Corporate social responsibility disclosure: The case of international shipping

Transportation Research Part E 71 (2014) 18–44

Contents lists available at ScienceDirect

Transportation Research Part E

journal homepage: www.elsevier .com/locate / t re

Corporate social responsibility disclosure: The caseof international shipping q

http://dx.doi.org/10.1016/j.tre.2014.08.0061366-5545/� 2014 Elsevier Ltd. All rights reserved.

q We thank Daniela Senkl and two anonymous referees for valuable comments.⇑ Corresponding author.

Wolfgang Drobetz a,⇑, Andreas Merikas b, Anna Merika c, Mike G. Tsionas d

a School of Business, University of Hamburg, Von-Melle-Park 5, Hamburg 20146, Germanyb Department of Maritime Studies, University of Piraeus, Karaoli & Dimitriou 40, 18534 Piraeus, Greecec Department of Economics, Deree College, The American College of Greece, 6 Gravias Street, 15342 Athens, Greeced Lancaster University Management School, Lancaster LA1 4YX, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history:Received 7 November 2013Received in revised form 23 July 2014Accepted 20 August 2014

Keywords:Corporate social responsibilityFirm valuationMarkov Chain Monte Carlo

Based on practices and legislation in the shipping industry, we construct a corporate socialresponsibility (CSR) disclosure index for listed shipping companies. We use Markov ChainMonte Carlo (MCMC) techniques for Bayesian inference, and we estimate the marginaleffects of firm characteristics on CSR disclosure for each firm. Our results show a positiverelationship between CSR disclosure and financial performance for each firm in our inter-national sample. Firm size, financial leverage, and ownership structure are also associatedwith CSR disclosure. Our findings suggest that a majority of listed shipping companies haveintegrated CSR practices into their strategic planning and operations.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Corporate social responsibility (CSR) practices have come under the spotlight of attention recently by regulators, policy-makers, and businesses. This development is partially attributable to the plethora of recent financial scandals, such as thecollapses of Enron, Parmalat, Bre-X Minerals, and Lehman Brothers, as well as catastrophic environmental accidents, suchas Exxon Mobil, Sea Star, Torrey Canyon, and several other oil spills. Moreover, an international survey of CEOs carriedout by the United Nations Global Compact in 2010 showed that the overwhelming majority believes CSR is an importantpath to higher profitability.

It is nowadays commonly believed that socially responsible firms, i.e., those that contribute both economically and eth-ically to the society and local communities they serve, are better positioned to grow in terms of reputation and revenues.Therefore, we address the following research question: Has increased corporate transparency through CSR disclosure prac-tices become an integral business process in the global shipping industry? If so, CSR disclosure should be positively associ-ated with corporate performance. This question is even more interesting given anecdotal evidence in Hong et al. (2012) thatannual CSR outlays of U.S. firms can reach the hundreds of millions of dollars, thus potentially outweighing any benefits.

The literature has examined the CSR puzzle in a unified institutional framework (Mattingly and Berman, 2006; Baronet al., 2011), and it has largely failed to detect a clear relationship between CSR disclosure and firm value (or profitability).The meta-analysis of Margolis et al. (2007) concludes that the empirical link between CSR transparency and corporate finan-cial performance is positive, but the effect is of only a small magnitude. In this study, we use the shipping industry as a plat-form to explore these questions further, because we believe it is an unusually distinct industry. Being globalized, extremely

http://crossmark.crossref.org/dialog/?doi=10.1016/j.tre.2014.08.006&domain=pdf

http://dx.doi.org/10.1016/j.tre.2014.08.006

http://dx.doi.org/10.1016/j.tre.2014.08.006

http://www.sciencedirect.com/science/journal/13665545

http://www.elsevier.com/locate/tre

W. Drobetz et al. / Transportation Research Part E 71 (2014) 18–44 19

competitive, and highly volatile in nature, it has progressed further than most land-based industries in developing a gener-ally accepted regulatory system for corporate behavior. The primary importance of shipping is naturally the integral role itplays in international trade. It is therefore often at the forefront of demands for greater transparency and accountability onissues such as global climate change, energy efficiency, waste management, worker safety and security, ocean and coastalhealth, and local community impacts and benefits (Coady et al., 2013). The IMO (the International Maritime Organization)and the ILO (the International Labor Organization) set the safety standards and oversee labor rights for the shipping industry,and they provide the common institutional framework for CSR disclosure practices. Flag states, classification societies, andport controls are also part of the corporate governance framework of the maritime sector, as these institutions serve enforce-ment roles.

We analyze the websites and annual reports of 111 internationally listed shipping firms during our 2002–2010 sampleperiod for potential determinants of CSR disclosure. Given this unique dataset, the contributions of our paper are threefold.First, our study quantifies the effects of corporate performance, but also of firm size, leverage, and ownership structure thatare associated with CSR disclosures in the shipping sector. Second, we construct a CSR disclosure index for each sample firmfollowing Tagesson et al. (2009) and Orens et al. (2010), and we apply Generalized Smoothly Mixing Regression (GSMR)models to capture heterogeneity among firms in this index (Geweke and Keane, 2007; Villani et al., 2012). We use MarkovChain Monte Carlo (MCMC) techniques for Bayesian inference, and we derive statistically significant and distinct marginaleffects for each firm in our sample. Using Tobin’s Q as a measure of financial performance, we find that the relationshipbetween CSR disclosure and firm valuation is positive for all firms in our sample. We also find a strong and positive associ-ation between both concentrated ownership and size with CSR disclosure for all firms. In contrast, we report a negative rela-tionship between financial leverage and CSR disclosure. Third, our results are consistent with the observation that CSRdisclosure is increasingly becoming the norm among shipping firms. This development suggests that enhanced CSR disclo-sure practices are not merely seen as the effect of sound financial performance, but perhaps more importantly, they arebecoming a requirement for it. This relationship may partially explain why we find that shipping firms are eager to increasethe amount of CSR information they disclose. Our analysis cannot unambiguously solve the issue of causality, but our resultsnevertheless offer some important policy implications for executives in the shipping industry.

The remainder of this paper is organized as follows: Section 2 contains a brief review of the extant CSR literature, andSection 3 discusses CSR in a shipping context. Section 4 develops our testable hypotheses. Section 5 describes the construc-tion of our CSR disclosure index, and Section 6 presents our sample and methodology. Section 7 discusses our main empiricalfindings and contains several robustness checks. Finally, Section 8 concludes and provides an outlook for further research.

2. Literature review

2.1. CSR, firm performance, and firm risk

It was a different era when Friedman (1970) and his followers claimed that people have social responsibilities, notcompanies. At the time, the reasoning was that the concept of CSR was patently incompatible with the nature and purposeof an enterprise, which was assumed to be solely shareholder wealth maximization. Nowadays, CSR is still far from beingwell-defined, thus managers cannot easily determine what level of social responsibility their firms should assume. However,shareholders themselves are able to decide whether their stock income sufficiently represents social awareness. Put briefly,the main concern by CSR skeptics is that the (immediate) costs associated with CSR improvements are likely to outweigh thefinancial benefits (mostly occurring in the distant future), which may ultimately make CSR inconsistent with the principlesof shareholder wealth maximization.

Nevertheless, since Friedman’s time, a large part of the corporate management literature has embraced CSR as ethical busi-ness behavior that can enhance shareholder value. Based on an analysis of the fundamental concepts of CSR practices, Brooks(2012) argues that CSR is compatible with a set of assumptions that underlie the free enterprise system. Freeman (1984) viewsCSR as an optimal choice that minimizes transaction costs and potential conflicts with stakeholders. In this sense, CSR may beinterpreted as an effective tool for improving firm reputation and mitigating the risks to a firm of becoming a victim of con-sumer activism and even legal action. Eichholtz et al. (2010) summarize the arguments in the management literature, andsuggest that firms with well-defined and aggressive CSR policies may outperform competitors for the following reasons:improved corporate reputations, less intrusion from activist and governmental organizations, reduced threat of regulation,and improved profitability through lower input costs and higher employee productivity. Moreover, CSR can be related to per-formance because it serves as a proxy for management skills (i.e., a forward-thinking and long-term-oriented management),lowers operating risks, and potentially reflects (technological) innovation (Guenster et al., 2011).

From a financial point of view, CSR can have an impact on firm performance if and only if it affects expected future cashflows and/or risk (Bouslah et al., 2013). For example, Sharfman and Fernando (2008) argue that improved environmental riskmanagement (e.g., a reduction in emissions and pollutants) reduces the probability of environmental crises that can nega-tively affect a firm’s expected cash flows (e.g., lawsuits, clean-up costs of environmental accidents, fines, and reputationdamage). Investing in CSR can thus create ‘‘moral capital,’’ or goodwill, which provides insurance-like protection in the eventof a crisis or a negative event (Godfrey et al., 2009).

In standard asset pricing models, investors evaluate assets based solely on payoffs and thus the total consumptionthey provide (Fama and French, 2007). Alternative theoretical models by Heinkel et al. (2001), Barnea et al. (2005), and

20 W. Drobetz et al. / Transportation Research Part E 71 (2014) 18–44

Mackey et al. (2007) relax the assumption of perfect capital markets, and assume some differences in investor preferences(segmented capital markets based on CSR). Traditional investors consider only financial criteria (return and risk) in theirinvestment decisions, while socially responsible investors consider both financial and CSR criteria. A general prediction ofthese models is the ‘‘neglect effect’’ caused by CSR, which leads to excess demand for socially responsible stocks and a lackof demand for stocks deemed ‘‘socially irresponsible.’’ As a result, the former tend to be overvalued (with initially higherreturns but lower expected returns and lower risk), while the latter tend to be undervalued (with initially lower returnsbut a higher expected return as a premium for the lack of risk-sharing opportunities and higher risk).

