Two Theoretical views
When they think of it at all, researchers in finance usually think of Socially Responsible Investing (SRI) as a sideshow. This is unfortunate. The theoretical problems raised by SRI are non-trivial, and many touch on longstanding areas of contention in finance and management theory, including value effects, the size effect, arbitrage pricing theory, market efficiency, and stakeholder theory.
SRI also deserves attention because of its prevalence. In 1997 over 800 money managers handled $529 billion in accounts using some type of social screen, according to the Social Investment Forum’s "Report on Responsible Investing Trends in the United States" . A study by Calvert Group and Yankelovich Partners  indicates that many additional investors would consider adopting social screens, if their concerns about relative performance were addressed.
This review of the literature focuses on the performance impact of SRI, with an emphasis on equity portfolios. Although there is no simple definition of SRI, screened portfolios have strong family resemblances, and the generalizations that follow are true for most widely used definitions (1). I have sought to present rigorous studies of SRI performance, but must forewarn the reader that only a few such studies exist. To supplement this material, I will also review relevant studies from the fields of economics, finance and management science. This body of work is unusually broad because social investors operate along a kind of theoretical fault line between financial and management theorists:
Financial theorists who believe in CAPM generally argue that SRI is likely to underperform over the long term because SRI portfolios are subsets of the market portfolio. Under CAPM, the market portfolio will outperform all subsets of itself if markets are efficient. Therefore, the more efficient the market, the more obvious the performance impact of SRI ought to become. This line of analysis is widely-taught in investment textbooks, and we will refer to it (perhaps a little unfairly) as the Markowitz View.
Meanwhile, some management theorists argue that SRI portfolios could outperform market benchmarks, because they incorporate important information not widely understood by the markets. Moskowitz  first raised this possibility, so we will refer to it as the Moskowitz View. (2)
It would seem that any difference of opinion could be resolved quickly by simply examining the performance of screened portfolios. If they were to underperform the market indexes, the Markowitz View would appear to hold. If they were to consistently outperform, we would be inclined to put more weight on the Moskowitz View.
The problem is that we observe a tie.
Most studies suggest that screened portfolios have about the same risk-adjusted returns as their unscreeened counterparts. Dhrymes  finds that "in the aggregate there are not perceptible and consistent differences in the (expected or mean) rates of return between firms which are deemed to be socially responsible." A socially screened index, the Domini Social Index (DSI), has had higher nominal returns than the S&P 500 since its inception in 1990; but DiBartolomeo  finds that, after adjusting for active risk exposures, its performance advantage disappears. Hamilton, Jo, and Statman  find no difference in returns between screened and unscreened mutual funds. Grossman and Sharpe  find that the risk-adjusted performance of South Africa-free portfolios was about the same as for unscreened portfolios, after adjusting for the small-cap effect.
These observations raise doubts about both the Moskowitz and Markowitz views. If Markowitz is right, why don’t we observe consistent underperformance? And if Moskowitz is right, why don’t we observe consistent outperformance? We may summarize the situation as follows:
It is tempting to frame the Markowitz and Moskowitz views as opposing positions in a debate. But, in light of the observed performance of screened portfolios, the two schools are better thought of as partners. The real opposing view, shown in the No/No box, is that there are no unaddressable diversification costs and no information effects. This view has few champions, and, if supported, would be relevant not only to social investors, but to anyone interested in how capital markets process information.
The Markowitz View:
SRI’s Impact on Diversification
SRI portfolios are active portfolios – their screens make them materially different from any widely used benchmark. If the screens exerted random effects, one could appeal to studies of diversification, such as Fisher  or Statman , which suggest that holding some number of randomly selected stocks can diversify away most unsystematic risk. (3)
But the impact of social screens is definitely non-random, and social screens can create uncompensated risk, even in very large portfolios. This is the most widely cited objection to SRI, and remains a significant obstacle to its implementation in accounts governed by ERISA.
Angel and Rivoli  review theoretical objections to SRI, and show how a methodology developed by Merton may be used to estimate the impact of a boycott, which the authors believe to be small in most cases. Grossman and Sharpe  find that a South Africa-free portfolio had a residual standard deviation of 2.51% relative to the New York Stock Exchange for the 1960-1983 time period. Freeman and Winchester  find that "simply removing SRI stocks from the [State of Connecticut] investable universe would have increased uncompensated risk by more than 2.0%," but note that adjustment strategies could offset this problem "substantially."
