Several empirical studies have been conducted into the performance of SRI funds in the US, including a study by Hamilton et al (1993) of 32 SRI funds and 320 randomly selected non-SRI funds in the US over the period 1981-1990. Since the number of funds increases sharply between 1982 and 1990, two sub-periods are used to test the mutual fund performance with the CAPM model (with the monthly market return for the CAPM being equal to the value-weighted NYSE index). For the first sub-period, 17 SRI funds are compared with 170 non-SRI funds (all established before 1985). The average monthly alpha for the SRI funds is recorded at -0.06%, which is higher than the average monthly alpha of -0.14% for the non-SRI funds. For the second sub-period, 15 SRI funds are compared with 150 non-SRI funds (all established after 1985). The average monthly alpha of these SRI funds is recorded at -0.28%, which is lower than the average monthly
alpha of -0.04% for the corresponding non-SRI funds. Hamilton et al (1993) also conclude that there is no statistically significant difference in the average of the alphas for SRI and non-SRI. Diltz (1995) and Sauer (1997) also fail to find any significant difference in the performance of SRIs and traditional investments.
Statman (2000) tests the performance of 31 SRI funds in the US against 62 non-ethical funds for the period 1990-1998, with both the SRI and non-ethical funds used being of similar sizes. The study records the average expense ratio for the SRI funds as 1.50%, compared with 1.56% for similar non-ethical funds. The S&P 500 Index and Domini 400 Social Index (DSI 400) are used as market indices for the CAPM model, with both indices returning the same results. The average monthly alpha is recorded at -0.42% for the SRI funds and -0.62% for the non-SRI funds. As such, the study concludes that the performance of SRI funds is not significantly different to that of non-SRI funds. Thus, the authors argue that socially responsible mutual funds do not earn statistically significant excess returns, and that the performance of such mutual funds is not statistically different from the performance of conventional mutual funds using a simple regression against a market index.
Due to the inconsistency in results from existing studies into SRI and non-SRI funds, comparing the average performance of these two types of fund provides investors with information of limited use.
Geczy et al (2003) examine diversification costs for investors by comparing optimal portfolios with and without SRI constraints for the period 1963-2001. The optimal portfolios are constructed for mean-variance investors according to short sale constraints. The optimal portfolios are selected from 35 SRI funds and 894 non-SRI funds, with the predictive distribution of fund returns being used for each optimisation. The study proposes that the difference in certainty-equivalent returns between portfolios with and without SRI constraints is an indication of the diversification costs of imposing SRI constraints. The study duly concludes that there are significant financial costs involved when mean-variance investors impose SRI constraints on their portfolios. However, these
costs also depend on whether investors choose to follow asset pricing models or managed funds. For example, for investors that follow the CAPM model rather than investing in a fund, the monthly cost of the SRI constraint is recorded at five basis points. For investors that follow multifactor asset pricing models (such as the three- or four-factor models), the monthly cost of the SRI constraint is found to be at least 30 basis points. For investors that put their faith in the skills of a fund manager, the SRI constraint imposes large costs on investors. Furthermore, for SRI funds void of “sin” stocks, the cost of the SRI constraint is also found to increase by an additional 10 basis points per month.
Geczy et al (2003) identify differences in both the basic characteristics and risk exposures of SRI and non-SRI funds. For example, the authors calculate the average expense ratio for the SRI funds to be 1.33%, which is higher than the 1.10% for non-SRI funds. The average annual turnover for SRI funds is found to be 81.5%, which is lower than the 175.4% for non-SRI funds. Moreover, the monthly abnormal return for the equally- weighted SRI portfolios is recorded at 0.21%, which is higher than the monthly abnormal return of 0.08% for the equally-weighted non-SRI portfolios. However, the difference between these two portfolios is not found to be significant. Since the study uses the four- factor model to test fund performance, the size factor of the SRI portfolio is 0.20, which is higher than the size factor of 0.16 for the non-SRI portfolios. Similar results are recorded for both the book-to-market and momentum factors for the two portfolios.
Each of the studies mentioned above examine the relationship between SRI funds and non-SRI funds. However, these studies do not consider investment screens for SRI funds in detail, which may also affect fund performance.
In contrast, Goldreyer et al (1999) use 29 equity funds, 9 bond funds and 11 balanced funds to test SRI funds performance for the period 1981-1997. A CAPM model is used to test the performance of each fund, with the model duly finding the average annual abnormal return of the 29 SRI equity funds to be -0.49%, while the average alpha of the 20 non-SRI equity funds is 2.78%. The authors also look at whether investment positive screens affect fund performance. They duly find SRI equity funds with positive screens to
have a monthly average abnormal return of -0.11%, which is significantly higher than the return of -0.81% for funds without positive screens. Thus, the authors conclude that SRI fund performance is indeed affected by investment screens. Barnett and Salomon (2006) find that the return of SRI funds decreases when the total number of social screens increases—until a maximum number has been reached (after the maximum number has been reached, the SRI fund returns start to increase again).
Several other studies have also been conducted into the performance of SRI portfolios using firm-level information to design the portfolios. Of these, when Grossman and Sharpe (1986) compare a South African free portfolio with unscreened portfolios, they find no significant difference in the returns of the two portfolios. Likewise, Guerard (1997) and Stone et al (2001) do not find any significant difference in the returns from SRI and non-SRI portfolios when using KLD data.