Similar predictions about neglected stocks and segmented markets can be derived from Merton’s (1987) ‘‘investor recog-nition theory.’’ If firms with higher (lower) CSR tend to have a larger (smaller) investor base, his model predicts two effects.First, due to the neglect effect and limited risk sharing, traditional asset pricing models will no longer hold, and idiosyncraticrisk will affect pricing. As a result, there will be a positive relationship between idiosyncratic risk and expected return wheninvestors fail to diversify their portfolios. Second, as a firm’s risk decreases with the size of its shareholder base due toincreasing risk-sharing opportunities, firms with higher (lower) CSR will have lower (higher) expected returns. Hong andKacperczyk (2009) confirm these predictions from the investor recognition model, and show that ‘‘sin’’ stocks (i.e., tobacco,alcohol, and gambling stocks) have higher expected returns (and risks) because they are neglected by social norm-constrained firms (e.g., pension funds). Lee and Faff (2009) conclude that the model is consistent with the notion that themarket values CSR-related risk management and transparency practices.

Thus far, the empirical results in the literature are not clear in establishing a definite relationship between CSR and cor-porate performance (Margolis and Walsh, 2003; Orlitzky et al., 2003; Margolis et al., 2007; Baron et al., 2011). However, anincreasing number of empirical studies have found a positive relationship between CSR and measures of corporate perfor-mance. For example, Pava and Krausz (1996) and Preston and O’Bannon (1997) report that CSR is positively associated withfinancial performance. Stanwick and Stanwick (1998) and Verschoor (1998) document synergies between corporate perfor-mance and good stakeholder relationships. Ruf et al. (2001) find that CSR is positively associated with sales growth andreturns on sales. Bauer et al. (2005) compare active strategies of ‘‘ethical’’ and traditional investment funds. They find mixedresults, but observe that a learning process seems to gradually improve the performance of ethical fund managers. Geczyet al. (2005) compute the cost of imposing socially responsible investment constraints in terms of risk-adjusted returns,and show how they depend on the total share of socially responsible investments as well as on fund managers’ ability.Barnea and Rubin (2010) show that the decision to invest in CSR is negatively correlated with insider ownership, which theyinterpret in the context of the overinvestment hypothesis. In particular, CSR is good for shareholder value up to a certainlevel, but above that level insiders may have an incentive to overinvest in CSR in order to improve their own reputationsas good citizens (managerial opportunism theory). In a different context, Eichholtz et al. (2010) report that buildings witha ‘‘green rating’’ command higher rental rates than otherwise identical buildings. Finally, Guenster et al. (2011) documentthat ‘‘eco-efficiency’’ (what they define as creating more value with fewer environmental resources and less environmentalimpact) influences operating performance, profitability, and the valuation of U.S. industrial firms (measured as Tobin’s Q).

A common problem with empirical research on the value implications of CSR is the issue of reverse causality. Studies thatregress returns of trading strategies or other annual measures of firm value on low frequency measures of CSR cannot reli-ably address whether firms ‘‘do well because they do good,’’ or whether they ‘‘do good because they do well.’’ Given that lessfinancially constrained firms spend more on goodness and have higher goodness scores, however, Hong et al. (2012) claimthat firms do good only when they do well in the sense of possessing financial slack (slack resources theory; see also McGuireet al., 1988). Krueger, 2013 addresses this methodological problem using an event study approach. He finds that investors onaverage react very negatively to negative CSR events, and weakly negatively to positive events. The reaction is particularlystrong when the information concerns employees, the environment, or communities. Investors react positively to positivenews about CSR if it is related to firms’ poor stakeholder relations. In contrast, they react negatively to positive events when-ever such events are likely to be the result of agency problems. In a related approach, Dimson et al. (2013) show that CSRengagements by activist investors with U.S. public companies are followed by a 1.8% one-year abnormal return, on average.Subsequent to successful engagements which address environmental, social, and governance concerns, firms experienceimprovements in operating performance, profitability, efficiency, and governance.

Most studies do not explicitly distinguish between the cash flow effect and the risk effect of enhanced CSR. However, somestudies focus more heavily on the risk dimension, and claim that CSR only affects systematic risk due to the main insight fromportfolio theory that only systematic risk is priced in the financial markets. Other studies suggest that CSR is firm-specific, andtherefore affects only idiosyncratic risk. For example, Lee and Faff (2009) document a negative relationship between idiosyn-cratic risk and an aggregate measure of CSR, while Luo and Bhattacharya (2009) find a negative relationship between CSR andboth idiosyncratic and systematic risk. Bauer et al. (2009) construct employee relations indices and show that firms withstronger employee relations have lower costs, higher credit ratings, and lower idiosyncratic risk. Bouslah et al. (2013) showthat not all CSR dimensions are relevant for firm risk. Most importantly, they find that the direction of causation between afirm’s risk and CSR strengths or concerns differs across the individual CSR dimensions.

2.2. CSR behavior and CSR disclosure

The prior literature has made an important distinction between CSR behavior and CSR disclosure and the relationshipwith firm performance. Empirical studies on CSR behavior usually involve a distinction between CSR practices that have a


positive social impact and those that have a negative social impact, i.e., they analyze the quality and performance of the mea-sures. In contrast, because we observe that shipping firms have become increasingly willing to voluntarily provide CSR infor-mation in recent years in spite of the industry’s already strong institutional framework (see Section 3), our analysis focuseson firms’ voluntary CSR disclosures. There is a common concern about the usefulness of this type of disclosure because ofnon-comparability and potential credibility issues, as well as opportunistic firm behavior (Ingram and Frazier, 1980;Hobson and Kachelmeier, 2005). Nevertheless, as Margolis and Walsh (2003) and Margolis et al. (2007) suggest, rather thanjudging the quality of their CSR behavior, CSR disclosure practices can provide important information about a firm’s trans-parency. Dhaliwal et al. (2006) argue further that quality ratings alone are unlikely to provide sufficient information forinvestors to assess firms’ overall CSR performance. They conjecture that voluntarily disclosing CSR activities is a demonstra-tion of firms’ confidence in their CSR performance, which sends a positive signal to investors, or, in the case of poor CSR per-formance, allows firms to offer explanations. They conclude that CSR disclosures contain information beyond what theperformance ratings can provide.

Corporate executives seem to believe that voluntarily communicating information can reduce their firms’ cost of capital(Graham et al., 2005). Most prior research focused on the relationship between financial disclosures and the cost of capital,and found evidence of a negative relationship (see Dhaliwal et al., 2006, for a survey). Greater disclosure can increase investorawareness of a firm’s existence and enlarge its investor base, which improves risk sharing and reduces the cost of capital(Merton, 1987). Higher quality or more precise firm-specific disclosures further decrease the covariance of a firm’s cash flowswith those of other firms, reducing the betas of individual firms and the cost of capital (Hughes et al., 2007; Lambert et al.,2007). In addition, as Verrecchia (2001) and Amihud and Mendelson (1986) note, greater disclosure can lead to reduced infor-mation asymmetry among investors or between managers and investors, which in turn may induce lower bid-ask spreads andtransaction costs, as well as a lower cost of equity capital. These mechanisms likely apply to both financial and non-financialdisclosure, such as CSR disclosure, as long as the information is value-relevant. Ultimately, however, it remains an empiricalquestion whether voluntary CSR disclosure can result in lower investor expected returns and thus lower cost of capital.

Dhaliwal et al. (2006) analyze stand-alone CSR reports issued by U.S. firms. They show empirical evidence to support thehypothesis that a potential reduction in the cost of equity motivates firms to publish voluntary CSR reports. In fact, CSR dis-closure by firms with superior CSR performance leads to a lower cost of capital. Supporting Merton’s (1987) investor recog-nition theory, initiating firms with superior CSR performance attract dedicated institutional investors and more analystcoverage. Brammer and Pavelin (2006) distinguish between the decision to make a voluntary environmental disclosureand the decision about the quality of such a disclosure. Their empirical results show that both types of decisions matter,but there is significant cross-sectional variation in the determinants of both the participation and quality decisions. Finally,several studies document that the disclosure of social performance activities mitigates any negative stock market shocks,e.g., from a disaster or some new regulation (Blacconiere and Patten, 1994; Freedman and Stagliano, 1991). This effect seemsto be independent of the accuracy of the disclosed information.

In our empirical analysis, we construct a CSR disclosure index in an attempt to reconcile the diversity of the shippingindustry. We measure CSR disclosure and its relationship with a set of firm characteristics. The distinct nature of the ship-ping industry also makes our disclosure index a quality measurement index, at least to a large extent. As elaborated in Sec-tion 5, the majority of the items included in the index (19 out of 23 items) exert a positive social, environmental, safety, oremployment impact.

3. CSR in the shipping industry

3.1. CSR failures in the shipping industry

Several cases of corporate social responsibility failures caused huge costs to individual firms as well as society in recentyears. For example, in April 2010, a 230 m-long ship carrying 975 tons of heavy fuel oil and 65,000 tons of coal was travellingat full speed when it hit a sandbank in a protected part of the Great Barrier Reef. Its fuel tank ruptured, causing a three km-long oil slick. The spill resulted in the worst environmental disaster in Australia’s history. Moreover, consider the oil traderTrafigura, which was fined one million Euros in 2010 for illegally exporting tons of toxic waste to West Africa. It marked thefirst time this London-based firm was convicted of criminal charges in an environmental scandal, in which 30,000 Africanswere made ill when the toxic waste was dumped in Ivory Coast. Finally, four senior executive at Vinalines, a major state-owned Vietnamese shipping firm, were arrested for mismanagement in summer 2012. They were accused of causing lossesof $80 million in 2009 and 2010 after purchasing technologically obsolete ships and making poor investments.

The above examples are just a few of many cases that caused CSR in shipping to come under greater scrutiny, with imple-mentation procedures continuously monitored and periodically revised. It appears that shipping must respond to society’sincreased expectations for CSR behavior and disclosure in order to maintain its good image. Coady et al. (2013) provide adetailed overview of current CSR activities in the shipping industry and a review of industry perceptions of CSR initiatives.Given its very visible environmental issues, we conjecture that shipping firms are more likely to make voluntary disclosuresthan firms from other industries, where the benefits of disclosing such information are lower. Presumably, increased CSRreporting enhances shipping firms’ transparency and lowers information costs on the part of investors, potentially leadingto positive financial effects.


3.2. CSR implementation in the shipping industry

Due to its natural character and dispersed location, we can characterize the shipping industry as one of the most inter-nationally oriented industries. In fact, globalization is part of the nature of shipping. Originally, vessels were registered byship owners in their home countries, employees and officers were selected from home countries, and a large amount of cargowas transferred from and to these countries. However, largely due to the oil crisis of the 1970s, ship owners were forced toflag offshore. Employees and officers now come from anywhere in the world, and the goal is simply to reduce overhead andoperating costs. Through globalization, in fact, the shipping industry has retained and strengthened its position as a keyplayer in world trade, and it has played a critical role in moving the entire world toward greater globalization. As a result,the shipping industry operates under a special international institutional framework.