Many studies find that screened portfolios tend to have smaller average capitalizations, higher price/book ratios, higher P/E ratios, and more favorable "excellence" ratios than their unscreened counterparts. (4) Recognizing that the prospective financial impact of these exposures is open to debate, on a historical basis one might summarize it as shown in Exhibit #1:
Exhibit #1 – Expected Impact of Typical SRI Exposures
Even allowing for wide differences of opinion on the impact of these four factors, it is hard to conclude that the existing body of knowledge favors SRI. No wonder that many practitioners have been skeptical that SRI portfolios can deliver competitive returns.
Besides these factor exposures, screened portfolios usually have significant industry bets. Wilk  and DiBartolomeo  report that these exposures were important drivers of relative performance for the Boston SAFE Index in the 1980s and the DSI in the 1990s.
SRI exposures can often be redressed, but not always. Using a fundamental factor model, one finds that underweighting in the market capitalization factor sometimes persists because there are not enough stocks big enough to substitute for frequently excluded names such as General Electric or Philip Morris. This has led to a search for better adjustment methods.
One promising area of investigation is the use of arbitrage pricing theory, as described by Ross and Roll [1976, 1995]. The APT provides SRI advocates with an appealing theoretical basis because it rejects the CAPM notion that the market portfolio must dominate all others in an efficient market. Under APT, two portfolios are good substitutes for one another, so long as APT factor exposures are equivalent and there are enough members in each portfolio to diversify away idiosyncratic risk. DiBartolomeo  shows that the returns of a portfolio of socially screened stocks, optimized to the S&P 500 using an APT factor model, would have had returns close to those of the S&P 500 (mean monthly difference of -0.01%) in the 1990s.
The same study finds that the DSI has a substantial oil price exposure in the APT framework. This is a particularly troublesome problem for environmentally oriented investors, for whom no oil company may be an acceptable investment. The exposure is caused by an underweighting of energy stocks, but also by overweighting in the retail sector, which tends to underperform when oil prices are rising, due to the impact of high energy costs on the prices of finished goods and disposable income.
This suggests that SRI may interfere with active management strategies. There is some evidence that this interference has occurred. Teper  reports that in the late 1980s risk-adjusted returns for social money managers and mutual funds were 80 basis points lower, on an annualized basis, than those of a backtest of the DSI. Today, as the DSI outperforms the S&P 500, Morningstar’s  Principia for Mutual Funds reports that of the twenty-two SRI mutual funds with three-year track records, nine were below-average relative to their peers, nine were average, and only four were above-average.
If interference with active management does occur, it could be asymmetrical by style. Kurtz  finds that, for a group of growth managers, the performance of screened and unscreened accounts was statistically indistinguishable. (7) Abramson  notes that cyclical companies "may present issues from an environmental perspective," but demonstrates that a value strategy may be successfully executed in a socially screened universe.
The Moskowitz View:
Evidence for and against an SRI Effect
Information Effects in SRI
The argument that socially responsible companies could outperform their less-responsible counterparts has been under discussion at least since Moskowitz  raised the topic. Moskowitz showed that a portfolio of socially screened stocks outperformed the Dow Jones Industrials on a nominal basis, although the time period was short. Subsequent studies showed large variations in relative performance.
Many SRI practitioners believe social screens contributed, at least in part, to the Domini Social Index’s strong performance in the 1990s. In so doing, they turn on its head an argument made by Burton Malkiel in 1971 against the South Africa boycott:
Innovative and growth-minded companies will generally want to market their products worldwide. Thus, there may be a systematic relationship between the expected profitability of an investment and the likelihood that the company will operate in all parts of the world, including southern Africa. To the extent that these corporate characteristics can be expected to affect future returns, altering the composition of the portfolio [by divesting from South Africa] might well reduce the yield of the endowment. (Ellis , p. 604)
Believers respond that innovative and growth-minded managements will be more likely to have superior environmental programs, good employee relations, and corporate citizenship - Waddock and Graves [1997b and 1999] find evidence of this. If so, these firms would be better represented in the Domini Social Index than in the S&P 500, and might thereby improve returns. Like Malkiel, supporters of this view are emboldened by historical performance that supports their position.
