The independent effects of environmental,social and governance initiatives on the performance of UK firms

Abstract

We investigate the effect of environmental, social and governance factors on the financial performance of UK firms. We examine the three factors separately to disentangle the relation of each with performance. We find no difference in the performance of firms with high or low environmental, social or governance rankings. The firms also do not differ in their systematic risks, book-to-market ratios or momentum exposures. However, high-rated firms are consistently larger. Our findings demonstrate that UK investors can incorporate environmental, social or governance criteria into their investment strategies without incurring any significant cost (or benefit) in terms of risk or return.

Keywords

Best of Sector corporate social performance environment governance social sustainability UK

1. Introduction

Sustainability investing recognizes that firms face environmental, social and governance (ESG) opportunities and risks that may significantly impact firm performance, and consequently affect shareholder value. Sustainability investing therefore requires that ESG factors be incorporated into investment analysis. ESG analysis is performed in conjunction with, not instead of, ‘traditional’ financial analysis in order to identify appropriate assets from which to form investment portfolios.

There has been increasing recognition globally of the need to incorporate ESG factors into investment analysis. Many countries’ stock exchanges now have mandatory ESG disclosure requirements, while other countries have voluntary requirements. For example, Australia, France, India, Malaysia, South Africa, Sweden, Taiwan and Thailand have introduced formal measures on ESG disclosure and stock exchanges in China have issued ESG guidelines for listed companies.¹

Recent mergers and acquisitions activity provides further evidence that ESG analysis is becoming more mainstream, and large data providers such as Bloomberg and Thomson Reuters now provide ESG data.² RiskMetrics recently acquired two leading ESG research firms, Innovest Strategic Value Advisors and KLD Research & Analytics, and MSCI then acquired RiskMetrics. MSCI’s chief executive Henry Fernandez stated that he believes ESG criteria, and not solely traditional measures such as liquidity, should be considered in index formation.³

The United Nations is perhaps the primary champion for ESG incorporation. The United Nations Principles for Responsible Investment (UN PRI) was launched in April 2006 and sets out six principles that signatories commit to. These primarily encompass incorporating ESG issues into investment and business practices. In January 2011 there were 867 UN PRI signatories, the majority of whom were investment managers.⁴ This represents more than US$21 trillion in assets globally, with most of the funds coming from Europe (US$12.5 trillion), followed by North America (US$5.5 trillion) and then Asia (US$1.96 trillion).

Further, the view of the relation between ESG integration and fiduciary duty is changing. Many are now challenging the ‘traditional’ view that the process of incorporating ESG factors into investment practices violates the fiduciary duty to maximize profits. The traditional view assumes that a negative financial impact must arise from integrating ESG factors into investment and business practices (see Moskowitz, 1972). A recent report by the United Nations Environment Programme Finance Initiative (2009, p. 11) challenges this view and states:

Fiduciaries have a duty to consider more actively the adoption of responsible investment strategies…[and] must recognise that integrating ESG issues into investment and ownership processes is part of responsible investment, and is necessary to managing risk and evaluating opportunities for long-term investment.⁵

Firms and investment managers may now need to demonstrate that they actively manage the opportunities and risks attributable to ESG factors as a part of their fiduciary duty (see Richardson, 2007 for a full discussion on ESG integration and fiduciary duty).

The United Kingdom is at the forefront of sustainability investing, with the Carbon Disclosure Project, the Institutional Investors Group on Climate Change and the secretariat of the UN PRI all based in London. In addition, the UK’s Companies Act (2006) requires the director’s report of a quoted company to include information on a range of ESG issues ‘to the extent necessary for an understanding of the development, performance or position of the company’s business’. There are currently 113 UK signatories to the UN PRI, which represents over US$3 trillion in assets under management.⁶ Indeed, more than 30% of the funds under management recorded by the UK’s Investment Management Association are managed by UN PRI signatories.⁷

It is clear, then, that there is large and growing interest in and support for ESG integration worldwide. The natural question is whether ESG factors impact financial performance. Surprisingly, there is a paucity of research into the issue, particularly in the UK. To date there is only one study, Brammer et al. (2006), that investigates the impact of ESG policies on the performance of UK firms.

In this paper we investigate whether there is any difference in the financial performance and risk of UK firms that have high environmental (E), social (S) or governance (G) scores compared with firms with low scores. We examine E, S and G separately in order to disentangle the relation of each with financial performance. Our findings suggest that to date there is neither a financial cost nor benefit from investing in firms with good E, S or G scores. In addition, these firms are not substantially different from firms with poor scores in terms of their systematic risk, book-to-market ratios or prior performance. However, we find that firms with high scores are larger than firms with low scores.

The rest of this paper proceeds as follows. Section 2 provides a review of the prior literature, Section 3 describes the data and Section 4 outlines the methodology. Section 5 provides a discussion of the results and robustness tests. Conclusions are presented in Section 6.

2. Prior literature

A number of studies examine the relation between overall ESG performance and financial performance. Results from prior research are, however, mixed. Some studies find that investing in ESG principles improves financial performance (see, for example, Filbeck et al., 2009; Lo and Sheu, 2007). Lee et al. (2009) find that ESG investment reduces financial performance, and argue that this could indicate a lower cost of equity capital for firms with high ESG rankings.⁸ A third group conclude that there is, in fact, no relation between ESG score and financial performance (Galema et al., 2008; Statman, 2006).

Galema et al. (2008) and Statman and Glushkov (2009) note that ESG score is made up of a number of factors, each of which may have a differing impact on performance. Some initiatives may be value enhancing, while others may be value destroying. Consequently, examining aggregated ESG scores may mask the effect each of the E, S and G sub-components has on performance, and therefore prohibit one from accurately determining the relation between ESG score and financial performance. It could therefore be worthwhile examining E, S and G sub-components separately to determine how each component affects performance. This is the approach taken by the following studies.

Derwall et al. (2005) find that portfolios of high-ranked eco-efficiency stocks significantly outperform low-ranked eco-efficiency stocks. Similarly, Klassen and McLaughlin (1996) and Gilley et al. (2000) find that stock prices rise (fall) when firms release good (bad) news on environmental issues. However, Brammer et al. (2006) do not find evidence of a significant relation between the E component and financial performance, and Galema et al. (2008) find a significant negative relation.