The first pillar of this international institutional framework is the IMO (the International Maritime Organization). It wascreated after World War II, and was formerly known as the IMCO until 1982. Its main purpose is to strengthen and supportthe ‘‘safety of life at sea’’ internationally (referred to as SOLAS). The second institutional pillar is the ILO (the InternationalLabor Organization), initially based on the U.S. Seamen’s act in 1915. It was created following the International Labor Con-ference in Geneva in 1920. Its main purpose is setting shipboard standards for seamen (i.e., the Seamen’s Code).

At a national level, the flag and port state controls, supported by classification societies, constitute an integral part of theinstitutional shipping framework. A flag state, defined by the UN Convention on the Law of the Sea (UNCLOS), has the respon-sibility for the implementation and enforcement of international maritime regulations for all ships granted the right to fly itsflag. It is a nation’s role as a flag state to be within the first line of defense against potentially unsafe or environmentally dam-aging ship operations. However, the flag state conducts many of its activities through entities located in other countries.Most national maritime administrations have multiple roles in their capacity as port and coastal states, which involve theenforcement of flag regulations through visiting foreign ships. The enforcement of IMO rules, aimed at supporting the imple-mentation of SOLAS and the International Seamen’s Code, as well as the management of IOPC (International Oil PollutionCompensation Fund) is the responsibility of IMO member governments in their capacity as flag states.

Classification societies are organizations that closely regulate a ship’s life during the design, building, operation, anddemolition stages. They therefore play a major role as drivers of technological innovation. More than 90% of the world’s bulkcarrier and tanker fleets are covered by classification societies, with construction as well as through-life compliance rulesand standards set by the thirteen member societies of the IACS (the International Association of Classification Societies Ltd.).

Finally, the port state control (or PSC) also has a major role in the context of CSR enforcement practices. It inspects andvisits foreign ships to verify their condition and to ensure that they are complying with international rules on safety, pollu-tion prevention, and seafarer living standards and working conditions. PSC also enforces compliance where the owner and/orflag state have failed in their responsibility to implement or ensure compliance. It can require that deficiencies found duringan inspection be rectified at either the next port of call or prior to departure. For very serious deficiencies, it can even detain aship. Therefore, PSC also serves as a port state’s defense against substandard ships visiting their ports.

Overall, we believe that the organizational structure of the shipping industry is more developed than many land-basedindustries. CSR behavior is vital and necessary for the safety of human life and cargo, and the prevention of environmentalpollution. Due to the fact that national governments cannot impose universal shipping rules and regulations, the implemen-tation of CSR in the shipping industry is to a large extent assured by the different institutional bodies involved. We hypoth-esize that a well-devised CSR strategy can be used by shipping firms for communication purposes with their customers. CSRdisclosures by shipping firms about issues and procedures whose effectiveness and implementation are already monitored bythe different institutional bodies may nevertheless contribute to a better overall reputation. They could also provide advan-tages during crisis periods, e.g., financial benefits by providing access to loans (Det Norske Veritas, 2004; Coady et al., 2013).

4. Testable hypotheses on the relationship between CSR and firm characteristics

We conjecture that voluntary CSR disclosure is an attempt to mitigate the information risks faced by potential and actualinvestors (see Section 2). If these risks (and the associated costs) vary across shipping firms, we also expect CSR disclosure todepend on a set of firm characteristics that influence the relative costs and benefits of disclosing such information. For exam-ple, firm size is related to the degree of asymmetric information (Banz, 1981; Blank et al., 2009). Large firms are more visibleto the general public, thus they tend to be subject to greater political and regulatory pressure from external interests. As aresult, large firms use CSR disclosures to demonstrate that their actions are legitimate and consistent with good citizenship(Brammer and Pavelin, 2006). In fact, CSR disclosure may also act as a counterweight to the potential political and socialpressures faced by relatively large firms, allowing them to resist further regulations and procedures that often result inhigher operating costs (Cooke, 1989). Larger firms are also more likely to use formal channels of communication to dissem-inate information about corporate activities. Therefore, our first hypothesis is:

Hypothesis 1. There is a positive relationship between a shipping firm’s size and the level of its corporate socialresponsibility disclosure.

A firm’s financial leverage is likely to be associated with its level of CSR disclosure. Recently, shipping firms, especiallywithin the dry bulk and container sectors, have come under increasing pressure from lenders and shareholders to provide


more detailed and more frequent financial disclosures. The need for more extensive disclosure is related to the inherentvolatility of the shipping markets (Drobetz et al., 2012), which implies a high level of uncertainty about shipping firms’revenues. As a result of these large shifts in market trends, there can be a growing mismatch between a ship’s book and mar-ket values (vessel price risk), and both lenders and shareholders are experiencing increased uncertainty about their invest-ment prospects (Albertijn et al., 2011).

‘‘Slack’’ (defined as financial flexibility due to low leverage; Graham and Harvey, 2001) may promote CSR disclosure sim-ply by providing the resources to meet the resulting administrative costs (‘‘slack resources theory’’; see McGuire et al., 1988).In addition, firms with lower leverage may experience less pressure from their creditors. Their financial flexibility makes iteasier to fund CSR disclosures, and they may also have the discretionary ability to focus on organizational activities that areonly indirectly linked to the financial success of the firm. In a non-shipping-related context, these arguments are corrobo-rated by the empirical results in Leftwich et al. (1981), Ferguson et al. (2002), and Brammer and Pavelin (2006). As CSR dis-closure adds an extra burden to highly leveraged firms, it may trigger hidden resistance from shareholders. These firms tendto suppress CSR expenses, and thus only report the minimum required items to limit their costs and to protect shareholders’residual claim. Accordingly, our second hypothesis is:

Hypothesis 2. There is a negative relationship between a shipping firm’s financial leverage and the level of its corporatesocial responsibility disclosure.

It is likely that similar arguments apply to the relationship between CSR disclosure and a firm’s profitability and valua-tion. When the profit rate is sufficiently high, a firm is more likely to possess the economic ability (or slack) to disclose CSRinformation; in this case, the firm is reassuring its shareholders about its healthy financial position, and the managers of thefirm may exploit it with respect to their own remuneration (Marston and Polei, 2002). For the same reason, one expects thata very high profit rate makes a firm more vulnerable to public inspection of its financial statements and its practices. Dis-closure of a sustainability report can thus serve as protection (Marston and Polei, 2002). As profit-seeking organizations,firms may also want to expand the social acceptance of their operations, especially to the extent that these operations invokenegative externalities on their stakeholders and the environment (Palazzo and Scherer, 2006). Brammer and Pavelin (2006)confirm a positive relationship between profitability and voluntary environmental disclosure.

Given the natural link between profitability and firm valuation in standard discounted cash flow (DCF) valuation models,we develop arguments that are borrowed from standard asset pricing theory to rationalize a link between CSR disclosure andfirm valuation levels. In particular, we posit that CSR disclosure is motivated by shipping firms’ attempts to reduce the risksfrom misinformation (asymmetric information risk), thereby attracting shareholders with a low cost of equity capital. As aresult, these firms may be considered less risky as investments, and investors may demand a lower (expected) rate of returncompared to non-disclosing firms. Because investors assign a lower discount rate to expected future cash flows for CSR dis-closing firms (due to their lower cost of adverse selection), these firms could benefit from higher valuation levels (e.g., insubsequent equity issuances; Dhaliwal et al., 2006). Based on these arguments for a relationship between firm valuation,profitability, and CSR disclosure, we test our third hypothesis:

Hypothesis 3. There is a positive relationship between the valuation level as well as the profitability of a shipping firm andthe level of its corporate social responsibility disclosure.

One caveat with this hypothesis is that it implicitly assumes that the market prices CSR disclosure efficiently. As Hamiltonet al. (1993) note, investors may find it overly complex to value the benefits and costs effectively, thus the relevant informa-tion may be only slowly reflected in stock prices. In particular, the stocks of CSR disclosing firms may be initially undervaluedrelative to those of non-CSR disclosing firms, and they may ultimately produce higher (ex post) risk-adjusted returns toachieve the correct valuation level (and the correct lower expected rate of return). We assume that the market has gainedsufficient experience over the years to efficiently assess the costs and benefits of firms with high CSR disclosure. Therefore,we do not examine this alternative ‘‘disequilibrium hypothesis’’ any further here.

Finally, the standard principal-agent model (Jensen and Meckling, 1976) predicts that when ownership is dispersed,adverse selection problems between managers and owners will be particularly strong. With a more disparate ownershipstructure, minority shareholders have little ability or incentive to monitor managers, resulting in a high degree of informa-tion asymmetry. Therefore, one would expect that a diffused ownership structure will provide an incentive for a firm to vol-untarily provide information to shareholders through CSR disclosures (Brammer and Pavelin, 2006). However, the shippingindustry in this respect is again distinct. Andrikopoulos et al. (2013) show that the degree of equity concentration positively(rather than negatively) influences the amount of financial information reported by shipping firms. Tsionas et al. (2012) doc-ument that the degree of ownership concentration is positively correlated with profitability in the shipping industry, thus weconjecture that the amount of CSR reporting increases with more concentrated ownership. In additions, ship owners want tobuild up and maintain their reputation in order to increase their chances of survival during crisis periods. Assuming that(founder) managers with high ownership stakes will attempt to hedge their idiosyncratic risks and protect their firm-specificinvestments (Stulz, 1996; Kim et al., 2013), we test our final hypothesis:


Hypothesis 4. There is a positive relationship between the level of a shipping firm’s ownership concentration and the levelof its corporate social responsibility disclosure.