Recent developments in stakeholder theory and management science have made the theoretical basis for such a hope more coherent. (8) Friedman  notwithstanding, corporations do not operate in a vacuum, and social responsibility could be one way for them to improve relations with constituencies important to their future success (Prahalad and Hamel ). Unfortunately, the theoretical work is supported by empirical research that can be best characterized as uneven. Many of the studies are intended for an academic audience rather than the financial community, and good SRI data were not available for many companies until the early 1990s.
Most attempts to measure a "social factor" directly have failed. DiBartolomeo  and Kurtz and DiBartolomeo  find that the returns of the DSI are fully explained by its fundamental factor bets. D’Antonio, Johnsen, and Hutton  arrive at the same conclusion when evaluating the performance of a screened bond portfolio. Grossman and Sharpe’s  multifactor work finds a small negative effect to the South Africa screen after accounting for all other factors, but this, too, proved elusive – from 1976 to 1983 the South Africa-free portfolio had a positive (although non-significant) alpha, probably because of the size effect. An event study by Teoh, Welch, and Wazzan  finds that the South Africa boycott had a negligible effect on financial markets.
Luck and Pilotte , however, find that the BARRA fundamental factor model is unable to completely account for the performance of the DSI for the 1990-1993 time period, and that an active return of 9 basis points per month remains unexplained. Luck  replicates this finding for the 1990-1998 time period, and reports that the stock-specific return grew steadily throughout the test period.
For several years, Luck and Pilotte  and Diltz  were the only rigorous risk-adjusted studies to report positive returns to social screens. But since 1997, a series of new studies using different methodologies, data sets, and time periods support the idea that some social screens may be associated with positive abnormal returns.
Guerard  finds that, in some subperiods, social screens improve the backtest results of a proprietary stock selection model with valuation and earnings revision components.
Feldman, Soyka, and Ameer  find that in a CAPM regression in which the assumption of uncorrelated residuals has been relaxed, ICF/Kaiser’s proprietary environmental rating models have significant explanatory power (p < 0.01). Russo  demonstrates a fundamental basis for these findings, finding that superior environmental performance is associated with higher returns on assets, after controlling for variables such as industry, firm size, and capital intensity; although no attempt is made to measure the capital markets’ reaction to this information.
In addition to these examples, many studies in the management sciences associate positive returns with progressive management practices. Exhibit #2 summarizes studies of the stock market impact of screens favored by social investors. In a few cases, works judged to have theoretical significance are included, although they did not explicitly measure stock market impact.
Note that not all findings imply positive returns to social investors. Crystal [1989-1999] shows a weak but measurable relationship between higher executive compensation and superior stock market performance, although the direction of causality is open to debate. A new study by Crystal  argues that companies with CEOs who are well-paid but have minimal stock incentives tend to materially underperform the S&P 500.
It is noteworthy that some of the screens that have been studied least, such as alcohol, gaming, military, and life ethics, are among the most widely used by social investors.
Exhibit #2 – Issue Studies of Interest to Social Investors
A critical approach is warranted in evaluating this literature, particularly with respect to the event studies. McWilliams and Siegel  review 29 recent event studies, many of which deal with corporate social responsibility, and find numerous methodological problems. The authors observe that three studies of corporate social responsibility "claim quite dramatic value effects," but in fact the results are not replicable after adjustments are made for methodological shortcomings. (11) Many of the other studies are contradicted by Dhrymes , who finds that none of the KLD social factors is a consistently good predictor of stock returns.
Moreover, although many issue studies include some form of risk adjustment, most fail to simultaneously adjust for other factors, which we know are likely to be present. The three multifactor studies of the DSI (Luck , Kurtz , and DiBartolomeo ) suggest that its performance is largely replicable by mimicking its factor exposures. (12)
Believers in an SRI effect would reject the reductionism implied by these findings. Their arguments usually take one of the following forms:
Sign of Skillful Management: Several authors argue for the possibility that socially responsible behavior is an indicator of management competence (e.g., Alexander and Bucholtz ). This makes sense in the abstract – good managers should be more likely to excel at social responsibility than their less-clever counterparts. Sandra Waddock and Samuel Graves of Boston College have done numerous studies supporting this idea. Waddock and Graves  demonstrates a positive relationship between social ratings and Business Week’s ratings of board quality. Waddock and Graves [1997b] finds a similar relationship with the Fortune "Most Admired" list. Graves and Waddock  evaluates the social records of companies featued in the book Built to Last, and finds they have superior employee relations and diversity records.
