Investment Barriers and Labeling Schemes for Socially Responsible Investments

In order to get a first impression to what extent different investor types face different investment barriers, we start with a bivariate analysis. To this end, we compare the four investor groups with respect to their responses regarding the potential investment barriers.²⁸ Fig. 1 illustrates the differences across the four investor groups with respect to the dummy variables used to capture factors related to information costs.

More than one third of the SRs strongly agreed with the statement “I feel poorly informed about sustainable investments.” as a reason why they do not invest more money in socially responsible investments. However, this share is still small compared to the shares in the other groups, which is confirmed by the corresponding t‑tests. Indeed, more than half of all SKs and considerably more than 60% of the INTs or CONVs, respectively, strongly agree that feeling poorly informed about sustainable investments is a reason why they do not invest (more strongly) in socially responsible investments. The results with respect to ‘Too little knowledge’ are quite similar. However, the share of respondents agreeing to the related statement is smaller for every group and there is a clear increasing trend from about 18% for SRs to about 60% for CONVs, i.e. the share of investors that self-assess their knowledge of SRI as too low is largest among CONVs. The results for ‘No offer by bank’ reveal that slightly more than 40% of the SRs or SKs, respectively, strongly agree that not having received an offer for a socially responsible investment product by their bank yet is a reason for not investing (more strongly) in socially responsible investments. The shares among INTs and CONVs even amount to about 73% and 65%, respectively. This result is impressive given that still approximately 50% of INTs and CONVs state that they consult their bank advisor before they invest in a financial product (see Table 3). In contrast to the variables before, the proportions among SRs and SKs, i.e. respondents that are currently invested in socially responsible investment products, are not significantly different from each other. However, there seems to be a large discrepancy between already invested and not (yet) invested investors. The results for ‘Distrust’ show a somewhat different pattern. Again two groups emerge, but the percentages for those who strongly agree with the statement that providers of socially responsible investment products do not follow the sustainability guidelines they represent in the investment information as reason for not investing (more strongly) in socially responsible investments is similar for future potential investors (i.e. SRs and INTs) and those who are not planning to invest in a socially responsible manner in the future (i.e. SKs and CONVs), respectively.²⁹ The share amounts to about 30% for the former group and about 44% for the latter group, which shows that the share of respondents that do not invest (more strongly) in socially responsible investments because they do not trust the information published by providers is not negligible.

Fig. 2 illustrates the relative frequencies for the variables capturing the perceived financial performance and transaction costs of socially responsible compared to conventional investments. Almost 50% of the SRs and even over 50% of the SKs, INTs, and CONVs, respectively, perceive the return of socially responsible investments to be lower compared to conventional investments. Concerning ‘Higher risk’, the percentage of SRs (19.6%) is significantly lower than the respective percentages of the other three investor groups. With respect to ‘Higher fees’, we see that about one quarter of those who are currently invested in socially responsible investments (i.e. SRs and SKs) think that fees of socially responsible investment are higher than that of conventional investments, whereas around 40% of those who are currently not invested in a socially responsible manner have this opinion. These values show that a considerable share of all investor groups perceive socially responsible investments to perform worse than conventional investments in terms of returns, but simultaneously the majority also does not expect higher risks, which indicates rather rational expectations.³⁰