5. Construction of the CSR disclosure index

Our initial sample consists of 118 shipping companies listed on major international stock markets, such as the NYSE, theNASDAQ, the London Stock Exchange, the Singapore Stock Exchange, and the Oslo Stock Exchange. The sample includesroughly two-thirds of all shipping companies listed on these markets. The remaining listed companies had only limitedinformation on their websites, thus we were unable to obtain the necessary information. We generally include firms inour sample if they own and/or operate merchant ships (excluding tourism, oil drilling, and conglomerate companies whereshipping is not the major activity). Therefore, in our sample, the merchant ships are tankers, dry bulk shippers, and contain-ers. To construct our CSR disclosure index, we search the websites and annual reports of our sample firms for each yearbetween 2002 and 2010. We include a total of 23 CSR items in the index, many of which are related to the best-practiceCSR framework for the shipping industry and its key performance areas (as per Coady et al., 2013). For each item, we simplyassign a value of one if it is reported, and zero otherwise. Accordingly, the index values for each firm and for each year rangebetween a minimum of zero and a maximum of 23. Table 1 lists the items included in our CSR index.

As shown in Fig. 1, the most frequently reported CSR item during our sample period is ‘‘accidents at work.’’ The next mostfrequently disclosed items are: job creation; employment opportunities and quality certification; health and safety in theworkplace; gifts, sponsorships, and investor relations; environmental certification; and human capital development. Fig. 2shows that CSR reporting nearly doubled during our sample period, thus CSR disclosure seems to become the norm amongshipping firms.

6. Data description and empirical methodology

6.1. Data collection

We collect financial data for the fiscal years from 2002 to 2010 (as of December 31), obtained from company records andwebsites. Panel A of Table 2 provides the definitions for all variables in our empirical analysis; all accounting figures are mea-sured in millions of dollars and double-checked by means of the Compustat Global database (whenever a firm is covered).Following Ettredge et al. (2002) and Debreceny et al. (2002), we approximate firm size by using the logarithm of a firm’s totalassets in each year (LTA). As in Leuz and Verrecchia (2000) and Dhaliwal et al. (2006), we measure the financial performanceof a firm by using Tobin’s Q (approximated by the market-to-book ratio, Tobin’s Q), and operating performance (profitability)

Table 1Items included in the CSR disclosure index.

# Items Definition

1. Employment opportunities Conditions of promotion2. Job creation Notifications for open seats3. Human capital development Training of employees4. Accidents at work Injuries and deaths5. Health and safety in the work

placeOHSAS 18001, safety regulations regarding human capital

6. Equal opportunities No discrimination on sex, religion, etc.7. Employee involvement Stock options to employees8. Regional development Contribution to local communities9. Gifts and sponsorships Gifts to the community (e.g., a library to a university) and sponsoring events

10. Service/product related incidents Piracy is considered a common problem for ships, so it was taken under consideration11. Service/product safety Safety regulations regarding service/product12. Human rights No discrimination, prohibition of child labor, care for vulnerable groups, etc.13. Community involvement Blood donations, raising awareness events, etc.14. Investor relations Assistance of current and/or new investors to learn about firm activities15. Code of conduct Rules that apply in dealing with employees16. Business ethics Awards earned for governance system, remuneration, compliance, internal transparency, corruption, and

ownership structure17. Quality certification ISO 9000 family of standards18. Environmental policy Plans and actions regarding the reduction of corporations effects on the environment19. Ship recycling Recycling of ship parts20. Environmental certification ISO 14001/14004/1406421. Discharge Ballast water, hazardous materials, sewage, exhaust emissions, and oil22. Corporations effects on the

environmentSize of environmental footprint left by the company’s actions in a year, biodiversity

23. Port detention Number of times any member of the fleet has been detained in various ports

0

100

200

300

400

500

600

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23T

otal

num

ber

of r

ecor

ding

sCSR disclosure index item

Fig. 1. Total number of recordings of CSR disclosure index items. This figure is based on the initial sample of 118 sample companies.

0

100

200

300

400

500

600

700

800

2002 2003 2004 2005 2006 2007 2008 2009 2010

Num

ber o

f rec

ordi

ngs

Sample year

Fig. 2. Recordings of all CSR disclosure index items by year. This figure is based on the initial sample of 118 sample companies.

Table 2Data description (final sample of 111 firms)

Panel A: Definition of variablesCSR CSR disclosure index. Total score for all 23 items (see Table 1)LTA Size. Natural logarithm of total assetsDEBT Leverage. Ratio of total liabilities over equityTobin’s Q Financial performance. Market-to-book ratioOWNERSHIP Concentrated ownership. Percentage of stakes held by the largest shareholderROE Return on equity. Ratio of net income over equityROA Return on assets. Ratio of net income over total assets

CSR LTA DEBT Tobin’s Q OWNERSHIP ROE ROA

Panel B: Descriptive statisticsMean 5.035 6.544 1.387 1.177 0.350 0.185 0.069Median 4.000 6.554 0.785 1.058 0.311 0.137 0.058Maximum 23.00 9.847 14.940 9.449 0.932 5.810 0.884Minimum 0.000 1.652 0.006 0.353 0.010 �4.673 �0.656Std. dev. 4.448 1.364 1.919 0.700 0.189 0.693 0.115


by either the return on equity (ROE) or the return on assets (ROA). Moreover, we measure leverage as the ratio of totalliabilities to equity (DEBT), and concentrated ownership as the proportion of equity held by the largest shareholder(OWNERSHIP). Panel B of Table 2 shows the descriptive statistics for these variables. Our dataset was initially 118 firms,but after excluding firms without the necessary data for constructing Tobin’s Q, our final sample comprises only 111 shippingfirms.

6.2. Econometric models

In our empirical analysis, we use Generalized Smoothly Mixing Regression (GSMR) models to capture the observed het-erogeneity in our index (Geweke and Keane, 2007; Villani et al., 2012). We begin with an intuitive description of our analysis,and proceed with a more formal discussion of this methodology.


6.2.1. Intuition of the analysisLet yit denote the dependent variable (CSR disclosure index), where xit is a k � 1 vector of explanatory variables. Note that

the regression model constitutes the workhorse in most empirical research. In the form yi ¼ x0ibþ ui, where the disturbancesare normally distributed with zero mean and constant variance, the linear regression model states: yijxi � N x0ib;r2

� �. There-

fore, the covariates have an effect on the mean of yi, which is given by the regression coefficients b. However, for most prac-tical purposes, the essential question is how the distribution of yi changes with variations in xi. This question is clearlyimportant for a change in the regressors that does not necessarily exhaust its effects on the mean of yi but rather on theentire distribution. To uncover such effects, we need to abandon the linear regression framework and adopt more generalmodels, such as a mixture-of-normals formulation, which can approximate many distributions. Instead of assuming thatyijxi � N x0ib;r2

� �, we can make the following assumption, using some notation adjustments:

yijxi � p1N x0ib1;r21

� �þ p2N x0ib2;r2

2

� �þ � � � þ pMN x0ibM;r2

M

� �¼XM

m¼1

pmN x0ibm;r2m

� �: ð1Þ

In this formulation, yi|xi is allowed to follow a normal distribution N x0ibm;r2m

� �, with given coefficients bm and variance r2

m butonly with probability pm. An alternative interpretation would be that the presence of heterogeneity is described by M differ-ent groups in the data. An observation (yi, xi) belongs to one of these groups, but the classification is unknown, i.e., an obser-vation belongs to each group with a certain probability. If we assume for simplicity that we have only one regressor, then theexpected value of yi|xi is:

EðyijxiÞ ¼XM

m¼1

pmbmxi; ð2Þ

and, therefore, the effect of the regressor on the mean is:

@EðyijxiÞ@xi

¼XM

m¼1

pmbm; ð3Þ

which is a weighted average of the coefficients b1, . . . , bM (the group-specific effects), with the weights being the probabil-ities. Although the mixture-of-normals model is flexible in terms of modeling mean effects, the conditional variance isvarðyijxiÞ ¼

PMm¼1 pmr2

m ¼ r2� , and is independent of the regressors. This restriction can be solved by assuming that each var-

iance is also a function of the regressors:

logr2m;i ¼ x0iam; m ¼ 1; . . . ;M: ð4Þ

Moreover, we can make the classification probabilities functions of the covariates instead of assuming that they are fixed, i.e.,there is a fixed (but unknown) proportion of observations that belong to each group. This change can be implemented byusing the logit transformation:

logpmi

1� pmi¼ x0icm; m ¼ 1; . . . ;M; ð5Þ

which ensures that pmi1�pmi

¼ expðx0cmÞ, which is always positive, or pmi ¼exp x0

icmð Þ

1þexp x0icmð Þ, which is always between zero and one. As a

result, all parameters of the mixture-of-normals distribution are now functions of the observed covariates, thus the finalmodel can be written, again with an adjusted notation, as follows:

yijxi �exp x0ic1

� �1þ exp x0ic1

� � � N x0ib1; expðx0ia1Þ� �

þ � � � þexp x0icM

� �1þ exp x0icM

� � � N x0ibM; exp x0iaM� ��

¼XM

m¼1

exp x0icm

� �1þ exp x0icm

� � � N x0ibm; exp x0iam� ��

: ð6Þ

The question arises what can be gained by making the parameters of the mixture-of-normals distribution a function of allcovariates. If we assume, again for simplicity, that we have only one regressor, the expressions for the conditional mean andthe conditional variance become:

EðyijxiÞ ¼XM

m¼1

expðcmxiÞ1þ expðcmxiÞ

� ðbmxiÞ; ð7Þ

varðyijxiÞ ¼XM

m¼1

expðcmxiÞ1þ expðcmxiÞ

� expðamxiÞ: ð8Þ

Now it is clear that the conditional expectation and the conditional variance are non-linear functions of the regressor xi, andthe effects @Eðyi jxiÞ

@xiand @varðyi jxiÞ

@xiare not constants and are not as simple to derive as they were initially. A mixture-of-normals can

approximate any distribution (if we allow the number of groups of classes M to increase). Therefore, the effects on the mean,


@Eðyi jxiÞ@xi

, can be approximated with greater accuracy if we depart from the classical linear regression model. Of course, the same

holds for the ‘‘second-order’’ effects of the regressor on the variance, @varðyi jxiÞ@xi

, which are similarly important but less

frequently the subject of empirical work. In fact, our testable hypotheses in Section 4 do not specifically emphasizehigher-order effects. Given the heterogeneous CSR disclosure practices across listed shipping companies, however, we alsoreport and interpret the marginal effects of firm characteristics on the conditional variance of our CSR disclosure index.

