Does Sustainability Assurance Improve Managerial Investment Decisions?

Abstract

This study analyzes the effect of sustainability assurance (SA) on managerial investment decisions in terms of sustainability investment (SI) efficiency. We hypothesize that SA improves the set of information available for managerial decision making, resulting in higher SI efficiency. Further, we argue that SA reduces information asymmetry between managers and investors, which enables investors to more effectively monitor a firm’s management, thus again leading to higher SI efficiency. Empirical findings for an international sample support these links. Moreover, we find weak evidence that SA provided by an auditor is associated with a stronger effect on SI efficiency. In additional analyses, we find weak evidence that the association of SA and SI efficiency is more pronounced if the SA is provided on a higher scope, on a higher level, and if other governance mechanisms are weak.

Acknowledgements

This paper has benefited from highly valuable comments and suggestions by Markus Arnold, Rolf Uwe Fülbier, Alexis Kunz, Wayne Landsman, Thomas Loy, Michel Magnan (the associate editor), Gary Monroe, William Rees, Cathy Shakespeare, Martin Wallmeier, and Stephen Zeff, two anonymous reviewers, 36th Annual Congress of the European Accounting Association in Paris, 2015 Schöller Workshop in Nuremberg, 78th Annual Conference of the German Academic Association of Business Research in Munich, and research workshops at the University of Bayreuth, University of Bern, and at the Friedrich-Alexander-University Erlangen-Nuremberg. We thank Wim Bartels, Global Head of Sustainability Assurance at KPMG International, and this team for highly valuable discussion and for the confidential provision of the raw data.

Notes

1 We define a firm’s sustainability as ‘adopting business strategies and activities that meet the needs of the enterprise and its stakeholders today while protecting, sustaining, and enhancing the human and natural resources that will be needed in the future’ (Deloitte & Touche & ISSD, 1992, p. 11). Consistent with this definition and being aware of the pluralism of overlapping terminology, we use the terms ‘sustainability reporting’ and ‘sustainability assurance’.

2 McWilliams and Siegel (2001) argue that for each firm an optimal level of SI exists that managers can determine via cost–benefit considerations. This approach is consistent with the view that managers should not invest (abandon an investment) in sustainability activities unless the economic benefits of conducting (leaving out) this investment exceed the costs, thus ultimately leading to maximized shareholder value.

3 De facto, all countries have established regulations that require a firm to disclose certain sustainability-related information (e.g. a firm’s corporate governance, employees, or greenhouse gas emissions) that are typically audited in a way conceptually similar to financial disclosure. However, these requirements refer to rather specific topics and do not mandate a comprehensive reporting on a firm’s overall sustainable activities. To our knowledge, France and South Africa are the only pioneering examples where firms have been recently required to publish (partially) assured information that explicitly refers to a firm’s overall sustainability activities (EY, 2015; KPMG, 2013).

4 Economic objectives of SI include enhanced revenue growth (e.g. Lev, Petrovits, & Radhakrishnan, 2010), employee satisfaction (e.g. Edmans, 2011), corporate reputation (e.g. Roberts & Dowling, 2002), and insurance-like protection against negative events (e.g. Godfrey et al., 2009). Although not uncontested, the empirical literature generally finds support for reaching these objectives via sustainability management.

5 Such deficits in sustainability-related information might result from managers’ lack of awareness of sustainability-related issues, a constrained accounting budget, and insufficient expertise for sustainability accounting.

6 For example, SA can facilitate managers’ selection of SI-related measures to adequately incorporate stakeholders’ claims in its strategic performance measurement system (e.g. Rodrigue et al., 2013).

7 Such misstatements can result from internal agency conflicts. For example, when the sustainability-related incentives of managers differ along a firm’s hierarchy, subordinated managers might intentionally forward incorrect sustainability information to maximize their own income. As a SA also evaluates whether subordinated managements faithfully operate according to a firm’s sustainability accounting guidelines, SA can diminish subordinated managers’ incentives for non-complying actions, leading to an improved set of information available to a firm’s top managers.

8 Referring to moral hazard, when the incentives of a firm’s managers differ from the preferences of its investors, managers might conduct investments that are not optimal for the investors. Referring to adverse selection, when a firm’s management disposes of a larger set of information than the investors, it can use private information to issue over-priced securities of a firm and use these funds for over-investment. However, a firm’s investors may respond rationally by rationing capital, which can lead to ex post under-investment.