In summary, the results of the bivariate analyses clearly support our first hypothesis. We see that SR investors have the highest level of knowledge compared to the other three investor groups. Further, a smaller share of SRs strongly agrees that the feeling of being poorly informed about SRI and distrust with respect to the information offered by the providers of socially responsible investments are reasons why they do not invest more in a socially responsible manner. They also perceive the financial performance of socially responsible compared to conventional investment products slightly more positive than the other investors, though they also tend to perceive that socially responsible investments have lower returns than conventional ones. The other end of the spectrum is marked by the CONVs. Compared to the other three investor groups, CONVs have the lowest level of knowledge, second highest level of distrust, and have the second lowest level with respect to ‘Poorly informed’ (after INTs). Furthermore, they exhibit the highest level regarding ‘Higher risk’ and ‘Higher fees’. SKs are a kind of hybrid of the other investor groups as they are similar to SRs regarding ‘No offer by bank’ and ‘Higher Fees’, but more like INTs regarding ‘Too little knowledge’, ‘Lower return’, and ‘Higher risk’, and more like CONVs with respect to ‘Distrust’ and ‘Poorly informed’. INTs most strongly differ from SRs and SKs with respect to ‘No offer by bank’. The feeling of being poorly informed about SRI is also a more important reason for them why they do not invest in a socially responsible manner compared to the former two groups. They are less knowledgeable, especially compared to SRs. Thus, having too little information on SRI seems to be the key barrier preventing them to invest in socially responsible investment products.

As the dependent variable ‘Investor type’ takes four different values on a nominal scale, we apply multinomial logit models (e.g. McFadden 1973) in order to identify the importance of the various potential barriers on the membership in the different investor groups. The parameters are estimated by using the maximum likelihood (ML) method. Table 6 reports the estimated average marginal and discrete probability effects with respect to the four different investor groups.

In line with our expectations, the estimation results reveal a highly statistically significant negative relationship between too little knowledge of SRI and the probability to be an SR. The estimated average discrete probability effect amounts to −13.5 percentage points³¹, which is larger compared to all other estimated discrete probability effects for this group, respectively. In contrast, the estimated probability for belonging to the group of CONVs is 18 percentage points larger if a respondent assesses their knowledge of SRI as too little. This supports our explanation that investors with only little knowledge of SRI face higher information costs in terms of, for example, time spent for processing information on SRI than more sophisticated investors, and hence are less likely to invest socially responsibly. We find no significant effect of ‘Too little knowledge’ for the two remaining groups.

Our expectations are further supported by the finding that an increasing number of information sources has a significantly positive effect on the membership in the group of SRs or INTs, and thus on being a potential future socially responsible investor. In contrast, we see that an increase of the number of self-reported information sources by one decreases the estimated probability for being an SK by approximately 2.6 percentage points and even by approximately 3.8 percentage points for being a CONV.

The perception of feeling poorly informed about SRI as reason for no (further) socially responsible investments is significantly positively related to belonging to the group of INTs and the corresponding estimated probability effect amounts to 11.6 percentage points. We find no evidence that this variable significantly affects the group membership of any other group. However, this result suggests that especially INTs have not reached a degree of information that makes them feel sufficiently comfortable to invest in socially responsible investments.

This interpretation is supported by our finding regarding ‘No offer by bank’ that shows that this variable increases the estimated probability for being an INT by approximately 15.8 percentage points. Thus, especially individual investors from this group are obviously confronted with too high information costs and demand more information as well as support from banks. The pivotal role of banks for investors that have not entered the market yet is further underlined by the significantly negative effects on the membership in the group of SRs and SKs. That is, investors that have once overcome the initial entry hurdle are significantly less affected by this potential barrier. This is partly also reflected by the share of people consulting the bank advisor before they make an investment, which is largest among INTs (see Table 3).

Distrust as a reason for no (further) socially responsible investments seems to play a severe role among SKs, INTs, and CONVs, as ‘Distrust’ significantly positively affects the probability for belonging to a group of investors that does not plan to invest socially responsibly in the future, i.e. SKs and CONVs, and significantly negatively affects the probability to belong to the group of INTs. Hence, distrust is not only a reason why CONVs do not plan to invest in a socially responsible manner, but also an explanation why some individual investors plan to withdraw from SRI.