Our model, known as Smoothly Mixing Regression (SMR), is very flexible and provides a large amount of information.With only a single regressor, the mean effect, @Eðyi jxiÞ

@xi� bi, is different for each observation (e.g., firm or time period, depending

on the nature of the data). More precisely, this effect is a function of the regressor, with @Eðyi jxiÞ@xi� bi ¼ biðxiÞ. We can compute

the mean effect for a given observation whose regressor takes the observed value xi, but we can also compute this effect for a

different value of the same firm, say @Eðyi jxi¼xoÞ@xi

¼ biðxoÞ. Therefore, the marginal effect on the expected value is in fact a func-

tion. Accordingly, the regression function in the SMR model, defined as f(x) = E(yi|xi = x), is no longer a linear function.If there are two regressors, we also have two marginal effects, @Eðyi jxi1 ;xi2Þ

@xi1¼ b1ðxi1; xi2Þ and @Eðyi jxi1 ;xi2Þ

@xi2¼ b2ðxi1; xi2Þ. At the

observed values of the regressors, we can easily compute the marginal effects (after the model has been estimated), andexamine:

1. Values of the marginal effects b1(xi1, xi2) and b2(xi1, xi2).2. Marginal sample distributions of the effect b1(xi1, xi2), which reveal the extent of heterogeneity in the sample (i.e., the

magnitude of differences in the marginal effects). If the sample distributions are concentrated around a given value,we should find no heterogeneity.

3. Bivariate distributions of the marginal effects b1(xi1, xi2) and b2(xi1, xi2), which reveal the most likely or representative val-ues of these effects, along with their spread or cross-sample variation. They can again be used to gain some appreciationof the existence and extent of heterogeneity.

All of these analyses attempt to understand the features of the joint distribution of marginal effects b1(xi1, xi2) andb2(xi1, xi2). When the number of regressors increases, the multivariate distribution of the effects becomes more difficult tovisualize, but we can still focus on univariate and bivariate distributions of the marginal effects to understand the complexinfluence of the regressors on the dependent variable. Similar analyses can be implemented to understand the underrated‘‘second-order’’ effects of the regressors on the variance. Our model can also be used to examine the effect of regressorson the log-odds of classification, i.e., the formulation log pmi

1�pmi¼ x0icm, which provides an automatic way to examine the influ-

ence of regressors on the classification into different groups and comprises the material foundation of heterogeneity in thesample. For example, a positive cmk indicates that xik contributes positively to the classification of the ith observation intogroup or class m e {1, . . . , M}. Clearly, we could have c1k > 0 and c2k < 0, implying that the kth regressor contributes positivelyto the classification into class 1, but negatively into class 2. As a result, class 1 should consist mostly of observations withhigher values of the kth regressor, while class 2 should consist mostly of observations with lower values of the kth regressor.

6.2.2. Formal discussionIn this section, we provide a formal discussion of our econometric model. Specifically, the SMR model is as follows:

yit jxit ; h � N x0itbm;r2m

� �; ð9Þ

with probability pm(xit, c), m = 1, . . . , M, parameterized as log pmðxit ;cÞPM

s¼1psðxit ;cÞ

¼ x0itcm;m ¼ 1; . . . ;M, where h generally denotes all

unknown model parameters. For identification purposes, we set bm = 0(k�1) and pmðxit; cÞ ¼ 1�PM

s¼1 psðxit ; cÞ. Therefore, thedependent variable, our CSR disclosure index, is a location-scale mixture-of-normal distributions. The means are given byregression models with coefficients bm for group m, different variances r2

m, and group probabilities pm that vary with theregressors. Specifically, note that the log-odds ratio is a linear function of the regressors. The heteroskedastic or generalizedSMR (GSMR) model is then given by:

yit jxit ; h � N x0itbm;r2m;itðxit; dÞ

� �;with probability pmðxit; cÞ; and

logr2m;itðxit ; dÞ ¼ x0itdm; m ¼ 1; . . . ;M: ð10Þ

The conditional density of observations is:

f ðyitjxit; hÞ ¼XM

m¼1

2pr2m;itðxit ; dÞ

� ��1=2exp �

yit � x0itbm

� �2

2r2m;itðxit; dÞ

" #pmðxit ; cÞ; ð11Þ

from which the log-likelihood function is:

log Lðh; y;XÞ ¼Xn

i¼1

XT

t¼1

log f ðyit jxit ; hÞ: ð12Þ


The log-likelihood function can be maximized using standard numerical techniques. The posterior probability of classi-fication into group m e {1, . . . , M} for a particular observation is given using Bayes’ theorem as:

Pmðxit; hÞ ¼2pr2

m;itðxit ; dÞ� ��1=2

exp � yit�x0itbmð Þ2

2r2m;itðxit ;dÞ

� �pmðxit; cÞ

PMs¼1 2pr2

s;itðxit; dÞ� ��1=2

exp � yit�x0itbsð Þ2

2r2s;itðxit ;dÞ

� �psðxit; cÞ

; m ¼ 1; . . . ;M; ð13Þ

where we can substitute the maximum likelihood parameter estimates for unknown parameters. The meaning of thisexpression is that the ‘‘prior’’ classification probability, pm(xit, c), is combined with the likelihood of the observation,

2pr2m;itðxit; dÞ

� ��1=2exp � yit�x0

itbmð Þ2

2r2m;itðxit ;dÞ

� �, to yield the ‘‘posterior’’ classification probability, Pm(xit, h). The denominator is simply

the normalizing constant.The GSMR allows for dependencies between our CSR disclosure index and the covariates not captured by relationships in

the mean, because it also incorporates possible effects on the variances of the classification groups. Another important aspectis that observations can be classified into groups using the posterior probabilities (for example, an observation belongs togroup 1 if Pmðxit; hÞP 1

2, with two groups in total). Moreover, the effects of covariates can be assessed using either the esti-mated coefficients b1, if the observation belongs to the first group, or b2 if the observation belongs to the second group. How-ever, we believe it is more suitable to calculate the effects of covariates in a different way. For the GSMR model, we have:

Eðyit jxit; hÞ ¼ x0itb ¼ x0itXM

m¼1

pmðxit ; cÞbm: ð14Þ

When the covariates change, we can estimate the marginal effects on the conditional expectation as follows:

rEðyitjxit ; hÞ ¼ bþ x0itXM

m¼1

rpmðxit; cÞbm; rEðyitjxit; hÞ �@Eðyitjxit ; hÞ

@xit: ð15Þ

There are thus two effects at work: first, the immediate effect, �b, which is the average coefficient; and second, the effect thata change in covariates induces a change in the classification probabilities. Furthermore, given the conditional variance:

Varðyit jxit; hÞ ¼XM

m¼1

pmðxit ; cÞr2m;itðxit; dÞ; ð16Þ

the marginal effects are:

rVarðyitjxit ; hÞ ¼XM

m¼1

rpmðxit; cÞr2m;itðxit ; dÞ þ

XM

m¼1

pmðxit; cÞrr2m;itðxit; dÞ: ð17Þ

In the SMR case, the effects would simplify to:

rVarðyitjxit ; hÞ ¼XM

m¼1

r2mrpmðxit; cÞ: ð18Þ

To summarize, a change in the covariates of GSMR models induces changes in both the conditional expectation and theconditional variance – changes that cannot be captured by linear models and reveal the potentially rich structure of depen-dencies between our CSR disclosure index and its explanatory variables. Without compromising the simplicity of linear mod-els, GSMR formulations provide a semi-parametric way to estimate complicated features of the conditional distribution thatcan be used for further analysis (Park et al., 2010).

6.5. Posterior analysis

To perform the computations, we use an adaptive Metropolis–Hastings algorithm (Tierney, 1994), which resembles theGibbs sampler in that it updates each parameter separately. The proposal distribution for each parameter is uniform, and in aclosed interval of the form [a, b]. During the transient, or ‘‘burn in’’ phase (consisting of one million iterations), the boundsare adjusted so that Metropolis–Hastings acceptance rates are between 20% and 30%. We take another 100,000 iterations,from which we retain only every other tenth in order to mitigate the impact of autocorrelation. We test convergence duringthe transient phase using Geweke’s (1992) diagnostics. These results are available upon request.

Fig. 3 illustrates the posterior classification probabilities into group or class 1 (with two groups in total) from both the SMRand the GSMR (heteroskedastic) model. If these probabilities are the same, all points in the graph should lie on the 45 degreeline. Most probabilities are quite similar in the extreme cases (with probabilities near one or zero). However, there is somedisagreement in the intermediate cases. We conclude that there is an important difference between homoskedastic and het-eroskedastic models, thus for the rest of our analyses we continue with the GSMR model to derive the marginal effects.

A caveat is that our analysis cannot unambiguously solve the issue of causality. In fact, causality is a general problem thatplagues the empirical corporate governance literature (Cremers and Ferrell, 2014; Wintoki et al., 2012). The main limitation

Fig. 3. Posterior probabilities of classification from SMR and (heteroskedastic) GSMR models.


of our methodology is that we are unable to test the bidirectional nature of the link between CSR disclosure and corporateperformance. Nevertheless, it allows us to analyze the heterogeneity in shipping firms’ CSR disclosure and to investigate howit relates to financial performance and other firm-level as well as industry-level characteristics.

7. Estimation and discussion of results

7.1. Main results

Fig. 4 shows sample distributions of the marginal effects of the explanatory variables on the conditional mean and theconditional variance of our CSR disclosure index. For example, the upper left diagram with the conditional mean effectsillustrates the sample distributions of dE(CSRi)/dLTAi = bi. Depending on the skewness of the distribution, the majority offirm-level effects will be either positive, negative, or zero. In the case of firm size (LTA), all marginal effects are positive. Fur-thermore, the lower left diagram shows the distributions of the effect of LTA on the conditional variance of our CSR disclosureindex. Again, this effect is positive in all cases. Overall, total assets (our proxy for firm size) exhibit positive mean and var-iance effects, implying that size contributes positively to shipping firms’ CSR disclosures, but the variance also increases.

