019 8,432 141,357

−0.198

−0.031∗∗∗ 0.015 0.024 −0.340∗∗ −0.311∗∗ −0.316∗∗ 0.019∗∗∗ 0.223∗∗∗

0.419∗∗∗ −0.005 0.030∗∗∗ 0.056∗∗∗ 0.052∗∗∗

0.257∗∗∗ 0.129∗∗∗ −0.023∗∗ 0.237∗∗∗ −0.163∗∗∗

Coef.

(0.206)

(0.010) (0.013) (0.045) (0.137) (0.139) (0.146) (0.003) (0.059)

(0.070) (0.012) (0.007) (0.005) (0.007)

(0.028) (0.029) (0.009) (0.067) (0.063)

SE

SR investors

(5)

This table presents the results from the analyses of the relation between individual investors’ flows and past positive and negative returns on conventional and SR funds. Models 1 to 5 represent the results from estimating a version of Equation (1) with Investor fund flowi,j,t as the dependent variable; these models differ with respect to the sample of investors and the individual-specific fixed effects. The standard errors (reported in parentheses) are clustered by individuals. *, **, *** indicate statistical significance at the 10%, 5% and 1%, respectively. See Table A1 for the variable definitions.

Ac ce p

Table VII Individual investors’ fund flows and past fund returns

29

Page 29 of 33

Controls Year FE Individual FE R2 Number of investors Observations

Investor f und f lowj,t−1 Investor f und f lowj,t−1 ×SRj P ositivej,t ×Rj,t N egativej,t ×Rj,t SRj P ositivej,t ×Rt−1 ×SRj N egativej,t ×Rt−1 ×SRj

661,510

0.006

Yes Yes

0.008 0.012 0.112∗∗∗ −0.068∗∗∗ −0.009 0.016 −0.029

∗∗∗

Coef. (0.002) (0.011) (0.017) (0.012) (0.006) (0.058) (0.049)

SE

All investors

(1)

∗∗∗

Yes Yes Yes 0.022 82,225 661,510

−0.086 0.012∗ 0.200∗∗∗ −0.085∗∗∗ −0.009 0.043 −0.026

Coef.

(3)

Yes Yes Yes 0.018 3,956 53,204

(4)

cr

Yes Yes Yes 0.030 3,863 42,689

−0.021 0.058∗∗∗ 0.271∗∗∗ −0.029 −0.146∗∗∗ −0.278∗∗ 0.170

∗∗∗

Coef.

Yes Yes Yes 0.018 7,819 95,893



Coef.

(0.003) (0.008) (0.027) (0.030) (0.009) (0.059) (0.067)

SE

SR investors

(5)

−0.007 −0.062∗∗∗ 0.214∗∗∗ 0.000 −0.013 0.096 −0.168∗∗

ip t

(0.005) (0.015) (0.043) (0.050) (0.019) (0.136) (0.163)

SE

Non-sticky SR inv.

us

(0.005) (0.009) (0.033) (0.036) (0.009) (0.059) (0.068)

SE

an

0.019 −0.132∗∗∗ 0.140∗∗∗ 0.023 0.060∗∗∗ 0.101∗ −0.118∗

Coef. ∗∗∗

Sticky SR inv.

M

(0.001) (0.007) (0.011) (0.012) (0.008) (0.052) (0.058)

SE

All investors

(2)

d

te

This table presents the results from the analyses of the relation between individual investors’ current fund flows, their past fund flows and past positive and negative returns on conventional and SR funds. Models 1 to 5 represent the results from estimating a version of Equation (1) with Investor fund flowi,j,t as the dependent variable and Investor fund flowi,j,t−1 as well as its interaction with SRj,t as independent variables; these models differ with respect to the sample of investors and the individual-specific fixed effects. The standard errors (reported in parentheses) are clustered by individuals. *, **, *** indicate statistical significance at the 10%, 5% and 1%, respectively. See Table A1 for the variable definitions.

Table VIII Individual investors’ current fund flows, their past fund flows and past fund returns

Ac ce p

Predictive Margins of SR with 95% CIs

-.18 -.16 -.14 -.12 -.1 -.08 -.06 -.04 -.02 Negative returns SR=0

0

0

.02

.08 .1 .12 Positive returns SR=0

.14

.16

.18

.2

SR=1

cr

Pr_(Selling) .24 .25 .26 .27 .28

Pr_(Selling) .23.24.25.26.27.28

Predictive Margins of SR with 95% CIs

-.18 -.16 -.14 -.12 -.1 -.08 -.06 -.04 -.02 Negative returns

0

0

SR=1

.02

.04

.06

.08 .1 .12 Positive returns SR=0

.14

.16

.18

.2

SR=1

an

SR=0

.06

SR=1

Predictive Margins of SR with 95% CIs

-.2

.04

us

-.2

ip t

.3

.3

Pr_(Buying) .32 .34 .36

Pr_(Buying) .32 .34 .36

Predictive Margins of SR with 95% CIs

Ac ce p

te

d

M

Figure 1. The marginal effects of the SR attribute on the past positive and past negative returns

30 Page 30 of 33

Appendix

Table A1 Variable Definition Definition

ip t

Variable

1 if fund j is an SR fund, 0 otherwise.

Sticky SRj,t

1 if fund j has a higher share of sticky SR investors, 0 otherwise.

P ositivej,t

1 if the market-adjusted return of fund j is positive in year t, 0 otherwise.

N egativej,t

1 if the market-adjusted return of fund j is negative in year t, 0 otherwise. √ Annualized standard deviation of fund j during year t, calculated as 250 × σdaily

us

Fund return volatility j,t

cr

SRj

Total value of a mutual fund calculated from position in the fund of individual in the sample.