In contrast to factors that influence the individual information and search costs, we only find weak evidence for an impact of the perceived financial performance and transaction costs on investor group membership. In case of ‘Lower return’ we find no significant effect on the membership to any of these groups. Similarly, ‘Higher fees’ seem to play a minor role, as we only find a weak significant and negative effect on the probability to belong to the group of SKs. Only the perception that socially responsible investments are riskier than conventional investments significantly negatively affects the probability of belonging to the group of SRs, while it has a significant positive effect on the affiliation to the group of CONVs. However, this is in line with previous empirical studies that also find few significant effects of these variable in multivariate regression approaches (e.g. Nilsson 2008; Wins and Zwergel 2016).

Also in line with previous studies, we only find weak statistical evidence for effects of sociodemographic variables. The results suggest that age positively affects the membership in the group of CONVs and a negative effect on being an INT, females are more likely in the group of SKs, and having a university degree significantly positively affects the affiliation to the group of SRs.³²

The results of the multivariate analysis above support our first hypothesis. In sum, SRs tend to be more sophisticated and better informed about SRI but also about financial products in general. Hence, they face lower information costs than other investor groups. Especially INTs are less likely invested in a socially responsible manner if they feel poorly informed about SRI and if they received no offer by their bank. Therefore, the barriers for INTs investors could potentially be reduced by targeted information disclosure about SRI and direct offers by banks. Further, SKs and CONVs seem to distrust providers of socially responsible investment products, which in turn prevents them from (future) socially responsible investments. However, both investor groups seem to be also generally worse informed about financial investment products, which leads to the idea that also trustworthy information disclosure about SRI might enhance their willingness to invest in a socially responsible manner.

For the analysis of the relevance of the attributes, and especially of the sustainability certificate and transparency logo, for the choice among the four equity funds in the SC experiment, we refer to multinomial discrete choice models that are based on random utility maximization theory (e.g. McFadden 1973). Accordingly, we assume that respondent i (i = 1, …, 801) chooses an equity fund j (j = 1, …, 4) in a specific choice set m (m = 1, …, 8) if the utility for this alternative is the largest among the utilities for all four alternatives. The utility of respondent i for equity fund j in choice set m is: \(U_{\mathrm{ijm}}=\beta _{i}` x_{\mathrm{ijm}}+\varepsilon _{\mathrm{ijm}}\).

The vector x_ijm = (x_ijm1, …, x_ijmK)’ comprises the explanatory variables that are based on the attributes and individual characteristics. The vector β_i = (β_i1, ..., β_iK)’ contains the unknown parameters that have to be estimated. The error terms ε_ijm summarize all unobserved factors for the choice of an equity fund. In the following analyses, we specify several models with a different number of explanatory variables, i.e. the dimension K of the parameter vectors varies across different model specifications. For example, we begin with basic model specifications that only include explanatory variables for the different attributes (see Table 7). We further consider model specifications that include the interactions terms, which we defined in Sect. 3.2., in order to analyze to what extent the preferences for a sustainability certificate or a transparency logo vary across investor groups (see Table 8) or potential investment barriers (Table 9). We further do not include alternative-specific constants as it is common in econometric analyses with data from unlabeled SC experiments (e.g. Goett et al. 2000; Hensher et al. 2005).

As multinomial logit models are mostly not appropriate for the analysis of SC experiments (e.g Hoyos 2010), we consider mixed logit models (e.g. Hensher and Greene 2003), which still assume independently and standard (type 1) extreme value distributed error terms ε_ijm, but are not based on the restrictive IIA assumption and allow for taste heterogeneity among the participants. Mixed logit models assume that the (random) parameters β_ik (i = 1, …, 801) of the explanatory variables are continuously distributed across i (e.g. Greene 2012) and we further assume that they are normally distributed. Thus, in contrast to the application of multinomial logit models, where only one parameter per explanatory variable is estimated, this approach leads to the estimation of the mean and the additional estimation of the standard deviation for each random parameter. In contrast, it is common practice to assume that the parameters of interaction terms are fixed (e.g. Goett et al. 2000; Hensher et al. 2005) and thus only one parameter per interaction term is estimated, as in the case of multinomial probit models. We use the Stata command “mixlogit” written by Hole (2007) for the simulated maximum likelihood (SML) estimation (e.g. Train 2009) of the parameters.³³