Social initiatives consider the firm’s relationships with its stakeholders, such as its employees and the wider community. Proponents argue that firms that invest in stakeholder engagement and management have a positive image in the community and also may be able to attract superior employees, giving them a competitive advantage (Lado and Wilson, 1994; Turban and Greening, 1997). Herremans et al. (1993) find that firms with better reputations outperform those with poorer reputations and also have lower risk. Filbeck et al. (2009) conduct an event study of the ‘100 Best Corporate Citizens since 2000’ composed by Business Ethics magazine and find these companies consistently outperform. However, Galema et al. (2008) find no relation between employee relations and performance.

In terms of the G component, Gompers et al. (2003) find that the risk-adjusted returns of firms with strong governance is 8.5% higher than firms with poor governance. Cremers and Nair (2005) find that both internal and external governance are important and find a positive relation between the G component and financial performance. In contrast, Statman and Gluskhov (2009) and Core et al. (2006) do not find evidence of a significant relation between the G component and financial performance.

Currently, there is only one study, Brammer et al. (2006), that investigates the link between ESG score and performance in the UK. Using data from the Ethical Investment Research Service (EIRIS) database, the authors do not find evidence of a significant relation. They also find little evidence of any significant difference in the performance of firms with high environment, employment or community scores relative to firms with low scores. Unfortunately, their ESG data is for only one year, 2002. The authors examine the financial performance of firms over the following one-, two- and three-year holding periods, which assumes that the ratings do not change over these periods. In addition, their ESG ratings do not include any G criteria, are quite granular with a 0–3 scale, do not distinguish between general and industry-specific ESG opportunities and risks, and the authors only examine equally weighted portfolios.

We extend Brammer et al.’s (2006) seminal paper and substantially refine their methodology. We examine a longer period, from 2002 to 2008. Our E, S and G ratings are formed using both general and industry-specific analysis and cover more than 300 metrics. Firms are given a more refined score, from 0 to 100. In addition, we value weight portfolios, which more accurately reflects the portfolios held by mainstream, sophisticated, long-term investors (e.g. pension and investment funds).

3. Data

3.1. ESG data

E, S and G ratings data are obtained from Sustainability Asset Management Group GmbH (SAM) for the period from October 2002 to September 2007.⁹ SAM is a proprietary company that specializes in sustainability research and analysis and is perhaps best known for providing the ESG research used to form the Dow Jones Sustainability Indexes. Each year SAM rates more than 1000 corporations from across the world.¹⁰ We select the UK portion of the database, which gives us a sample of 249 firms.

SAM analyses the E, S and G opportunities and risks faced by each company according to general and industry-specific criteria. General criteria are applicable to all industries and include up to 14 categories.¹¹ Industry-specific criteria are designed to recognize that there are specific E, S and G opportunities and risks that a particular industry may face. Industry-specific criteria are very specialized, for example, ‘fleet age’ and ‘sustainable fisheries’, and may apply to only one particular industry. There are more than 300 industry-specific criteria and each industry has an average of 6.4 industry-specific criteria over the sample period. SAM gives each company a score from 0 to 100 across each of the general or industry-specific criteria.

For the first time in any study, we are able to access the raw data for both the general and industry-specific criteria from SAM. This means we are able to examine the impact of general or industry-specific E, S and G factors on financial performance. This will assist in isolating how general (more systematic) or industry-specific (more idiosyncratic) factors contribute to the relation between E, S and G factors and financial performance. This is especially important given the widely held view that ESG analysis should include an industry-specific component.¹²

We form general and industry-specific E, S and G scores by equally weighting the relevant criteria. We also calculate a combined (total) score which equally weights the general and industry-specific E, S or G scores. To summarize, we have nine scores for every firm: for each of the E, S and G dimensions, every firm has a general score, an industry-specific score and a combined score.¹³ This process is illustrated in Figure 1.

Figure 1.

ESG data. This figure shows how the raw environmental (E) social (S) and governance (G) data from Sustainability Asset Management Group GmbH (SAM) are aggregated to give general, industry-specific and combined scores for each firm.

3.2. Financial data

Monthly data on total returns, size and industry classification are extracted from Datastream for all firms in our sample over the period from October 2002 to September 2008. Monthly data on size, industry and returns for all firms comprising the FTSE All-Share Index are extracted from Datastream in order to form our factors, as are the monthly returns of the FTSE All-Share Index, FTSE SmallCap Index, FTSE 100 Index and the three-month UK Treasury Bill.¹⁴

4. Methodology

4.1. Portfolio formation

SAM announces its ESG scores in September every year. We therefore form portfolios in October to avoid a ‘look ahead’ bias. Each year, three portfolios are formed across each of the E, S or G metrics: a ‘high-ranked’ portfolio comprising those stocks with the best performance in that area; a ‘low-ranked’ portfolio comprising those stocks with the worst performance in that area; and a ‘difference’ portfolio, which is a portfolio long in the best scoring stocks and short in the worst. The difference portfolio enables us to easily determine whether there is any significant financial cost or benefit for those firms that invest more heavily in E, S or G initiatives relative to those firms that do not.¹⁵

Recall that our aim is to determine whether there is any financial cost or benefit to firms engaging in E, S or G initiatives. One factor that needs to be taken into account is that some industries, due to the nature of their operations, will exhibit higher E, S or G scores than others. Table 1 shows the mean and standard deviation of E, S and G scores for each of the 20 FTSE industries. It is evident from the table that there are considerable differences in the E, S and G scores across industries. This means that portfolios comprising firms with the very best or worst E, S or G scores may take on significant industry tilts (see also Lee and Faff, 2009). EuroSIF (2008) notes that one of the major participants in the ESG market is occupational pension funds and, as already noted, most of the UN PRI signatories are investment managers. Large institutional investors like these generally prefer to hold diversified portfolios and try to avoid significant portfolio biases, such as industry tilts (Hawley and Williams, 2007). Consequently, to reflect the portfolios these large investors will hold, we try to minimize the industry biases in our portfolios. Specifically, we form high-ranking (low-ranking) portfolios from the 10% of stocks that have the best (worst) E, S or G scores within each of the 20 FTSE All-Share industries. This ensures that every industry is represented in the portfolios and it is generally referred to as a Best of Sector approach. For robustness, we also investigate alternative portfolio formation methodologies that do not take the industry tilts into account.