But tautological pitfalls remain – is the company successful because of good management, or is management perceived as being of high quality because of the buoyant stock price? Moreover, social ratings appear to be relatively stable over time, while the history of the Fortune "Most Admired" lists testifies to the transient nature of management reputation.
Result of Prosperity: Social responsibility could be a by-product of financial success. Prosperous businesses have less incentive to cheat and more of a stake in protecting their good names. This alone would justify avoiding criminal behavior, treating employees well, and investing in environmental programs and philanthropy.
Researchers seeking evidence for this argument would need to explain why social responsibility is not simply a coincident indicator, and also why some of the most-excluded companies are notable for generating significant free cash flows, such as Philip Morris and General Electric.
Catalyst for Change: There is some anecdotal evidence (see Petzinger ), that SRI programs, particularly in the environmental area, can stimulate operating efficiencies. Meyers and Nakamura  suggest a theoretical basis for this effect, arguing that increased environmental regulation can lead to greater capital turnover and innovation, offsetting the apparent costs of such policies. Russo  and Feldman  present a theoretical arguments for this at the microeconomic level, and report positive returns to their environmental variables at the income statement level.
Prior to these studies, it appeared that companies in countries with stronger environmental regulations faced higher costs than their less-regulated competitors, but that these costs were not so burdensome as to promote industrial flight (Cropper and Oates ).
Signaling Mechanism: Socially responsible behavior could be interpreted as a signal from management. Suppose a manager knows that her business will have exceptional results over the next three years, resulting in a significant increase in free cash flows with minimal incremental investment. The manager values financial flexibility (per Myers ; Waddock and Graves [1997b] is also relevant), so chooses not to allocate all these funds to dividend increases or share buyback. Instead she seeks to allocate the capital to areas that will improve customer and employee perceptions of the company, but from which it can be reclaimed, if necessary, with minimal fuss. A one-time philanthropic grant or environmental project fills those requirements nicely.
Employee Relations – A Cautionary Tale
The case of employee relations provides an example of the difficulties faced by researchers who wish to demonstrate an SRI effect. There is some theoretical justification for such a project. Weitzman  attributes the success of Japanese companies up to that time to their extensive cash profit sharing and no-layoff policies. Krugman  comments that firms might pay higher wages and offer better benefits in exchange for loyalty from their workforce. These ideas contradict the conventional view that gains to labor necessarily come at a cost to shareholders, first outlined by Hotelling in 1932 (see Abowd ).
One could argue that market participants have focused too much on the Hotelling thesis, overlooking the possibility that progressive employment policies make the pie larger, so that workers’ gains need not come at the expense of shareholders. This should be especially true for companies whose value depends on the production of intellectual property.
Amgen, the biotechnology company, gives every employee stock, offers generous childcare benefits, and provides a company-subsidized cafeteria. Management reports that these policies further the objective of retaining and focusing a skilled work force. (13) Even the former management of Columbia/HCA, hardly free-spenders, made investments in employee relations. They told investors that patients’ perceptions of quality of care were closely related to employee morale. (14)
Recent advances in social research have enabled systematic study of the stock market impact of employee relations policies. One would expect to find significant relationships between these newly available variables and stock returns. Unfortunately several studies have failed to do so, notably Diltz [1995a], which used data from the Council on Economic Priorities, and other unpublished projects using data from Kinder, Lydenberg, Domini & Co. (KLD). Kurtz and Luck  do find that when the investment portfolio is narrowed to a small group of exceptional companies (based on The 100 Best Companies to Work For in America database), an historical performance advantage is observed for the 1985-1999 time period, even after adjusting for beta, industry, and other financial factors. It remains to be seen whether the effect is significant enough that this information could be incorporated into an active management strategy.
It is possible that the market is already correctly pricing employee relations information. This is good news, in the sense that this is what SRI advocates ostensibly want. The bad news is there isn’t much evidence that understanding these issues confers an investment advantage.
These problems are not unique to the area of employee relations. SRI embodies many different disciplines, each with its own unique measurement challenges and subjective inputs. And the risk that the market already understands the relationship, or will come to understand it, is always present.
Markowitz/Moskowitz vs. No Effect
Given the difficulties researchers have had in proving a diversification cost or positive SRI effect, it is worth considering whether the "No Effect" hypothesis deserves closer attention.