The conditional mean effect of Tobin’s Q is positive and concentrated around 0.4, while for a change in Tobin’s Q, on aver-age, there is no significant change in the variation of CSR disclosure. The conditional mean effect of concentrated ownership(OWNERSHIP) is mostly positive, and the conditional variance effect is strongly positive. Finally, financial leverage (DEBT) hasa negative effect on both the conditional mean and the conditional variance, and the distributions are highly asymmetric. Infact, the posterior distributions are generally asymmetric, suggesting that the assumption of normality is not valid. This evi-dence indicates that the conditional distribution of our CSR disclosure index given the covariates is complex and cannot besummarized by using linear regression analysis, where the marginal effects are constant and the same for all firms, andwhere the variance is independent of the covariates.

Our sample distributions do not exhaust all the information we can derive from the data regarding the relationshipbetween our CSR disclosure index and the covariates. Each firm exhibits a different marginal effect, but the parametersare unknown and must be estimated using MLE. To cope with parameter uncertainty, we implement Markov Chain MonteCarlo (MCMC) techniques for Bayesian inference. We use flat priors for all parameters in the interval [�100, 100]. MCMC isimplemented using 500,000 passes, where the first 400,000 are discarded to mitigate start-up effects and any possible non-convergence. To reduce the impact of autocorrelation, we thin out every other tenth draw to produce a final set of 10,000parameter draws from the posterior distribution of the GSRM model. We assess the marginal effects on the conditional meanand the conditional variance for each firm. The posterior distributions of marginal effects are illustrated in Fig. 5, and Table 3shows the MCMC results for each firm.

Table 3 shows the conditional mean and variance effects (MCMC) for each firm. Given the strongly positive link betweenCSR disclosure and Tobin’s Q (see Fig. 4), we expect this relationship to be positive for each firm in our sample. This findingwould then provide support for our main hypothesis that shipping firms have integrated CSR disclosure practices into theirstrategic planning and operations. In fact, we observe that the marginal effect of Tobin’s Q (our measure for financial

Fig. 4. Sample distributions of marginal effects: Main results.


performance) on the conditional mean of our CSR disclosure index is positive for all firms in our sample. The marginal effectof Tobin’s Q on the conditional variance of the index is also positive for all firms. Overall, these results corroborate the firstpart of our Hypothesis 3, suggesting a positive relationship between a shipping firm’s valuation level and its amount of CSRdisclosure.

Confirming Hypothesis 1, Table 3 further shows that the marginal effect of firm size (LTA) on the conditional mean of ourCSR disclosure index is positive for all sample firms, albeit with considerable variability. The marginal effect of financialleverage (DEBT) on the conditional mean of CSR is negative for all sample firms, supporting Hypothesis 2 that there is a neg-ative relationship between a shipping firm’s leverage and its amount of CSR disclosure. Similarly, the impact of financialleverage on the conditional variance of CSR disclosure is negative for all firms in our sample, suggesting that there is reducedvariation across firms from their respective means with increasing leverage. As expected under Hypothesis 4, the marginaleffect of concentrated ownership (OWNERSHIP) on the conditional mean of CSR disclosure is positive in the majority of cases(107 firms); it is negative for only four firms. CSR disclosure increases for ship owners with higher ownership stakes, whichmay be part of the corporate risk management process to mitigate idiosyncratic risks faced by dominant ship owners (Stulz,1996; Kim et al., 2013).

Finally, to confirm the statistical significance of our results, Table A1 in Appendix A shows the 95% Bayes probability inter-vals for the mean and variance effects of Table 3. For example, the first line of the table is interpreted as follows: for samplefirm 1, we are 95% confident that the marginal effect of LTA on the mean of our CSR disclosure index lies between 0.82 and1.48. Moreover, the marginal effect of LTA on the variance of the index lies within 0.27 and 0.58 with a 95% probability.

7.2. Robustness tests

As a robustness check, we repeat our analysis using the return on equity (ROE) and the return on assets (ROA) as our oper-ating performance measures (replacing Tobin’s Q as a financial performance measure). To test this second part of our Hypoth-esis 3, we use the initial sample of 118 firms in the remaining analyses, for which the two profitability measures areavailable. Figs. 6 and 7 illustrate the joint sample distributions of the marginal effects on our CSR disclosure index fortwo relationships (LTA versus ROA or ROE, and DEBT versus OWNERSHIP). In particular, following Geweke and Keane

Fig. 5. Posterior distributions of marginal effects for all firms (MCMC): Main results.


(2007), we present aspects of the multivariate distribution of mean and variance effects in the form of bivariate distributions.The darker reddish colors denote areas of higher posterior probability mass (location of the posterior mode). For example,the upper left panel in Fig. 6 depicts (contours of) the joint posterior distributions of the marginal mean effects of ROAand LTA. In classical econometrics, this would correspond to the contours of the joint distribution of b1 and b2. The point(0,0) receives zero posterior probability mass, and the distributions are reasonably symmetric compared to the distributionsof the mean effects shown in Fig. 5. The lower panels in Figs. 6 and 7 indicate the non-normality of the marginal effects forthe variance, as evidenced both by the curvature and the location of the posterior mode toward the ‘‘tails’’ of the graphs.

Comparing Figs. 6 and 7, we assess that ROA is superior to ROE for capturing the joint effects on our CSR disclosure index.In fact, ROA generally seems to be a better metric of operating performance than ROE. It explicitly takes into account theassets used to support business activities, and it determines whether a firm is able to generate an adequate return on theseassets rather than simply showing a return on equity (Hagel et al., 2010). Asset-intensive firms like those in the shippingindustry require a higher level of net income to support their businesses than asset-light firms, where already thin marginscan generate a healthy return on assets. Even if our results for the operating performance covariates ultimately turn out to bemarginal, we consider this test important because it may reveal insights into how the shipping industry operates.

As discussed, the sample distributions do not exhaust all the information we can obtain from the data on the associationbetween our CSR disclosure index and the covariates. Therefore, we again implement Markov Chain Monte Carlo (MCMC)techniques for Bayesian inference. The posterior distributions of the marginal effects are shown in Fig. 8. For shipping firms,a higher ROA is attributable to (i) operational improvements through a reduction in the non-operating days of ships, leadingto an increase in cash inflows, and/or (ii) financial effects through lower financing costs. We expect the relationship betweenCSR disclosure and ROA to be positive for each firm. However, we find that the marginal effect of ROA on the conditionalmean of our CSR disclosure index is positive for 61 firms in our sample, and it is negative for the remaining 57 firms. Theimpact on the conditional variance of an increase in ROA is negative for each firm. Our methodology of modeling heteroge-neous behavior indicates that, despite the fact that for the majority of firms the relationship between CSR disclosure and ROAis positive, there nevertheless remains an unexpectedly high percentage of firms where it is negative.

To shed further light on this issue, we examine whether firms with a negative relationship differ with respect to size,leverage, or ownership concentration. As we cannot definitively determine such differential characteristics, we proceed

Table 3Conditional mean and variance effects (MCMC) for all firms.

Firm Conditional mean effect Conditional log-variance effect

LTA Tobin’s Q DEBT OWNERSHIP LTA Tobin’s Q DEBT OWNERSHIP

1 1.18 0.58 �0.70 2.49 0.43 0.18 �0.22 1.772 0.80 0.39 �0.39 1.25 0.33 0.12 �0.10 1.323 0.88 0.43 �0.47 1.52 0.36 0.14 �0.14 1.434 1.61 0.80 �1.04 3.85 0.56 0.25 �0.37 2.345 1.32 0.63 �0.81 2.86 0.51 0.20 �0.28 2.256 1.73 0.86 �1.16 4.11 0.56 0.26 �0.39 2.417 1.23 0.61 �0.74 2.69 0.44 0.18 �0.24 1.758 0.69 0.35 �0.32 0.92 0.29 0.11 �0.09 1.039 1.17 0.59 �0.69 2.55 0.40 0.17 �0.21 1.56

10 1.15 0.58 �0.66 2.59 0.37 0.16 �0.19 1.4311 1.71 0.86 �1.14 3.96 0.58 0.26 �0.40 2.5012 1.39 0.70 �0.85 3.44 0.43 0.19 �0.26 1.7213 1.60 0.83 �0.99 4.59 0.47 0.23 �0.33 1.9514 1.54 0.77 �0.97 3.91 0.48 0.22 �0.31 1.9515 1.14 0.54 �0.64 2.40 0.35 0.13 �0.13 1.5816 1.24 0.59 �0.73 2.75 0.34 0.14 �0.13 1.5617 1.14 0.58 �0.66 2.56 0.35 0.15 �0.17 1.3518 0.45 0.22 �0.12 0.16 0.19 0.05 0.03 0.6519 1.13 0.53 �0.63 2.38 0.33 0.13 �0.11 1.5420 0.53 0.24 �0.16 0.42 0.21 0.05 0.03 0.8821 0.53 0.26 �0.18 0.43 0.20 0.06 0.02 0.7522 1.31 0.65 �0.79 2.99 0.42 0.18 �0.23 1.7223 1.26 0.61 �0.76 2.72 0.46 0.19 �0.25 1.9224 0.90 0.47 �0.48 1.74 0.31 0.13 �0.13 1.1425 1.46 0.75 �0.90 3.81 0.46 0.22 �0.31 1.7726 0.88 0.47 �0.46 1.79 0.31 0.14 �0.15 1.0527 1.82 0.95 �1.21 4.84 0.54 0.28 �0.42 2.4728 1.43 0.72 �0.89 3.44 0.49 0.23 �0.32 1.8729 0.23 0.09 0.07 �0.58 0.10 0.00 0.14 0.3730 0.46 0.23 �0.12 0.28 0.18 0.05 0.02 0.6431 1.81 0.97 �1.22 4.66 0.53 0.28 �0.44 2.5332 1.82 0.96 �1.26 4.32 0.54 0.28 �0.43 2.4933 1.60 0.82 �1.00 4.35 0.47 0.23 �0.32 1.9834 0.90 0.46 �0.47 1.75 0.29 0.11 �0.10 1.0835 1.41 0.68 �0.87 3.18 0.49 0.20 �0.27 2.0836 0.75 0.38 �0.36 1.22 0.26 0.10 �0.07 0.9437 1.01 0.51 �0.55 2.15 0.31 0.13 �0.13 1.2138 0.58 0.31 �0.22 0.69 0.25 0.10 �0.06 0.7839 1.22 0.59 �0.71 2.75 0.35 0.14 �0.15 1.5240 1.75 0.89 �1.15 4.35 0.57 0.27 �0.41 2.4341 0.80 0.41 �0.40 1.38 0.28 0.11 �0.10 1.0042 1.81 0.94 �1.24 4.32 0.56 0.28 �0.43 2.5143 0.55 0.28 �0.19 0.57 0.19 0.06 0.01 0.7044 1.76 0.90 �1.19 4.36 0.51 0.24 �0.36 2.3045 0.69 0.36 �0.30 1.08 0.23 0.08 �0.04 0.8446 1.65 0.83 �1.07 4.11 0.52 0.24 �0.35 2.2147 0.15 0.05 0.13 �0.86 0.09 �0.01 0.15 0.2848 1.15 0.59 �0.67 2.71 0.35 0.16 �0.18 1.3449 0.63 0.31 �0.26 0.78 0.22 0.07 �0.01 0.8750 1.42 0.73 �0.87 3.74 0.43 0.21 �0.28 1.6951 0.91 0.41 �0.46 1.67 0.36 0.12 �0.10 1.6752 1.68 0.85 �1.08 4.40 0.50 0.24 �0.35 2.1553 0.98 0.48 �0.53 1.88 0.38 0.14 �0.16 1.5354 1.83 0.92 �1.19 5.02 0.56 0.28 �0.43 2.4355 1.51 0.79 �0.92 4.22 0.44 0.21 �0.30 1.8356 1.71 0.83 �1.12 4.24 0.57 0.27 �0.40 2.2857 1.26 0.65 �0.75 3.04 0.40 0.18 �0.23 1.5358 1.65 0.84 �1.04 4.55 0.45 0.22 �0.31 1.9759 1.03 0.50 �0.57 2.03 0.36 0.14 �0.14 1.4760 1.21 0.62 �0.72 2.80 0.41 0.18 �0.23 1.5661 0.57 0.29 �0.22 0.56 0.25 0.08 �0.04 0.9062 0.40 0.19 �0.08 �0.01 0.17 0.04 0.05 0.6063 1.22 0.61 �0.73 2.69 0.45 0.19 �0.25 1.7664 1.45 0.74 �0.89 3.79 0.44 0.21 �0.28 1.7565 1.05 0.54 �0.60 2.26 0.35 0.16 �0.18 1.2666 1.59 0.78 �1.03 3.92 0.47 0.22 �0.30 2.0067 1.60 0.81 �1.01 4.12 0.52 0.25 �0.36 2.08