Table 1.

Sustainability Asset Management Group GmbH (SAM) and FTSE All-Share Index industry representation. This table shows the average (Av.) and standard deviation (Std. dev) statistics for the combined (general plus industry-specific) environment (E), social (S) and governance (G) ratings for each of the FTSE All-Share Index industries. E, S and G ratings are out of a possible maximum rating of 100. The number of firms (# of firms) within each industry is provided for the SAM and the FTSE All-Share Index constituent firms over the sample period. The final two columns correspond to the percentage of firms within each industry relative to the total number of firms from the SAM or FTSE All-Share Index samples. The sample period is 2002–2007.

FTSE industry group	Environment		Social		Governance		# of firms		% of firms
	Av.	Std. dev.	Av.	Std. dev.	Av.	Std. dev.	SAM	FTSE	SAM	FTSE
Automobiles & parts	52.29	44.33	43.50	22.31	20.28	13.33	1	1	0.40	0.16
Banks	49.93	49.72	58.59	16.97	21.25	20.28	10	10	4.02	1.57
Basic resources	62.47	57.54	67.26	27.17	22.68	29.23	8	23	3.21	3.62
Chemicals	61.98	51.58	51.37	26.89	25.00	24.40	4	5	1.61	0.79
Construction & material	31.52	33.55	38.81	19.55	17.15	12.26	7	12	2.81	1.89
Financial services	30.15	31.62	37.54	18.43	15.61	17.15	15	193	6.02	30.39
Food & beverage	45.34	39.81	41.93	19.29	17.90	18.23	9	17	3.61	2.68
Healthcare	60.36	52.12	64.08	31.45	23.58	19.99	8	22	3.21	3.46
Industrial goods & services	37.30	35.14	40.26	22.99	18.22	21.16	42	108	16.87	17.01
Insurance	41.15	40.14	40.38	19.44	16.60	17.55	11	21	4.42	3.31
Media	53.08	42.26	42.50	27.25	26.16	22.30	16	26	6.43	4.09
Oil & gas	59.94	59.81	57.81	24.11	21.06	21.50	8	25	3.21	3.94
Personal & household goods	43.25	35.57	50.25	22.91	19.28	18.40	10	17	4.02	2.68
Real estate	39.69	40.59	46.34	23.46	21.94	20.02	13	30	5.22	4.72
Retail	41.17	37.66	44.48	25.87	20.74	17.99	22	32	8.84	5.04
Technology	28.25	20.98	32.11	15.31	10.53	11.76	15	32	6.02	5.04
Telecommunications	56.94	55.05	49.72	34.91	30.45	21.12	9	9	3.61	1.42
Tobacco	62.59	53.96	54.78	30.18	23.77	21.82	3	3	1.20	0.47
Travel & leisure	42.82	35.61	42.31	23.60	19.76	18.71	24	35	9.64	5.51
Utilities	55.27	51.97	47.24	18.99	21.66	19.53	14	14	5.62	2.20
All industries	44.72	40.94	45.65	25.02	22.08	21.18	249	635	100	100

4.2. Empirical framework

We test the impact of E, S and G factors using a number of widely used performance models. First, we use a one-factor model as follows:

R_{p, t} - R_{f, t} = α_{p} + β_{p} (R_{m, t} - R_{f, t}) + ε_{p, t}

(1)

where R_p,t, R_m,t and R_f,t are the returns for portfolio p, the market portfolio (the value-weighted FTSE All-Share Index) and the risk-free asset (the three-month UK Treasury Bill) in month t respectively.

Prior research has shown that firms with high ESG scores may systematically differ from those with low scores in terms of their size, book-to-market ratios and momentum exposures (see Derwall et al., 2005; Galema et al., 2008). Consequently, we add size, book-to-market and momentum mimicking factors into the equation and examine performance using the four-factor Carhart (1997) model:

R_{p, t} - R_{f, t} = α_{p} + β_{p} (R_{m, t} - R_{f, t}) + s_{p} S M B_{t} + h_{p} H M L_{t} + m_{p} U M D_{t} + ε_{p, t}

(2)

where R_p,t, R_m,t and R_f,t are as above and SMB, HML and UMD are the monthly returns on the mimicking size, book-to-market and momentum factors, respectively.

In line with Elton et al. (1996), we form SMB from the monthly return on the FTSE SmallCap Index minus the monthly return on a large cap index, the FTSE 100. We obtain the UK HML series from the Kenneth R. French data website.¹⁶ To form UMD, each year we rank all firms on the FTSE All-Share Index according to their prior 12-month return and calculate the difference in the returns on portfolios of the top and bottom 30% of firms.¹⁷

The coefficients on alpha from Equations (1) and (2) indicate whether the portfolios outperform or underperform, given their systematic risks. The alphas on the difference portfolios are of particular interest, because they indicate whether companies with high E, S or G scores have better or worse financial performance than companies with low E, S or G scores.

As discussed previously, portfolios with high or low E, S or G rankings are likely to exhibit substantial industry tilts. This is especially true of those portfolios formed purely on E, S or G rankings (i.e. not within industry). In addition, although our Best of Sector approach ensures that each portfolio comprises an equal number of firms in each industry, because we value weight returns, even our high- and low-ranked Best of Sector portfolios may have different industry weightings relative to each other and also the market portfolio. We therefore control for industry biases by including additional industry factors into the four-factor model. We use principal components analysis to form three industry factors (see Derwall et al., 2005). The model is as follows:¹⁸

R_{p, t} - R_{f, t} = α_{p} + β_{p} (R_{m, t} - R_{f, t}) + s_{p} S M B_{t} + h_{p} H M L_{t} + m_{p} U M D + \sum_{i = 1}^{3} i_{i} I P_{p, t} + ε_{i, t}

(3)

where $\sum_{i = 1}^{3} i_{i} I P_{p, t}$ are the three industry factors and all other variables are as above.