9 To substantiate our assumption that information in SR facilitates stakeholders’ monitoring, we examine the overall effect of SR on SI efficiency. Therefore, we adapt the approach described in Section 6.1.1 for SR and estimate this model for SR and non-SR firms. We find that SR is positively associated with SI efficiency, which suggests that SR – irrespective of SA – is likely to contain information that enhances stakeholders’ ability to monitor SI.

10 Expertise is particularly related to the likelihood that an attestation service provider discovers deficient aspects in the client’s accounting system. The conditional probability of reporting on detected deficiencies and/or requesting corrections before giving a favorable opinion is related to the provider’s objectivity, which is interpreted as a key consequence of a SA provider’s independence from a given client (Watts & Zimmerman, 1981).

11 For this argumentation, an increase in actual SR quality (i.e. higher accuracy of information provided in SR) and/or an improvement in stakeholders’ perception of the SR quality can establish the assumed link. Even in the absence of a change in actual SR quality, SA can increase stakeholders’ confidence in SR, leading to an improved decision usefulness of SR because information would have been discounted/ignored in the absence of a SA (e.g. Casey & Grenier, 2015).

12 Consistent with this argumentation, the prior literature suggests that audit quality is positively related with financial reporting quality (e.g. DeFond & Zhang, 2014), which also constitutes a link between audit quality and managerial investment efficiency (e.g. Bae & Choi, 2012; Kausar, Shroff, & White, 2016; Lu & Sapra, 2009).

13 Cohen and Simnett (2015) note further aspects that make the provision of SA challenging and potentially diminish SA quality (e.g. lack of analytic rigor and of well-developed criteria for materiality judgments).

14 For example, The Guardian unfavorably commented on EY’s role in the preparation and assurance of Barclays PLC’s 2012th SR: ‘social auditors failed to expose the bank’s rotten culture’ (The Guardian, 2012).

15 The KPMG raw datasets consist of the 100 largest public and private firms from 21 (volume 2008), 34 (volume 2011), 41 (volume 2013), and 45 countries around the world (volume 2015). The conclusions remain unchanged (i) if we require inclusion of a firm in all four volumes and (ii) if we conduct our analyses for each volume separately.Our sample also includes firms from the financial service industry. We argue that our measurement approach for SI efficiency is applicable to all industries after controlling for industry effects because the level of disclosure and the types of SI are relatively comparable within industries (Lys et al., 2015). We note that the tenor of our findings remains unchanged if we exclude firms from the financial service industry.

16 We attribute data from the 2008 (2011, 2013, and 2015) volume of KPMG’s study to the years 2006–2007 (2008–2010, 2011–2012, and 2013–2015). We argue that this approach can reasonably be applied because neither our theoretical arguments nor our empirical identification strategy requires that the effect of SA on SI occurs in one specific year. Different to studies on the market effects of the first-time issuance of (assured) SR which require strong time-lag assumptions (e.g. Dhaliwal et al., 2011), we argue that both SA and SI efficiency are rather sticky-firm attributes. This assumption is supported by the findings (i) that the decision to purchase SA is rarely reversed (in 4% of our sample) and (ii) that our measure of SI efficiency is highly one- and two-year-autocorrelated (correlations > 0.6) which indicates a low extent of short-term SI reversals.

17 Thus, our sample contains (multiple statements possible) stand-alone SR (68.2%), separate chapters of the annual reporting (37.6%), and definable reporting on corporate websites (7.9%). All conclusions are unaffected when we conduct our analyses only based on stand-alone SR.

18 We note that each approach has its advantages and disadvantages and that the validity of the findings depends on the appropriateness of the assumptions of each approach (Fu et al., 2012).

19 Even though there are doubts whether this assumption is fully met, we apply OLS regression analysis as it remains ultimately unclear whether other examination techniques produce preferable results (Larcker & Rusticus, 2010).

20 The ideal IV should be highly related to the potentially endogenous explanatory variable (in our case, Sa_i,t) and unrelated to the dependent variable (in our case, Sie_i,t) except through the instrumented potentially endogenous explanatory variable. A particular challenge of our search for an appropriate instrument is that the prior literature provides evidence that the incentives to issue SR overlap with those to purchase SA (Cormier, Magnan, & Van Velthoven, 2005; Simnett et al., 2009).

21 This approach also reduces self-selection concerns as the firms from the treatment and control group are close with respect to the key drivers to purchase SA. Consequently, changes in SI efficiency attributed to the newly purchased SA are less likely to be endogenously induced by a firm’s characteristics included in Equation (3).