Table 7 reports the basic SML estimation results in mixed logit models for the choice among four equity funds, i.e. we only include the explanatory variables for the attributes and no interactions terms. The estimation results are based on five different subsamples. The estimation results in column two are based on all 801 respondents (and thus 801 × 8 = 6408 observations), while the results in the other columns are only based on the decisions of SRs (i.e. 109 respondents, 872 observations), SKs (i.e. 83 respondents, 664 observations), INTs (i.e. 248 respondents, 1984 observations), and CONVs (i.e. 361 respondents, 2888 observations), respectively. The upper part of the table reports the estimated means of the random parameters and the lower part reports the estimated standard deviations of the random parameters. The estimated standard deviations of the random parameters must not be confused with the estimated (robust) standard errors of the estimated means, which are reported in parentheses.³⁴ Almost all estimated means and standard deviations of the random parameters are significantly different from zero.³⁵ While the former result shows that almost all attributes significantly affect the investors’ choices among the four equity funds, the latter result indicates high unobserved heterogeneity among the respondents, and thus also within every investor group.

The three financial attributes have the expected significant effects on the choice among the four equity funds, i.e. the net return in the last year and the average annual net return in the last five years have significant positive effects, whereas the value of the subscription fee has a significant negative effect. The higher estimated means of random parameters for the average return in the past five years than for the net return of the last year indicate that the respondents put a stronger focus on the mid-term than on the short-term financial performance of equity funds.

With respect to the explanatory variables of main interest, Table 7 reveals for all investor groups negatively estimated means of the random parameters for ‘Consideration of sustainability criteria without certificate’ as well as ‘Conventional fund’. On average, this indicates that individual investors have lower stated preferences for socially responsible equity funds without sustainability certificate or conventional equity funds than for socially responsible equity funds with sustainability certificate. The inclusion of a transparency logo has a highly significant positive effect on the individuals’ equity fund choice, i.e. individual investors have higher stated preferences for an equity fund with a transparency logo than for an equity fund without one. This result again applies to all investor groups.

As in the case of multinomial logit models (e.g. Swait and Louviere 1993), it is not possible to directly compare the size of the estimated means of the random parameters across different model specifications or subsamples due to different scale parameters. One solution would be the estimation of mean willingness to pay for sustainability certificates or transparency logos. This, however, is beyond the scope of this paper as we only aim to analyze whether the different investor groups have different stated preferences towards these attributes and whether certificates and/or labels could overcome certain investment barriers.³⁶ Therefore, we use the interaction terms that we defined in Sect. 3.2, which allow us to compare stated preferences across different investor groups as well as potential investment barriers.

Table 8 reports the SML estimation results in the mixed logit model that includes the explanatory variables for the three financial attributes, the consideration of sustainability criteria without certificate, the consideration of sustainability criteria with certificate, and the transparency logo as well as the six interaction terms of the latter two variables and the dummy variables indicating SRs, SKs, and INTs, respectively. For fixed parameters (i.e. the interaction terms) only one value is estimated, whereas for random parameters (i.e. for the variables of the attributes) estimates for the mean and for the standard deviation are reported.

Table 8 shows that the stated preferences for certified socially responsible equity funds (compared to conventional equity funds) and transparency logos (compared to equity funds that have no transparency logos) vary across investor groups. While we still see that the consideration of sustainability criteria with certificate as well as transparency logos positively affect the choice among the four equity funds for all groups, SRs and INTs have significantly higher stated preferences for these attributes (compared to conventional equity funds or equity funds with no transparency logo, respectively) than CONVs. We only find weak empirical evidence that SKs prefer the consideration of sustainability criteria with certificate over conventional equity funds more than CONVs, but no significant difference with respect to transparency logos. Hence, we see that especially potential future socially responsible investors have significantly positive preferences than CONVs for equity funds with these labels compared to equity funds without these labels. This shows that not only investors that are already active in SRI, i.e. SRs, but also those who have not managed to enter the market yet have positive relative preferences for both types of labels.