For robustness, we also add an idiosyncratic risk mimicking factor to Equation (3) to examine if there are differences in the idiosyncratic risks of firms with high and low ESG rankings. Results from this model are very similar to those from Equation (3) and the coefficient on the idiosyncratic risk factor is almost never significant. For brevity we therefore do not display the results from this model.¹⁹

5. Results

The last four columns of Table 1 show the industry composition of the firms that SAM rates and those listed on the FTSE All-Share Index. The SAM sample is generally representative across industries, but is substantially underweight in the financial services sector. Closer examination of the data indicates that many of the financial services firms in the FTSE All-Share Index are small equity and real-estate investment trusts, which are of little concern to SAM, which is primarily interested in the ESG performance of corporations. Overall, then, investors wishing to adopt an ESG investment strategy can use the SAM universe and still obtain sufficient industry diversification.

Table 2 presents descriptive statistics on firm size and E, S or G scores for the high- and low-ranked portfolios. Consistent with prior research we find that high-ranked E, S and G portfolios are comprised of significantly larger firms than their low-ranked counterparts.²⁰ We also note that the mean E, S and G scores for the high-ranked portfolios are more than twice those of the low-ranked portfolios.

Table 2.

Descriptive statistics on size and score. This table provides the descriptive statistics for the firms that comprise the high- and low-ranked environmental, social and governance portfolios. Size Av. and Size Med. are the average and median market capitalization in millions of pounds. Av. Score is the average rating and is out of a possible 100. Results coincide with the annual September rankings from 2002 to 2007. Results for the full period are also presented.

		2002	2003	2004	2005	2006	2007	Full period
Panel A – environmental portfolios

High-ranked portfolio	Size Av.	8433.6	12,361.7	12,051.7	13,955.4	15,821.8	16,272.6	13,149.5
	Size Med.	2932.2	3928.7	4315.0	4462.9	5709.8	5778.4	4521.2
	Av. Score	68.8	68.8	65.2	71.0	73.4	69.1	69.4
Low-ranked portfolio	Size Av.	1779.9	4475.4	1753.1	2372.2	3016.4	3214.7	2768.6
	Size Med.	864.4	1582.9	1295.1	1815.5	2306.9	2080.4	1657.5
	Av. Score	32.8	32.8	27.7	31.5	30.8	20.4	29.3

Panel B – social portfolios

High-ranked portfolio	Size Av.	8607.0	14,497.4	12,149.1	13,950.3	15,661.6	16,316.9	13,530.4
	Size Med.	2283.0	4193.9	4379.3	4446.1	5246.5	5689.6	4373.1
	Av. Score	45.1	65.1	60.3	61.4	65.2	63.7	60.1
Low-ranked portfolio	Size Av.	1534.9	2739.2	1656.9	2377.3	3179.9	2995.8	2414.0
	Size Med.	939.6	1582.9	1378.7	1775.3	2269.0	2080.4	1671.0
	Av. Score	11.2	31.5	26.3	25.9	27.7	26.8	24.9

Panel C – governance portfolios

High-ranked portfolio	Size Av.	8536.8	12,876.5	11,653.4	13,563.8	16,140.5	16,509.6	13,213.4
	Size Med	2125.6	3682.5	4031.3	3607.1	6258.8	5824.7	4255.0
	Av. Score	51.5	62.8	61.7	75.2	61.1	65.5	63.0
Low-ranked portfolio	Size Av.	1605.2	4272.3	2032.1	2763.7	2704.6	2996.3	2729.0
	Size Med	953.3	1689.6	1400.2	1940.7	2202.8	2080.4	1711.2
	Av. Score	18.3	35.7	24.9	37.4	29.5	34.3	30.0

In Table 3 we provide descriptive statistics on the risk and return of the portfolios. The mean returns of the low-ranked portfolios appear slightly higher than those of the high-ranked portfolios. However, t-tests indicate that these differences are not significant. Further, F-tests indicate no significant difference in the standard deviations of the high- and low-ranked portfolios.

Table 3.

Descriptive statistics on risk and return. This table provides descriptive statistics on the risk and return characteristics of the portfolios. High- (low-) ranked portfolios represent portfolios comprised of firms with the 10% best (worst) environmental, social or governance ratings within each of the 20 FTSE All-Share industries. All results are monthly values. Mean return is the arithmetic mean return. Std. dev. is the standard deviation. p value on diff. are the p values from t- and F-tests on differences in high- and low-ranked portfolios’ means and standard deviations respectively. R/R ratio is the risk/reward ratio, calculated as the mean return divided by standard deviation, and Sharpe ratio is the return in excess of the risk-free rate divided by the standard deviation. Portfolios are formed each year in October and held until September the following year. The sample period is from October 2002 to September 2008.

	Mean %	p value on diff.	Std. dev. %	p value on diff.	R/R ratio	Sharpe ratio
Panel A – environmental portfolios

High-ranked portfolio	0.66%	0.89	3.80%	0.97	0.17	0.07
Low-ranked portfolio	0.74%		3.82%		0.19	0.10

Panel B – social portfolios

High-ranked portfolio	0.65%	0.59	3.72%	0.89	0.18	0.08
Low-ranked portfolio	1.01%		3.66%		0.28	0.18

Panel C – governance portfolios

High-ranked portfolio	0.86%	0.85	3.90%	0.44	0.22	0.12
Low-ranked portfolio	0.99%		4.27%		0.23	0.14

Regression results for portfolios formed on E scores are presented in Table 4. None of the coefficients on alpha are significant, suggesting that there is neither a cost nor a benefit to investing in E initiatives. The coefficients on beta are significant across all models, and close to one. This is unsurprising, as we have formed portfolios of stocks that incorporate all of the FTSE industries, meaning our portfolios are likely to resemble the market portfolio. In addition, the systematic risk of high- and low-ranked portfolios is not significantly different – the coefficients on beta of the difference portfolios (Diff(H –L)) are insignificant.

Table 4.