Fund return rankj,t

A categorical variable with values range from 1 to 10, the category is assigned by the decile of the return on fund j in year t.

an

ln(Net asset valuej,t−1 )

Female

Dummy variable equal to 1 if individuals are female, 0 otherwise.

Single

1 if individuals are married or co-habit, 0 otherwise. 1 if individuals have children, 0 otherwise.

Born 1973

1 if individuals are born in 1973, 0 otherwise.

Educational attainment

Categorical variable with a unique value for the highest level of education for each individual: “Compulsory education” (0), “High school” (1), “Bachelor or master” (2) and “Ph.D.” (3).

Economics degree

1 if individuals complete a higher education in economics discipline, 0 otherwise.

Salary

Annual income before tax from wage (labor). Millions of SEK.

d

Net wealth calculated from cash at bank, investment in funds, bonds, stocks, real estate wealth, minus the total value of liabilities. Millions of SEK.

Number of equity mutual funds in an individual’ portfolio.

Ac ce p

Number of funds in portfolio

te

Net wealth

M

Children

31 Page 31 of 33

32

Page 32 of 33

Controls Year FE Individual FE R2 Number of investors Observations

Bottom decile of Rj,t ×SRj

Top decile of Rj,t ×SRj

Bottom decile of Rj,t

Top decile of Rj,t

N egativej,t × Rj,t × SRj

P ositivej,t × Rj,t × SRj

SRj

N egativej,t × Rj,t

Fund performance P ositivej,t × Rj,t

764,540

0.022

Yes Yes

0.263∗∗∗ (0.011) −0.049∗∗∗ (0.010) 0.010 (0.007) −0.043 (0.034) 0.238∗∗∗ (0.038) −0.030∗∗∗ (0.004) −0.039∗∗∗ (0.003) 0.027 (0.051) 0.029∗ (0.017)

(1)

(2)

Yes Yes Yes 0.008 88,419 764,540

0.168∗∗∗ (0.009) −0.013∗ (0.008) 0.016∗∗∗ (0.004) 0.004 (0.032) 0.100∗∗∗ (0.031) −0.009∗∗∗ (0.003) −0.012∗∗∗ (0.002) 0.007 (0.036) −0.005 (0.015)

All investors (3)

108,970

0.022

Yes Yes

d Yes Yes Yes 0.006 7,858 108,970

676,107

0.045

Yes Yes

us

Yes Yes Yes 0.035 82,300 676,107

cr

101,995

0.069

Yes Yes

(8)

Yes Yes Yes 0.048 7,576 101,995

−0.023 (0.021) −0.004 (0.021) −0.023∗∗∗ (0.005) 0.025 (0.039) 0.074∗ (0.039) 0.018∗∗ (0.008) 0.006 (0.006) −0.009 (0.039) 0.013 (0.017)

SR investors

0.106∗∗∗ (0.028) 0.035 (0.026) −0.058∗∗∗ (0.007) 0.096∗∗ (0.046) 0.077∗ (0.043) 0.012 (0.010) 0.031∗∗∗ (0.007) −0.074 (0.048) 0.041∗∗ (0.017)

(7)

ip t

Selli,j,t

−0.030∗∗∗ (0.009) −0.007 (0.008) −0.022∗∗∗ (0.005) −0.000 (0.034) 0.045 (0.033) 0.018∗∗∗ (0.003) −0.002 (0.002) 0.047 (0.036) 0.019 (0.016)

( 6)

All investors

0.123∗∗∗ (0.012) 0.032∗∗∗ (0.010) −0.044∗∗∗ (0.006) 0.065 (0.041) 0.046 (0.037) 0.011∗∗∗ (0.004) 0.021∗∗∗ (0.003) 0.024 (0.050) 0.057∗∗∗ (0.017)

(5)

an

M

(4) 0.204∗∗∗ (0.023) 0.065∗∗∗ (0.022) 0.014∗∗∗ (0.005) 0.033 (0.038) 0.035 (0.038) −0.018∗∗ (0.008) −0.002 (0.006) −0.014 (0.039) −0.017 (0.017)

SR investors

0.263∗∗∗ (0.028) 0.094∗∗∗ (0.027) 0.052∗∗∗ (0.008) −0.007 (0.040) 0.134∗∗∗ (0.046) −0.017∗ (0.010) −0.003 (0.008) −0.047 (0.051) −0.012 (0.018)

te

Buyi,j,t

This table presents the results from the analyses of the relation between individual investors’ buying and selling decisions and extreme past positive and negative returns on conventional and SR funds. Models 1-8 represent the results from estimating Equation (1) with indicator variables for extreme past fund returns and their interaction with SRj ; these models differ with respect to the sample of investors and the individual-specific fixed effects. The standard errors (reported in parentheses) are clustered by individual. *, **, *** indicate statistical significance at the 10%, 5% and 1%, respectively. See Table A1 for the variable definitions.

Ac ce p

Table A2 Individual investors’ buying and selling decisions and extreme past returns on conventional and SR funds

ip t cr us an

-

M

-

d

-

SR hold more diversified portfolios SR investors are less likely to sell SR than conventional funds as past negative returns decrease. the aggregated flows of SR funds for which the majority of their investors are sticky SR investors are less sensitive to past positive returns. Sticky SR investors' flows (at the individual level) are more sensitive to the past positive returns and are less sensitive to the past negative returns Sticky SR investors are less likely to reinvest in their SR than in their conventional funds.

Ac ce pt e

-

Page 33 of 33