Table 3 (continued)

Firm Conditional mean effect Conditional log-variance effect

LTA Tobin’s Q DEBT OWNERSHIP LTA Tobin’s Q DEBT OWNERSHIP

68 1.70 0.87 �1.07 4.90 0.46 0.23 �0.33 2.0469 1.73 0.88 �1.12 4.71 0.48 0.24 �0.34 2.1570 1.62 0.83 �1.03 4.21 0.51 0.24 �0.35 2.1571 1.22 0.59 �0.72 2.68 0.39 0.16 �0.18 1.6472 1.16 0.55 �0.65 2.48 0.36 0.13 �0.13 1.6473 1.81 0.96 �1.22 4.58 0.54 0.27 �0.42 2.4874 1.80 0.94 �1.18 4.90 0.53 0.27 �0.41 2.4075 1.41 0.69 �0.87 3.31 0.43 0.18 �0.23 1.9076 1.71 0.87 �1.09 4.48 0.55 0.26 �0.39 2.2977 1.42 0.71 �0.88 3.43 0.46 0.20 �0.28 1.8978 1.48 0.73 �0.93 3.50 0.50 0.22 �0.31 2.0779 1.00 0.51 �0.56 1.97 0.34 0.15 �0.16 1.2580 1.26 0.65 �0.75 3.09 0.39 0.18 �0.22 1.4981 0.99 0.51 �0.55 2.07 0.35 0.14 �0.16 1.2682 0.17 0.05 0.12 �0.80 0.09 �0.01 0.15 0.3183 1.55 0.78 �0.98 3.98 0.51 0.24 �0.34 2.0084 1.25 0.63 �0.74 2.93 0.40 0.17 �0.21 1.5885 0.91 0.44 �0.47 1.68 0.29 0.11 �0.07 1.2086 1.25 0.65 �0.74 3.08 0.40 0.19 �0.25 1.4987 0.96 0.48 �0.51 1.94 0.32 0.12 �0.12 1.2188 0.90 0.46 �0.48 1.72 0.32 0.13 �0.14 1.1489 1.09 0.55 �0.61 2.37 0.36 0.14 �0.17 1.4190 1.12 0.58 �0.65 2.52 0.37 0.17 �0.20 1.3491 1.02 0.54 �0.57 2.32 0.35 0.17 �0.20 1.2292 0.61 0.28 �0.22 0.72 0.22 0.06 0.02 0.9993 0.69 0.34 �0.30 0.98 0.24 0.08 �0.02 0.9494 1.76 0.89 �1.16 4.46 0.56 0.27 �0.40 2.4095 1.26 0.64 �0.76 2.96 0.42 0.19 �0.25 1.5696 1.01 0.49 �0.57 1.90 0.34 0.14 �0.13 1.3797 1.24 0.60 �0.74 2.68 0.47 0.19 �0.25 1.9898 1.10 0.56 �0.63 2.36 0.39 0.16 �0.21 1.4899 0.40 0.19 �0.06 0.05 0.14 0.03 0.08 0.55

100 0.42 0.19 �0.08 0.11 0.14 0.02 0.09 0.59101 0.56 0.28 �0.20 0.58 0.22 0.07 �0.01 0.76102 0.60 0.30 �0.24 0.62 0.26 0.09 �0.05 0.94103 0.60 0.30 �0.24 0.66 0.25 0.08 �0.03 0.92104 0.96 0.48 �0.52 1.88 0.35 0.14 �0.15 1.34105 1.11 0.57 �0.65 2.44 0.38 0.17 �0.21 1.36106 1.06 0.50 �0.58 2.13 0.34 0.13 �0.12 1.49107 0.59 0.28 �0.21 0.65 0.20 0.06 0.03 0.85108 0.71 0.33 �0.30 1.02 0.25 0.07 �0.01 1.10109 1.14 0.59 �0.66 2.60 0.39 0.17 �0.21 1.47110 1.81 0.99 �1.26 4.19 0.54 0.29 �0.45 2.52111 0.57 0.29 �0.21 0.59 0.22 0.07 �0.02 0.77


by investigating external differences, e.g., in the institutional framework. First, we consider the classification society, a majordriver of CSR in the shipping industry. Our sample firms belong to classification societies that are all members of IACS, thusthere are again no relevant institutional differences. Second, we analyze institutional variation in terms of the flag state. Weclassify firms by flag state as those that carry convenience flags, and those that do not. By analyzing the marginal effects,panel A of Table 4 shows that only 15% of the shipping companies with convenience flags exhibit a positive relationshipbetween our CSR disclosure index and ROA; the remaining 85% show a negative relationship. According to Vilsted (2004),CSR is intricately related to the flag state. By law the flag state has the right and the obligation to exercise all necessaryinspections and implementations of technical, administrative, and social matters. In some cases, however, particularly inthe case of ‘‘open registries’’ or what are otherwise considered convenience flags, there is a lack of well-trained personneland a lack of political will to enforce the laws (Kovats, 2006). Moreover, open registries imply lower tax rates as well as lowerregistration and tonnage fee factors, which tend to be the driver behind shipping companies’ choices to adopt certain flags(Tenold, 2003). The flag state thus seems to be naturally associated with the degree of CSR reporting that each firm assumes.In particular, in the case of convenience flags, CSR reporting is left partly to the management team, who may even use it as anentrenchment strategy to serve short-term management goals. For example, the firms in our sample under convenience flagsconsistently do not report the number of port detentions (index item 23), which could be used by shareholders as an argu-ment against paying out management bonuses.

Third, we use stock exchange listings to analyze institutional variations. In our sample, 37 shipping firms are listed onAnglo-American exchanges, and 81 on Euro-Asian exchanges. Panel B of Table 4 summarizes the main findings for this

Fig. 6. Examples of joint sample distributions of conditional effects (using ROA).

Fig. 7. Examples of joint sample distributions of conditional effects (using ROE).


grouping. Based on the marginal effects for each firm, we find that only 35% of the firms listed on Anglo-American exchangesshow a positive relationship between CSR disclosure and ROA, while 59% of the firms listed on Euro-Asian exchanges do. Itthus appears (and this notion is supported by anecdotal evidence) that the latter exchanges are more demanding and tend tomonitor CSR reporting in shipping more closely. If we tie this observation to the dominant flag carried by the fleet of eachlisted company, we find that 73% of the companies listed on the Anglo-American exchanges carry convenience flags, under

Fig. 8. Posterior distributions of marginal effects for all firms (MCMC): Robustness test.

Table 4Marginal effect of the return on assets (ROA) on the conditional mean of the CSR disclosure index.

Positive effect (%) Negative effect (%)

Panel A: Convenience flagNo convenience flag 52 48Convenience flag 15 85

Panel B: Convenience flag and stock exchange listingAnglo-American listing 35 65Euro-Asian listing 59 41Anglo-American listing and convenience flag 5 95Euro-Asian listing and convenience flag 22 78


which CSR reporting tends to be weakly implemented, while only 33% listed on the Euro-Asian exchanges do (not reported).Firms listed on Anglo-American stock exchanges which fly convenience flags have a strong tendency to reduce their amountof CSR disclosure when profitability increases (in fact, 95% of them do so, as shown in panel B of Table 4), arguably becausethey believe it only serves short-term objectives. In contrast, firms listed on the Euro-Asian exchanges usually do not carryconvenience flags. For the majority of these firms, when profitability increases, they also exhibit increased CSR disclosure tocomply with enhanced investor monitoring.

8. Conclusion

Corporate social responsibility (CSR) disclosures are vital tools for stakeholders to use in evaluating shipping companies’operations. We studied a comprehensive panel dataset of 111 listed shipping companies during the 2002–2010 sample period,which had about 8850 ships registered. Using items suggested in the earlier literature to measure CSR (Carroll, 1979, 1999;Waddock and Graves, 1997; and many others), as well as additional shipping-specific items, we constructed a CSR disclosureindex and assumed heterogeneity in CSR behavior among firms. We approximated CSR disclosure behavior for each firm in oursample, and we assessed the sign and the statistical significance of the relationships with various firm characteristics.