22 We explicitly control for SA provider size because of the considerable overlap between SA providers’ professional background and SA providers’ size (i.e. the four largest SA providers are the big four auditors). Theoretical literature suggests that the size of a provider is positively related to attestation quality. This effect is likewise attributed to higher expertise and higher objectivity of larger attestation service providers (e.g. Watts & Zimmerman, 1981).

23 Besides the (total) investment efficiency literature (e.g. Biddle et al., 2009), splitting payments in an optimal and a deviation component is a widespread approach (e.g. in the management compensation literature; e.g. Core, Holthausen, & Larcker, 1999).

24 Specifically, to test their investment hypothesis, Lys et al. (2015) model the economically optimal level of sustainability-related expenditures based on economic factors. This approach is conceptually in line with McWilliams and Siegel (2001) who argue that an ‘ideal’ level of SI exists which managers can determine via cost–benefit analyses.

25 SR generally does not provide direct data on sustainability expenditures (Moser & Martin, 2012). Consistent with the prior literature, we assume that a firm’s sustainability score is related to a scale measure of a firm’s actual SI (Servaes & Tamayo, 2013). As we control for industry differences, this assumption is reasonable as the types of SI and the level of disclosure are relatively comparable within industries (Lys et al., 2015). A firm’s environmental score refers to resource reduction, emission reduction, and product innovation. A firm’s social score incorporates employment quality, health safety, training development, diversity opportunities, community, human rights, and product responsibility. We use the environmental and social score because these dimensions of sustainability are connected with the notion of SI.

26 We do not include advertising expenses due to data unavailability. However, we note that all findings remain unchanged when we include selling, general, and administrative expenses which, among others, incorporates advertising expenses.

27 Si_i,t is defined in the interval from 0 to 100 while higher values indicate higher SI. We note that Si_i,t (Sie_i,t) has a considerably high volatility in terms of quartile-range from 68.5 to 91.0 (−10.8 to –2.7) and standard deviation of 21.2 (7.4). As Sie_i,t is our major measurement variable, we carefully investigate the robustness of the findings against alternative estimation approaches. We note that all conclusions remain unchanged when we (i) estimate Equation (7) for the full sample with/without country fixed effects, (ii) remove each of the explanatory variables one at a time from the model (Lys et al., 2015), (iii) include additional explanatory variables (e.g. for growth expectation; Biddle et al., 2009), and (iv) estimate Equation (7) separately for the overall environmental and social score provided in ASSET4.

28 The validity of our 2SLS approach is supported by the following tests (Larcker & Rusticus, 2010). As the over-identification restriction test does not reject the appropriateness of Index_i,t as an instrument for Sa_i,t, we conduct a classic Hausman (1978) for endogeneity which suggests that Sa_i,t cannot be assumed to be exogenous.

29 We also conduct analyses to find out when the hypothesized effect of SA on SI efficiency actually occurs. Conceptually in line with Christensen, Hail, and Leuz (2016), for the sub-sample of firms that newly decide to purchase SA during our investigation period, we counter-factually shift the firm-specific date of the first-time purchase of SA around the date of the correct first-time purchase. The annual shifts range from minus three to plus three years, respectively. We find that the detected effect of SA on SI efficiency has the highest magnitude if we use the dates of our above-explained attribution scheme. We note that the association of SA and SI efficiency is partially washed out to the extent of our counter-factual shifting. This evidence supports the validity of our attribution scheme.

30 For this comparison, we re-run the model given in Equation (2) for the sample used for Equation (9) (5278 obs.) to avoid biases from different sample sizes because bid–ask spreads are not available for all observations in our full sample.

31 We note the following important limitation of this analysis. For our overall conclusions, we keep in mind that this interpretation only holds for the very restrictive requirements that the bid–ask spread metric fully and unbiasedly captures the external information environment and that absolutely no other link than the two modeled ones is in place. Thus, we take this analysis only as a descriptive indication of the importance of the link through the internal set of information.

32 To evaluate the robustness of our findings, we re-conduct our analyses for the subsample without firms that receive SA from non-auditors. This analysis is particularly motivated by prior evidence which suggests that auditors use more caution language in the SA opinions/reports (O’Dwyer & Owen, 2005; Perego & Kolk, 2012) which points to the higher conservatism of SA providers from the audit profession. However, our findings do not substantiate this concern.

Additional information

Funding

We thank the Schöller Research Center for financial support.