Results for portfolios formed on environmental scores. This table reports the empirical results from the capital asset pricing model (CAPM) and multifactor regression models defined in the text. Independent variables comprise: returns on the market portfolio (FTSE All-Share Index); SMB, HML and UMD represent size, book-to-market and momentum mimicking factors, respectively. Industry factors (not reported) are derived from a principal components analysis on the 20 FTSE All-Share Index industry-sorted portfolios’ returns that are not captured by the single-factor market model. α is the monthly alpha intercept. The dependent variables are portfolios formed on environmental (E) scores. High- (low-) ranked portfolios are formed using the 10% of firms with the best (worst) E score in each industry. The general (industry-specific) criteria portfolios are formed using general (industry-specific) ratings data. Combined portfolios are formed by combining general and industry-specific ratings. Diff(H – L) are the difference portfolios. All regressions use Newey–West (1987) HAC consistent standard errors. The sample period is from October 2002 to September 2008.

	Portfolio	α	β	SMB	HML	UMD	Adj R2
Panel A - CAPM

Combined portfolios	High-ranked	−0.0013	0.97***				0.909
	Low-ranked	−0.0001	0.90***				0.770
	Diff(H – L)	−0.0012	0.07				−0.002
General criteria portfolios	High-ranked	0.0000	0.96***				0.877
	Low-ranked	0.0023	0.97***				0.863
	Diff(H – L)	−0.0023	–0.01				−0.014
Industry-specific criteria portfolios	High-ranked	0.0002	0.90***				0.879
	Low-ranked	−0.0013	0.85***				0.742
	Diff(H – L)	0.0015	0.05				−0.008

Panel B – four-factor market model

Combined portfolios	High-ranked	−0.0009	0.99***	−0.16***	−0.07	0.01	0.924
	Low-ranked	−0.0008	0.87***	0.08	0.14	0.02	0.776
	Diff(H – L)	−0.0001	0.12	−0.23**	−0.21*	0.00	0.112
General criteria portfolios	High-ranked	0.0002	0.98***	−0.14***	−0.04	0.04	0.887
	Low-ranked	0.0023	0.96***	0.20***	0.02	−0.02	0.882
	Diff(H – L)	−0.0021	0.02	−0.34***	−0.06	0.06	0.191
Industry-specific criteria portfolios	High-ranked	0.0006	0.92***	−0.19***	−0.10*	−0.02	0.904
	Low-ranked	−0.0018	0.82***	0.13	0.11	0.02	0.750
	Diff(H – L)	0.0024	0.10	−0.32***	−0.20	−0.04	0.176

Panel C – four-factor market model with industry factors

Combined portfolios	High-ranked	−0.0017	1.00***	−0.15**	−0.06	0.01	0.928
	Low-ranked	−0.0009	0.89***	0.08	0.15	0.05	0.768
	Diff(H – L)	−0.0007	0.11	−0.24**	−0.21	−0.04	0.104
General criteria portfolios	High-ranked	−0.0010	1.02***	−0.13**	−0.01	0.06	0.893
	Low-ranked	0.0020	0.97***	0.20***	0.03	−0.03	0.882
	Diff(H – L)	−0.0030	0.05	−0.32***	−0.04	0.09	0.177
Industry-specific criteria portfolios	High-ranked	0.0001	0.93***	−0.19***	−0.09*	−0.03	0.904
	Low-ranked	−0.0019	0.84***	0.14	0.12	0.04	0.740
	Diff(H – L)	0.0021	0.09	−0.32***	−0.21*	−0.07	0.155

***, **, * indicate statistical significance at the 0.01, 0.05 and 0.10 levels, respectively.

Turning to the other factors (Table 4, Panel B), the coefficients on SMB for high-ranked portfolios are always significantly negative. Some of the SMB coefficients for the low-ranked portfolios are significantly positive and all coefficients on the difference portfolios are significantly negative. These results indicate that companies with higher E scores are systematically larger than low-ranked companies. There are a number of reasons why this may be the case. Firstly, larger companies are likely to be more visible and therefore face more pressure from outsiders to behave responsibly (Moskowitz, 1972; Ullmann, 1985). Secondly, it has long been argued that larger companies have more resources than smaller companies and are therefore able to afford to invest in ESG initiatives (Freeman, 1984).

Coefficients on HML are mostly insignificant, although there is some weak evidence in a few models that high-ranked portfolios have lower book-to-market ratios. None of the coefficients on UMD are significant. These results therefore indicate that portfolios of firms with high and low E rankings are generally no different in terms of their book-to-market ratios or exposure to momentum. Forming portfolios using general, industry-specific or combined E criteria does not alter the results.

Table 5 presents results for portfolios formed using S scores. Results on alpha from the capital asset pricing model (CAPM) and four-factor models (Panels A and B) are generally insignificant, although the low-ranked combined portfolios show some weak evidence of outperformance. Results from the model that includes industry factors (Panel C) provide some evidence of high-ranked S stocks underperforming, with the underperformance in the combined portfolio appearing to be driven by underperformance in the high-ranked industry-specific portfolio. It is therefore interesting to note that high industry-specific S scores appear to detract from performance in this instance – albeit the evidence of underperformance is weak at best.²¹

Table 5.

Results for portfolios formed on social scores. This table reports the empirical results from the capital asset pricing model (CAPM) and multifactor regression models defined in the text. Independent variables comprise: returns on the market portfolio (FTSE All-Share Index); SMB, HML and UMD represent size, book-to-market and momentum mimicking factors, respectively. Industry factors (not reported) are derived from a principal components analysis on the 20 FTSE All-Share Index industry-sorted portfolios’ returns that are not captured by the single-factor market model. α is the monthly alpha intercept. The dependent variables are portfolios formed on social (S) scores. High- (low-) ranked portfolios are formed using the 10% of firms with the best (worst) S score in each industry. The general (industry-specific) criteria portfolios are formed using general (industry-specific) ratings data. Combined portfolios are formed by combining general and industry-specific ratings. Diff(H – L) are the difference portfolios. All regressions use Newey–West (1987) HAC consistent standard errors. The sample period is from October 2002 to September 2008.