To summarize the results reported in Table 7 and 8, we find that both kinds of labels positively influence the investors’ investment decisions. Although this result generally holds for all groups, there is obviously heterogeneity across investor groups and we find the highest positive stated preferences for transparency labels as well as for sustainability certificates among SRs and INTs. Hence, particularly investors who plan to invest in socially responsible investments in the future show interest for these labels. This finding is in line with the previous result that these two groups tend to use more information sources before they make an investment. Hence, it seems to be natural that these groups are also more interested in information provided via labels or certificates, which in turn decrease participation costs and therefore increase the likelihood for SRI. Nevertheless, the results suggest that labels could also provide an opportunity to win SKs and CONVs for SRI as they prefer equity funds with these labels over sustainable funds without label and even conventional investments. Thus, in sum, we find empirical support for our second hypothesis.

In a last step, we directly analyze whether investors that state to face certain investment barriers have higher stated preferences for equity funds with labels than their counterparts. Therefore, we consider model specifications that include the same attribute variables as in Table 8 and add the 22 interactions terms between the dummy variables for the two labels and the explanatory variables described in Table 2, as defined in Sect. 3.2.

Table 9 reports the corresponding SML estimation results in mixed logit models for the five different subsamples. Due to missing values, the total number of respondents dropped from 801 to 530 respondents (i.e. from 6408 to 4240 observations), and particularly from 109 to 91 respondents among SRs, from 83 to 60 respondents among SRs, from 248 to 194 respondents among INTs, and from 361 to 185 respondents among CONVs. As in Table 7, the upper part reports the estimated means of the random parameters and the lower part the corresponding estimated standard deviations. Table 9 reveals again that almost all standard deviations of the random parameters are significantly different from zero, which indicates the presence of unobserved heterogeneity and supports the application of mixed logit models instead of conditional logit models. We further see that the general results regarding the attributes, which are not interacted with other explanatory variables, remain qualitatively the same. That is, the three financial attributes as well as the consideration of sustainability criteria without certificate significantly affect the choice among the four equity funds. In contrast, the estimated means of the random parameters for the consideration of sustainability criteria with certificate as well as for transparency logo are not significantly different from zero in all five (sub-)samples. As the interpretation for these two variables refers to the case that all interaction terms are zero, this result suggests that some of the 11 explanatory variables can explain the estimated strong preferences for sustainability certificates or transparency logos.

The second column in Table 9 further reveals that too little knowledge of SRI has a significant negative effect on the stated preferences for sustainability certificates³⁷, which is obviously driven by the stated preferences for sustainability certificates of SRs and CONVs (see column three and six). We find no significant effects on the stated preferences of SKs or INTs for sustainability certificates. Further, we find no significant effect of too little knowledge of SRI on the stated preferences for transparency logos for any investor group. Our results suggest that a certain degree of knowledge of SRI is required in order to make sustainability certificates work. However, the stated preferences for transparency logos seem not to be affected.

This result is supported by the findings with respect to the stated number of information sources used by the respondents before they make investment decisions. For SRs and CONVs, we again see that the stated number of information sources is significantly positively related to the stated preferences for sustainability certificates. With one exception, we find no further significant effects on the stated preferences for sustainability certificates or transparency logos, respectively. Only for CONVs, the stated number of information sources also have a (weak) significant positive effect on the stated preferences for transparency logos. This suggests that financially literate respondents from the groups of SRs and CONVs who already use several information sources appreciate a sustainability certificate, i.e. a further information source which particularly confirms that an equity fund indeed considers sustainability criteria. We might find no similar significant effects with respect to transparency logos as investors who use several information sources already might perceive their degree of information about the investment strategy as sufficient. It seems as if they just demand a confirmation for issues they already know.