Do social screens simply have no impact on performance, after accounting for factor exposures, as DiBartolomeo  suggests? If so, the portfolio manager could simply apply social screens and equalize factor exposures to the benchmark portfolio using a fundamental factor or APT model. Clients would benefit because the screened portfolio would be both consistent with their values and offer a competitive return. This would make compliance with statutory diversification requirements, such as ERISA, relatively easy. It appears that few portfolio managers currently take this approach, however.
It could also be that both Markowitz and Moskowitz are right. The DSI (which, like most SRI portfolios, is not optimized to any broader benchmark) could suffer diversification cost which have been offset by information effects. But if this were the case, SRI practitioners would be faced with a dilemma – information effects will likely be arbitraged away eventually, while diversification costs would be permanent in unadjusted portfolios if CAPM holds.
At present, it appears that portfolio managers wishing to eliminate the active exposures of social screens may do so without incurring material costs. Those who choose to accept active exposures, believing that an SRI effect will more than offset the diversification costs, have not yet shown why this is a compelling investment strategy.
Areas for Further Study
Many other questions about SRI remain unanswered, some of which have significant theoretical implications. In our view, the most important are:
Ruf  finds evidence of this. An important implication of this idea is that to profit from such an effect, an investor would need to focus on companies with below-average social records - the exact opposite of what most social investors currently do.
There are many ways to invest responsibly. In addition to screened portfolios, investors may choose to invest with managers focusing only on those companies with the best social records, or in loan funds paying market rates of interest or below-market rates. All of these options must be balanced against charitable giving, in terms of both financial and social impact. The field needs theoretical and empirical work that will help managers and clients select which, if any, of these options to choose.
With the advent of high-quality social data and the introduction of prizes for good studies, we can expect more analysis of SRI than we have seen in the past. But it is unlikely that there will ever be a final word on the subject. It is in the nature of SRI that it responds to the changing political and social landscape. Companies, too, change. Industries that are today excluded from SRI portfolios may one day be included, altering the currently observed relationships between social screens and conventional investment factors.
As researchers address these questions, the findings should matter not only to practitioners of socially responsible investing but also to all investors.
1. For a definition of social investing and a discussion of the rationale for commonly used screens, see Kinder, Lydenberg, and Domini .
2. See, for example, Zvi Bodie, Alex Kane, and Alan Marcus, Investments, Chicago: Irwin, 1993 and Frank K. Reilly, Investment Analysis and Portfolio Management, Fort Worth: Dryden, 1994. (Both texts are required reading for CFA candidates.) Bodie et al include a brief discussion of socially responsible investing in this context.
3. Fisher suggests that this figure is 30. StatmanŐs marginal utility analysis implies that it is more like 200.
4. For a full discussion of "excellence" ratios, see Clayman .
5. Black  asserts that "just after the small-firm effect was announced, it seems to have vanished." Some analysts, e.g., Reinganum , view the small-cap effect as an outgrowth of the January effect, which Haugen  argues is still operative.
6. Although occasionally cited as evidence that a strategy of buying "excellent" companies could backfire because of regression toward the mean, note that in both of ClaymanŐs studies, a portfolio of "excellent" companies has a positive alpha when returns are compared with the S&P 500.
7. Note that Angel and Rivoli  predict that a boycott would have greater impact on the value of a fast-growing stock than on a slower-growing stock.
8. For a more complete review of the literature of corporate social responsibility, see Griffin and Mahon . Waddock and Graves [1997a, 1997b] also include a review and discussion of this literature.
9. Based on Social Investment Forum [1995 and 1997], conversations with social analysts and money managers.
10. The most common screen in the U.S. until the end of the international boycott in 1993.
11. See Wright et al  and Meznar, Nigh, and Kwok .
12. A common criticism of fundamental factor models is that some of the factors are statistical constructs, and difficult to relate to the real world. Conversely, the chief criticism of social screens is that they are based on the real world, but have little investment relevance. Now we find that there is some correlation between the two. Which is likely to be a better investment tool? Is it easier to predict the trend of a companyŐs environmental record, or the trend of the price/book variable in a multifactor model?
13. According to a conversation with Gordon Binder, CEO of Amgen, June 23, 1997.
14. According to Columbia/HCA Presentation .
15. The author is indebted to Lisa Leff of Trillium Asset Management for this observation.