	Portfolio	α	β	SMB	HML	UMD	Adj R2
Panel A – CAPM

Combined portfolios	High-ranked	–0.0012	0.95***				0.912
	Low-ranked	0.0026*	0.89***				0.826
	Diff(H – L)	–0.0039	0.06				–0.005
General criteria portfolios	High-ranked	0.0000	0.96***				0.906
	Low-ranked	0.0011	0.95***				0.698
	Diff(H – L)	–0.0011	0.01				–0.014
Industry-specific criteria portfolios	High-ranked	–0.0008	1.00***				0.901
	Low-ranked	0.0023	0.88***				0.825
	Diff(H – L)	–0.0031	0.12				0.019

Panel B – four-factor market model

Combined portfolios	High-ranked	–0.0010	0.96***	–0.08	–0.04	0.00	0.914
	Low-ranked	0.0027*	0.89***	0.18***	–0.01	0.06	0.842
	Diff(H – L)	–0.0038	0.07	–0.27***	–0.03	–0.06	0.080
General criteria portfolios	High-ranked	0.0002	0.98***	–0.15***	–0.06	–0.05	0.919
	Low-ranked	–0.0005	0.87***	–0.02	0.27*	–0.17	0.744
	Diff(H – L)	0.0007	0.10	–0.13	–0.33*	0.11	0.114
Industry-specific criteria portfolios	High-ranked	–0.0013	0.97***	–0.20***	0.08	–0.15***	0.935
	Low-ranked	0.0024	0.88***	0.15**	–0.01	0.07	0.835
	Diff(H – L)	–0.0038	0.09	–0.36***	0.09	–0.22*	0.237

Panel C – four-factor market model with industry factors

Combined portfolios	High-ranked	–0.0017	0.97***	–0.07	–0.04	0.01	0.916
	Low-ranked	0.0030*	0.89***	0.17***	0.00	0.05	0.843
	Diff(H – L)	–0.0047*	0.08	–0.24**	–0.04	–0.03	0.100
General criteria portfolios	High-ranked	–0.0002	0.98***	–0.15***	–0.06	–0.06	0.917
	Low-ranked	0.0006	0.86***	–0.04	0.27*	–0.18	0.747
	Diff(H – L)	–0.0008	0.12	–0.11	–0.33*	0.12	0.125
Industry-specific criteria portfolios	High-ranked	–0.0022**	1.00***	–0.19***	0.09*	–0.12***	0.937
	Low-ranked	0.0025	0.89***	0.15**	0.00	0.06	0.830
	Diff(H – L)	–0.0047*	0.11	–0.34***	0.10	–0.19	0.231

***, **, * indicate statistical significance at the 0.01, 0.05 and 0.10 levels, respectively.

The differences in the systematic risks (betas) of high- and low-ranked S portfolios are not significant. Coefficients on SMB again suggest that high-ranked stocks are larger than low-ranked stocks. However, this appears to be driven by the industry-specific criteria, as the Diff(H – L) SMB coefficients are not significant using general criteria. Most coefficients on HML are insignificant, although the general criteria portfolios indicate some weak evidence of higher ranked portfolios having lower book-to-market ratios. Coefficients on UMD for the industry-specific criteria portfolios suggest that firms with high S rankings may have been poorer performers in the prior period.

The results from forming portfolios on G scores are presented in Table 6. None of the coefficients on alpha are significant. Betas are again significant and close to one and there is no significant difference in the systematic risk of high- and low-ranked portfolios.

Table 6.

Results for portfolios formed on governance scores. This table reports the empirical results from the capital asset pricing model (CAPM) and multifactor regression models defined in the text. Independent variables comprise: returns on the market portfolio (FTSE All-Share Index); SMB, HML and UMD represent size, book-to-market and momentum mimicking factors, respectively. Industry factors (not reported) are derived from a principal components analysis on the 20 FTSE All-Share Index industry-sorted portfolios’ returns that are not captured by the single-factor market model. α is the monthly alpha intercept. The dependent variables are portfolios formed on governance (G) scores. High- (low-) ranked portfolios are formed using the 10% of firms with the best (worst) G score in each industry. The general (industry-specific) criteria portfolios are formed using general (industry-specific) ratings data. Combined portfolios are formed by combining general and industry-specific ratings. Diff(H – L) are the difference portfolios. All regressions use Newey–West (1987) HAC consistent standard errors. The sample period is from October 2002 to September 2008.

	Portfolio	α	β	SMB	HML	UMD	Adj R²
Panel A – CAPM

Combined portfolios	High-ranked	0.0007	0.99***				0.899
	Low-ranked	0.0020	0.96***				0.709
	Diff(H – L)	–0.0014	0.02				–0.013
General criteria portfolios	High-ranked	0.0005	0.93***				0.865
	Low-ranked	0.0006	0.94***				0.805
	Diff(H – L)	–0.0002	0.00				–0.014
Industry-specific criteria portfolios	High-ranked	–0.0005	0.96***				0.870
	Low-ranked	0.0037	0.99***				0.740
	Diff(H – L)	–0.0042	–0.04				–0.012

Panel B – four-factor market model

Combined portfolios	High-ranked	0.0009	1.00***	–0.14**	–0.07	–0.09	0.911
	Low-ranked	0.0005	0.90***	–0.03	0.27	–0.11	0.745
	Diff(H – L)	0.0004	0.10	–0.12	–0.34*	0.02	0.102
General criteria portfolios	High-ranked	0.0005	0.94***	–0.14**	–0.03	–0.10	0.875
	Low-ranked	0.0008	0.94***	0.15**	–0.03	0.01	0.810
	Diff(H – L)	–0.0003	0.00	–0.29***	–0.01	–0.11	0.065
Industry-specific criteria portfolios	High-ranked	–0.0003	0.97***	–0.08	–0.05	0.00	0.870
	Low-ranked	0.0024	0.93***	–0.01	0.24	–0.12	0.768
	Diff(H – L)	–0.0026	0.04	–0.07	–0.29	0.12	0.070

