Corporate Environmental Responsibility and the Cost of Capital: International Evidence

Economist, 5 March 2014.

Economist, 8 February 2014.

Economist, 26 September 2015.

Economist, 3 October 2015.

A 2013 survey by KPMG reveals that 82 % of Fortune Global 250 firms release corporate responsibility information (either in standalone reports or as part of annual financial reports), as opposed to 78 % in 2011. As the report further indicates, “Most G250 CR reports (87 percent) identify at least some social and environmental changes (or ‘megaforces’) that affect the business. Climate change, material resource scarcity and energy and fuel are the most commonly mentioned” (p. 13). Consistent with the strategic importance of CER, the report stresses that

Many companies no longer see corporate responsibility as a moral issue, but as core business risks and opportunities. More and more investors accept that environmental and social factors put company value at stake. This leads to the question of what the potential financial impacts of those risks and opportunities could be and what the company is doing to mitigate or maximize them (p. 14).

Sharfman and Fernando (2008) find a positive relation between environmental performance and the cost of debt, which contradicts their evidence for equity pricing. Given that the risk channels through which CER affects the cost of equity are inherently different from those that affect the cost of debt (Sharfman and Fernando 2008), in this paper we focus on shareholders’ perception of CER.

Consistent with potential cross-country differences, the 2013 KPMG survey of corporate responsibility indicates that among the world’s largest 250 companies, those from Europe

are the most likely to discuss in detail the environmental and social impacts of their products and services. Almost three quarters (73 percent) of reporting companies in Europe do so with a further 23 percent providing limited information. In the Americas, less than half (49 percent) provide detailed information on downstream impacts and the figure drops to less than one third (32 percent) in Asia Pacific (p. 17).

Prior research finds that CSR engagement is inversely related to firm risk (Boutin-Dufresne and Savaria 2004; Lee and Faff 2009; Jo and Na 2012; Kim et al. 2014).

We emphasize that the environmental costs we analyze are external costs, that is, costs that affect a party (in our context, society) that did not choose to incur them (Buchanan and Stubblebine 1962). Thus, the environmental costs that we study in this paper are not accounting costs (Jo et al. 2015b). Jo et al. (2015b) find an insignificant negative correlation between external environmental costs and accounting costs for the manufacturing industry.

Below, we check the robustness of our main evidence to alternative models of the cost of equity.

See El Ghoul et al. (2011) for a discussion of the advantages of the implied cost of capital approach.

For an overview of this literature, see Orlitzky et al. (2003), Margolis et al. (2007), and Baron et al. (2011).

A notable exception is Brammer et al. (2006), who investigate the relationship between corporate social performance and stock returns of UK firms. They observe that firms with higher environmental performance realize lower stock returns. We present cross-country evidence on the relation between CER and the cost of equity capital for an institutionally diverse sample of 30 countries from 2002 to 2011.

A handful of studies examine the effects of CER on debt financing costs. Focusing on the most polluting US industries—chemical and pulp and paper—Schneider (2011) argues that toxic emissions increase firm’s bankruptcy risks and thus lead to more expensive bond prices. Graham et al. (2001) investigate new bond issues over the 1990–1992 period and find environmental liability information negatively influences bond ratings. Similarly, Bauer and Hann (2010) demonstrate that environmental incidents constitute meaningful risks for investors in the non-secured publicly traded debt market. They find that CER is generally associated with a lower cost of debt and higher credit ratings. Chava (2014) provides evidence that bank lenders charge a significantly higher interest rate on loans to firms with environmental concerns (such as hazardous chemical, substantial emissions, and climate change concerns).

Canadian and US firms’ financial statement data are from the Compustat North America file, while data for firms from the rest of the world are from the Compustat Global file.

Input–output modeling shows the amount of resources required to produce a unit of output, and where this output is sold. Trucost uses a global input–output model based on detailed government census and survey data on resource use and pollutant releases, industry data and statistics, and national economic accounts for over 700 environmental resources (Trucost 2008).

For more details on Trucost methodology, we refer the reader to http://​www.​trucost.​com/​methodology.

See “Appendix 2” section for a detailed explanation of the Trucost data.

In other words, the environmental costs of high CER firms should be lower.

Firm-level environmental costs are directly related to firm size. For example, large firms have generally higher absolute environmental costs than small firms. Thus, we measure environmental costs relative to firm size, i.e., we normalize environmental costs by total assets to control for size effects (Kim et al. 2015). As a test of robustness, we re-estimate our baseline regression using environmental costs–sales and the logarithm of environmental costs as alternative proxies of environmental costs. Our results are robust to using these alternative proxies of environmental costs. We discuss these tests in more detail later in the paper.

We proxy for firm risk using the volatility of stock returns instead of beta following Hail and Leuz (2006, 2009). This design choice allows us to avoid taking a position on whether international equity markets are integrated. Specifically, if equity markets are segmented, one should use a local equity index to estimate a firm’s beta. However, if equity markets are integrated, one would use a world equity index. Nonetheless, we find that our evidence remains when we control for beta instead of the volatility of stock returns.

Since firm fixed effects would be perfectly correlated with industry and country fixed effects, we do not include firm fixed effects in our equity pricing regressions (Khurana and Raman 2004; Lawrence et al. 2011).

Recall that our main independent variable of interest, ENVCOST, reflects the level of CER because increasing CER investment lowers (external) environmental costs.

FEPS _t+1 is forecasted earnings for year t + 1 and P _t is stock price measured 10 months after the fiscal year-end.

In our main analysis, we assume that the perpetual growth rate is equal to the future inflation rate when we estimate the cost of equity following Claus and Thomas (2001) and Ohlson and Juettner-Nauroth (2005).

In untabulated tests, we examine whether our evidence is robust to alternative specifications for the cost of equity estimates. We use the median and the first principal component instead of the average of the four individual cost of equity models and employ the ‘real’ cost of equity by subtracting the inflation rate from the cost of capital. Our evidence remains intact.

For instance, Ali et al. (1992) argue that analysts have a tendency to react gradually to publicly available information.

In untabulated tests, we eliminate the top 5, 10, and 25 % of firm-year observations in the long-term growth forecast (LTG) distribution, respectively. We continue to find a significant positive relation between ENVCOST and the cost of equity capital.

Alternatively, we employ the industry average environmental costs–total assets in the first year of data and a dummy variable for whether prior year’s earnings are negative as instruments. We use a negative earning dummy variable because when previous year’s earnings are negative, the firm has fewer resources to invest in CER. At the same time, it is unlikely that previous year’s earnings will affect contemporaneous cost of equity. The (untabulated) results provide similar evidence to that reported in Table 7.

The Box–Cox transformation of ENVCOST is (ENVCOST ^λ  − 1)/λ. Our estimate of λ is 0.747.

In an alternative (untabulated) test to assess whether the heterogeneity in the number of observations across countries affects our results, we run a weighted least squares (WLSs) regression where the weight is the inverse of the number of firm-year observations per country. We continue to estimate a positive and significant coefficient on ENVCOST.

Interestingly, we note that the coefficient on ENVCOST is less significant for emerging countries relative to developed countries, and is less significant for North America relative to Asia Pacific and Europe.