With respect to the feeling of being poorly informed on SRI as reason why the respondents do not invest (more strongly) in a socially responsible manner, we only find a highly significant positive effect on the stated preferences of CONVs for sustainability certificates. However, this result suggests that sustainability certificates might especially decrease the information costs for this specific subgroup of CONVs. Further, this again suggests that sustainability certificates might even help to activate individual investors who actually do not want to invest in socially responsible investments, as already discussed above.

The results regarding ‘No offer by bank’ show that SRs and INTs who indicate that they do not invest (more strongly) in socially responsible investments as their bank has not offered socially responsible investment products to them yet have significantly higher stated preferences for sustainability certificates. In all other cases, we find no further significant effects on the stated preferences for sustainability certificates or transparency logos. However, this result indicates that sustainability certificates could especially help individual investors who assessed themselves as future socially responsible investors. Given that particularly INTs represent the largest share of respondents who consult their bank advisor before they make investments (see Table 3) and who strongly agree that their bank has not offered socially responsible investments to them is a reason why they do not invest socially responsibly (see Fig. 1), sustainability certificates could especially lower their entry costs and serve as substitute for banks who do not totally fulfill their role as intermediaries. That is, we again find support for the second hypothesis.

However, we also see that not only a sufficient degree of financial sophistication, but also a certain degree of trust is necessary to make sustainability certificates work. Table 9 reveals that individual investors, who indicate that they do not invest (more strongly) in socially responsible investments because they do not trust that providers of socially responsible investment products indeed follow their own sustainability guidelines, have significantly lower stated preferences for sustainability certificates. These results are driven by individual investors who have not invested in a socially responsible manner yet, i.e. INTs and CONVs. Hence, the results suggest that a sustainability certificate works worse for individual investors with low levels of trust, although it is the intention of sustainability certificates to address particularly this issue. In contrast, it is intuitively comprehensible that we find no significant effect of this variable on the stated preferences for transparency logos, because this logo only indicates that extensive information on each investment strategy was published, but it does not confirm that the providers follow their own sustainability strategies. Thus, our results suggest that distrust is a crucial barrier and that just the introduction of a label might not be sufficient to overcome this barrier. As mentioned by previous studies in other contexts (e.g. Bonroy and Constantatos 2014), a label itself as well as its provider need to be perceived as credible by the individual investors. This depends on several factors, such as the level of awareness or reputation of the provider, and it could require some time until a label is established. In this respect, future research could analyze on the basis of fund flows or experimental data whether the provider of a label has an effect on the demand of investors.

We also find some evidence that the perceived relative financial performance or transaction costs of socially responsible compared to conventional investments affect the stated preferences for sustainability certificates or transparency logos, respectively. For SRs, we find that the perception of higher risks of SRI compared to conventional investing is related to significantly lower stated preferences for sustainability certificates. Similarly, we find for INTs that the perception of higher fees of SRI compared to conventional investing is associated with (weakly) significantly lower stated preferences for sustainability certificates. We also see for INTs and CONVs that the perception of relatively lower returns for SRI compared to conventional investing has a significantly negative effect on the stated preferences for transparency logos. Thus, these results suggest that financial motives play an important role for at least some individual investors. Given the results of the meta analyses mentioned above (e.g. Friede et al. 2015; Rathner 2013), which indicate that socially responsible funds perform better than or equal to conventional funds in most cases, individual investors seem to need more information not only about the degree of sustainability of socially responsible investments, but also about the relative financial performance of SRI compared to conventional investing in order to make sustainability or transparency labels work.

Finally, we also find some significant effects of socio-demographic control variables on the stated preferences for sustainability certificates or transparency logos. For SKs and INTs, we find that age is significantly positively associated with the stated preferences for sustainability certificates. Age is also significantly positively related to the stated preferences of SRs for transparency logos. For INTs and CONVs, we find that females have significantly positive stated preferences for sustainability certificates and transparency logos, respectively. For CONVs, the results show that the degree of education is negatively related to the stated preferences for transparency logos.