Panel C – four-factor market model with industry factors

Combined portfolios	High-ranked	0.0002	1.03***	–0.14**	–0.05	–0.08	0.914
	Low-ranked	0.0015	0.90***	–0.04	0.27	–0.10	0.752
	Diff(H – L)	–0.0013	0.13	–0.10	–0.33*	0.02	0.135
General criteria portfolios	High-ranked	–0.0008	0.98***	–0.13**	–0.01	–0.08	0.886
	Low-ranked	0.0015	0.95***	0.15*	–0.02	0.03	0.815
	Diff(H – L)	–0.0023	0.03	–0.27***	0.02	–0.11	0.142
Industry-specific criteria portfolios	High-ranked	–0.0006	0.98***	–0.09	–0.04	–0.03	0.875
	Low-ranked	0.0031	0.95***	–0.02	0.25	–0.10	0.778
	Diff(H – L)	–0.0038	0.03	–0.07	–0.29	0.07	0.089

***, **, * indicate statistical significance at the 0.01, 0.05 and 0.10 levels, respectively.

We once more observe negative coefficients on SMB for the high-ranked and difference portfolios, with low-ranked portfolios exhibiting less negative or slightly positive SMB coefficients. However, the coefficients on SMB for the difference portfolios are only significant for the general criteria portfolios. Therefore there is some evidence that high-ranked G stocks are larger, but it is not as pervasive as the evidence for high-ranked E or S portfolios. Almost all coefficients on HML are insignificant and coefficients on UMD are never significant. Results for the combined, general and industry-specific criteria portfolios are not substantially different.

In summary, our results indicate that there is very little evidence of a difference in the performance of stocks ranked according to E or G criteria. We do, however, find some weak evidence that firms that invest in S initiatives may underperform, although this appears to be driven by the industry-specific component of the S ratings. The performance of portfolios formed on the basis of E or G factors is not impacted by using general or industry-specific analysis. In addition, high-ranked stocks do not appear to differ substantially from low-ranked stocks in terms of their systematic risk, book-to-market ratios or momentum. However, we do find strong evidence of higher performing stocks being larger, particularly when portfolios are formed using E and S criteria.

5.1. Robustness tests

Recall that our analysis above uses a Best of Sector approach, where stocks are ranked according to E, S or G criteria within industries. This approach is appropriate for investors who desire a diversified portfolio, but may not result in portfolios that strictly have the best and worst E, S or G scores. For robustness, we therefore investigate two alternative portfolio formation methodologies where firms are ranked purely on their E, S or G score, without taking industry biases into account. They are designed to reflect the stock holdings of investors who are more concerned with E, S or G factors than with industry representativeness and diversification.²² Firstly, we form portfolios by splitting the firms at the median E, S or G score and we refer to these as our broad portfolios. Secondly, we form high- (low-) ranking portfolios from the 40 stocks with the best (worst) E, S or G ranking. We call these our E, S and G conviction portfolios. For brevity, we do not display the tabulated results.²³

Results for E scores are similar to the initial results. Coefficients on alphas and betas of the difference portfolios are never significant. Again there is clear evidence of a size bias and none of the coefficients on HML are significant. However, there is some weak evidence that high-ranking stocks are past losers.

Results for S now show insignificant alphas on the difference portfolios, and beta results are almost identical to those of the initial specification.²⁴ The size results are much stronger, with significantly negative coefficients on SMB for every difference portfolio. Coefficients on HML for the broad portfolios are similar to the Best of Sector results, but the conviction portfolios do not load onto HML at all. There is some evidence of high-ranked S stocks being prior losers, although these results are only particularly strong for the broad portfolios.

For G scores, the broad portfolios have significantly negative alphas on high-ranked portfolios for both the four-factor model and the four-factor model with industry factors. However, this only translates into a significantly negative alpha on the difference portfolio in one instance.²⁵ Alphas on the conviction portfolios are generally insignificant.²⁶ All alphas on the difference portfolios are insignificant except the industry-specific portfolio using the four-factor model. In contrast to the initial results, all SMB coefficients on the difference portfolios are significantly negative across all models. HML coefficients are insignificant in the broad portfolios, but there is evidence of high-ranking stocks having lower book-to-market ratios in the conviction portfolios. Coefficients on UMD suggest that high-ranking governance portfolios may have been losers in the prior year.

To summarize, our robustness tests in the main confirm our initial results. We find little evidence of a difference in the performance of stocks with high and low E, S or G scores. High- and low-ranked E, S and G portfolios also do not appear to differ significantly in their systematic risk or book-to-market ratios. There is some limited evidence that prior poor financial performers may invest more in E, S or G initiatives, but this finding is not consistent across all our models. The most pervasive result throughout the analysis is that high-ranked portfolios across all E, S or G criteria appear to consist of significantly larger stocks than their low-ranked counterparts.

6. Conclusion

There is increasing interest in sustainability investing, or incorporating E, S and G factors into investment decisions, worldwide. The UK is one of the world leaders in this area, but to date there has been surprisingly little research into the impact of ESG factors on the UK market. In this study, we investigate whether E, S or G initiatives have a significant impact on the performance of UK firms.

We find no difference in the performance of portfolios comprised of firms with high and low E or G rankings. We initially find some weak evidence of firms with strong S scores underperforming, but this is not consistent across all of our models and the finding is not upheld in robustness tests.

In terms of stock characteristics, there is little evidence of a difference in the systematic risk or book-to-market ratios of high- and low-ranked E, S or G stocks. In our robustness tests we find a few instances of high-scoring stocks being poor financial performers in the prior period, perhaps suggesting that firms invest in E, S or G initiatives as a way of salvaging their reputations within the community. However, the finding is not observed across all our models, so we do not wish to labour this point. One consistent finding is that firms with high E, S or G rankings tend to be larger. This is perhaps unsurprising. Larger firms are likely to have the resources to invest in ESG initiatives. In addition, larger firms are more visible to the public and therefore more likely to face pressure from outside stakeholders to behave responsibly.

Our results therefore suggest that there is neither a financial cost nor benefit to incorporating E, S or G factors into investment decisions. Our results are robust to using general, industry-specific or combined E, S or G criteria.

We provide two possible interpretations of our findings. Firstly, the UK market is efficient with respect to ESG information. Secondly, it is possible that an E, S or G price premium did not arise over our 2002–2008 sample period. Prior research by Heinkel et al. (2001) and Hong and Kacperczyk (2009) suggests that 10%–20% of the market needs to divest itself of ‘unacceptable’ firms before a material impact on firms’ prices is observed. As previously mentioned, there are currently 113 UK signatories to the UN PRI, representing over US$3 trillion in assets under ESG management. We also find that these signatories manage more than 30% of the UK’s Investment Management Association’s recorded funds under management (as at March 2010). These figures suggest that there may currently be a sufficient proportion of UK assets managed according to ESG principles to move prices. However, the size of the UK’s ESG market over much of our sample period was substantially smaller. For example, in May 2006 there were only 73 UN PRI signatories worldwide compared with 867 today – there were no signatories prior to 2006. Consequently, it is possible that over our sample period the size of the ESG market was not sufficiently large enough to impact asset prices.

Consequently, it is plausible that a price premium effect may occur in the future as assets managed according to ESG criteria continue to grow. Whether investors that currently undertake ESG strategies will benefit from a future increase in the market’s demand for high-rated ESG firms is an interesting question for future research.

Our results are good news for investors who wish to hold stocks of companies that are responsible corporate citizens, or indeed for those institutional investors whose mandates require them to consider ESG factors in investment decision making. Choosing stocks with good E, S or G ratings does not result in lower returns nor appear to breach one’s fiduciary duty to maximize profits. Finally, stocks with good E, S or G ratings are likely to be larger and therefore more liquid and easier to trade, making them more desirable for large institutional investors.

Footnotes

Acknowledgements

The authors thank Paul Dou, Doug Foster, Tom Smith, Terry Walter and seminar participants at Melbourne University, the Australian National University research camp 2010, and the 23rd Australasian Finance and Banking Association Conference 2010 for helpful comments. We also thank SAM and Christophe Churet for providing the sustainability ratings data.

Funding

This work was supported by the Accounting and Finance Association of Australia and New Zealand.

Notes

References

Brammer

Brooks

Pavelin

(2006) Corporate social performance and stock returns: UK evidence from disaggregate measures. Financial Management 35(3): 97–116.

Carhart

(1997) On persistence in mutual fund performance. Journal of Finance 52(1): 57–82.

Core

Guay

Rusticus

(2006) Does weak governance cause weak stock returns? An examination of firm operating performance and investors’ expectations. Journal of Finance 61(2): 655–687.

Cremers

KJM

Nair

(2005) Governance mechanisms and equity prices. Journal of Finance 60(6): 2859–2894.

Derwall

Guenster

Bauer

Koedijk

KCG

(2005) The eco-efficiency premium puzzle. Financial Analysts Journal 61(2): 51–63.

Elton

Gruber

Blake

(1996) The persistence of risk-adjusted mutual fund performance. Journal of Business 69(2): 133–157.

EuroSIF (2008) European SRI Study 2008. Available at http://www.eurosif.org/publications

Filbeck

Gorman

Zhao

(2009) The ‘best corporate citizens’: are they good for their shareholders? Financial Review 44(2): 239–262.

Freeman

(1984) Strategic Management: a Stakeholder Approach. Boston: Pitman.

10.

Galema

Plantinga

Scholtens

(2008) The stocks at stake: returns and risk in socially responsible investment. Journal of Banking and Finance 32(12): 2646–2654.

11.

Gilley

Worrell

Davidson

III El-Jelly

(2000) Corporate environmental initiatives and anticipated firm performance: the differential effects of process-driven versus product-driven greening initiatives. Journal of Management 26(6): 1199–1216.

12.

Gompers

Ishii

Metrick

(2003) Corporate governance and equity prices. Quarterly Journal of Economics 118(1): 107–155.

13.

Hawley

Williams

(2007) Universal owners: challenges and opportunities. Corporate Governance: An International Review 15(3): 415–420.

14.

Heinkel

Kraus

Zechner

(2001) The effect of green investment on corporate behaviour. Journal of Financial and Quantitative Analysis 36(4): 431–449.

15.

Herremans

Akathaporn

McInnes

(1993) An investigation of corporate social responsibility reputation and economic performance. Accounting, Organizations and Society 18(7–8): 587–604.

16.

Hon

Tonks

(2003) Momentum in the UK stock market. Journal of Multinational Financial Management 13(1): 43–70.

17.

Hong

Kacperczyk

(2009) The price of sin: the effects of social norms on markets. Journal of Financial Economics 93(1): 15–36.

18.

Hwang

(2007) Cross-sectional stock returns in the UK market: the role of liquidity risk. In: Satchell

(ed.) Forecasting Expected Returns in the Financial Markets. Oxford: Elsevier, 191–213.

19.

Klassen

McLaughlin

(1996) The impact of environmental management on firm performance. Management Science 42(8): 1199–1214.

20.

Lado

Wilson

(1994) Human resource systems and sustained competitive advantage: a competency-based perspective. Academy of Management Review 19(4): 699–727.

21.

Lee

Faff

(2009) Corporate sustainability performance and idiosyncratic risk: a global perspective. Financial Review 44(2): 213–237.

22.

Lee

Faff

Langfield-Smith

(2009) Revisiting the vexing question: does superior corporate social performance lead to improved financial performance? Australian Journal of Management 34(1): 21–49.

23.

Lee

Humphrey

Shen

(2010) The cost of meeting an ESG mandate. Working paper.

24.

Sheu

(2007) Is corporate sustainability a value-increasing strategy for business? Corporate Governance: An International Review 15(2): 345–358.

25.

Moskowitz

(1972) Choosing socially responsible stocks. Business and Society Review 1(1): 71–75.

26.

Richardson

(2007) Do the fiduciary duties of pension funds hinder socially responsible investment? Banking and Finance Law Review 22(2): 145–201.

27.

Statman

(2006) Socially responsible indexes. Journal of Portfolio Management 32(3): 100–109.

28.

Statman

Glushkov

(2009) The wages of social responsibility. Financial Analysts Journal 65(4): 33–46.

29.

Turban

Greening

(1997) Corporate social performance and organizational attractiveness to prospective employees. Academy of Management Journal 40(3): 658–672.

30.

Ullmann

(1985) Data in search of a theory: a critical examination of the relationships among social performance, social disclosure, and economic performance of U.S firms. Academy of Management Review 10(3): 540–557.