Most importantly, we documented a positive relationship between the amount of CSR disclosure and financial perfor-mance, as measured by Tobin’s Q, for all shipping firms in our sample. In a robustness test, we uncovered a negative relation-ship between CSR disclosure and operating performance (as measured by return on assets) for some firms in our sample.However, we also showed that this effect is (at least partially) attributable to the choice of the flag state and the stockexchange listing. Overall, our results corroborate the findings in prior studies that identify corporate performance, firm size,leverage, and ownership concentration as the main factors related to CSR disclosure, and we find it necessary to accentuatethese factors in the shipping industry.

Table A1Bayes probability intervals (95%) for mean and variance effects.

Variable Firm Mean effect Variance effect

Upper bound Lower bound Upper bound Lower bound

LTA 1 0.82 1.48 0.27 0.58Tobin’s Q 1 0.19 1.00 �0.05 0.47DEBT 1 �1.12 �0.24 �0.46 �0.02OWNERSHIP 1 0.42 4.61 0.78 2.81LTA 2 0.54 1.02 0.13 0.49Tobin’s Q 2 0.10 0.71 �0.10 0.41DEBT 2 �0.70 �0.12 �0.29 0.07OWNERSHIP 2 �0.40 2.84 0.04 2.75LTA 3 0.62 1.11 0.17 0.52Tobin’s Q 3 0.13 0.78 �0.09 0.43DEBT 3 �0.84 �0.14 �0.37 0.04OWNERSHIP 3 �0.20 3.25 0.17 2.72LTA 4 1.00 2.11 0.39 0.75Tobin’s Q 4 0.22 1.38 �0.01 0.56DEBT 4 �1.68 �0.32 �0.71 �0.08OWNERSHIP 4 0.25 6.96 1.18 3.92LTA 5 0.66 1.90 0.24 0.85Tobin’s Q 5 0.14 1.17 �0.06 0.58DEBT 5 �1.37 �0.21 �0.62 0.04OWNERSHIP 5 0.12 6.23 0.87 4.75LTA 6 1.05 2.27 0.38 0.74Tobin’s Q 6 0.15 1.51 �0.01 0.56DEBT 6 �1.84 �0.37 �0.69 �0.09OWNERSHIP 6 �0.18 7.43 1.26 3.90LTA 7 0.90 1.50 0.30 0.58Tobin’s Q 7 0.22 1.01 �0.04 0.45DEBT 7 �1.18 �0.24 �0.48 �0.04OWNERSHIP 7 0.67 4.67 0.81 2.69LTA 8 0.34 0.90 0.04 0.48Tobin’s Q 8 0.09 0.64 �0.10 0.37DEBT 8 �0.64 �0.05 �0.30 0.06OWNERSHIP 8 �1.52 2.46 �1.11 2.53LTA 9 0.92 1.38 0.25 0.55Tobin’s Q 9 0.21 0.98 �0.04 0.42DEBT 9 �1.10 �0.23 �0.43 �0.03OWNERSHIP 9 0.93 4.09 0.36 2.54LTA 10 0.90 1.40 0.17 0.53Tobin’s Q 10 0.22 0.93 �0.04 0.39DEBT 10 �1.04 �0.22 �0.39 �0.01OWNERSHIP 10 1.21 3.92 �0.12 2.52LTA 11 0.98 2.23 0.38 0.77Tobin’s Q 11 0.15 1.51 �0.01 0.58DEBT 11 �1.85 �0.32 �0.75 �0.07OWNERSHIP 11 �0.29 7.42 1.23 4.24LTA 12 1.01 1.80 0.24 0.59Tobin’s Q 12 0.25 1.14 �0.02 0.43DEBT 12 �1.30 �0.29 �0.49 �0.04OWNERSHIP 12 1.53 5.32 0.59 2.65LTA 13 1.12 2.18 0.20 0.66Tobin’s Q 13 0.27 1.38 0.00 0.48DEBT 13 �1.57 �0.29 �0.64 �0.03OWNERSHIP 13 2.41 6.95 0.53 2.96LTA 14 1.07 2.00 0.31 0.64Tobin’s Q 14 0.25 1.29 �0.01 0.48DEBT 14 �1.50 �0.32 �0.58 �0.06OWNERSHIP 14 1.45 6.28 1.03 2.83LTA 15 0.70 1.80 0.18 0.56Tobin’s Q 15 0.15 0.97 �0.08 0.44DEBT 15 �0.88 �0.30 �0.26 0.03OWNERSHIP 15 0.26 5.36 0.51 2.93LTA 16 0.74 2.02 0.16 0.52Tobin’s Q 16 0.15 1.10 �0.08 0.42DEBT 16 �1.00 �0.36 �0.26 0.02OWNERSHIP 16 0.23 6.09 0.39 2.86LTA 17 0.89 1.39 0.13 0.53Tobin’s Q 17 0.20 0.96 �0.05 0.39DEBT 17 �1.00 �0.23 �0.36 0.00


Table A1 (continued)



OWNERSHIP 17 1.21 3.88 �0.41 2.51LTA 18 0.14 0.72 �0.06 0.43Tobin’s Q 18 0.01 0.48 �0.16 0.35DEBT 18 �0.27 0.02 �0.10 0.11OWNERSHIP 18 �2.29 2.00 �1.76 2.64LTA 19 0.66 1.88 0.15 0.55Tobin’s Q 19 0.13 1.00 �0.10 0.44DEBT 19 �0.89 �0.30 �0.25 0.05OWNERSHIP 19 0.17 5.61 0.41 2.99LTA 20 0.30 0.83 0.00 0.40Tobin’s Q 20 0.01 0.53 �0.18 0.37DEBT 20 �0.30 0.01 �0.09 0.15OWNERSHIP 20 �1.12 1.74 �0.68 2.90LTA 21 0.29 0.75 �0.02 0.41Tobin’s Q 21 0.04 0.51 �0.15 0.36DEBT 21 �0.31 �0.05 �0.08 0.10OWNERSHIP 21 �1.62 1.91 �1.38 2.69LTA 22 0.94 1.66 0.29 0.56Tobin’s Q 22 0.22 1.07 �0.04 0.44DEBT 22 �1.19 �0.29 �0.42 �0.05OWNERSHIP 22 0.89 5.01 0.82 2.59LTA 23 0.82 1.62 0.29 0.64Tobin’s Q 23 0.18 1.08 �0.05 0.50DEBT 23 �1.23 �0.26 �0.52 �0.02OWNERSHIP 23 0.32 5.22 0.88 3.19LTA 24 0.63 1.10 0.07 0.50Tobin’s Q 24 0.16 0.77 �0.07 0.37DEBT 24 �0.79 �0.14 �0.31 0.03OWNERSHIP 24 �0.09 3.16 �0.95 2.46LTA 25 1.07 1.82 0.19 0.65Tobin’s Q 25 0.24 1.29 �0.01 0.48DEBT 25 �1.47 �0.25 �0.62 �0.02OWNERSHIP 25 1.99 5.59 0.10 2.81LTA 26 0.37 1.17 �0.01 0.54Tobin’s Q 26 0.13 0.86 �0.07 0.39DEBT 26 �0.87 �0.07 �0.43 0.06OWNERSHIP 26 �1.21 4.42 �1.63 2.48LTA 27 1.22 2.33 0.31 0.74Tobin’s Q 27 0.12 1.65 0.01 0.55DEBT 27 �1.92 �0.38 �0.76 �0.09OWNERSHIP 27 0.90 7.84 1.34 3.61LTA 28 1.04 1.75 0.30 0.65Tobin’s Q 28 0.23 1.24 �0.02 0.49DEBT 28 �1.46 �0.26 �0.62 �0.05OWNERSHIP 28 1.27 5.46 0.73 2.85LTA 29 0.02 0.55 �0.12 0.36Tobin’s Q 29 �0.13 0.40 �0.25 0.35DEBT 29 �0.09 0.21 0.01 0.23OWNERSHIP 29 �2.82 1.55 �2.03 3.01LTA 30 0.16 0.76 �0.07 0.43Tobin’s Q 30 0.02 0.50 �0.16 0.35DEBT 30 �0.28 0.01 �0.12 0.11OWNERSHIP 30 �2.25 2.49 �1.86 2.60LTA 31 1.10 2.35 0.21 0.75Tobin’s Q 31 �0.03 1.74 0.01 0.57DEBT 31 �2.01 �0.28 �0.78 �0.07OWNERSHIP 31 �0.62 8.07 1.30 3.75LTA 32 1.05 2.38 0.25 0.76Tobin’s Q 32 �0.01 1.69 0.00 0.56DEBT 32 �2.08 �0.30 �0.76 �0.07OWNERSHIP 32 �1.00 7.86 1.34 3.77LTA 33 1.10 2.18 0.26 0.65Tobin’s Q 33 0.27 1.35 0.00 0.47DEBT 33 �1.55 �0.33 �0.61 �0.05OWNERSHIP 33 2.00 6.87 0.95 2.90LTA 34 0.67 1.12 0.06 0.48Tobin’s Q 34 0.16 0.77 �0.08 0.36

(continued on next page)





DEBT 34 �0.71 �0.18 �0.24 0.03OWNERSHIP 34 0.21 3.12 �0.99 2.50LTA 35 0.86 1.92 0.31 0.69Tobin’s Q 35 0.20 1.18 �0.04 0.52DEBT 35 �1.35 �0.31 �0.53 �0.04OWNERSHIP 35 0.19 6.16 0.99 3.64LTA 36 0.46 0.95 0.02 0.46Tobin’s Q 36 0.12 0.66 �0.10 0.36DEBT 36 �0.58 �0.11 �0.21 0.05OWNERSHIP 36 �0.85 2.74 �1.29 2.48LTA 37 0.79 1.27 0.10 0.49Tobin’s Q 37 0.20 0.83 �0.07 0.37DEBT 37 �0.84 �0.21 �0.29 0.01OWNERSHIP 37 0.88 3.46 �0.64 2.50LTA 38 0.11 0.89 �0.07 0.50Tobin’s Q 38 0.05 0.62 �0.11 0.37DEBT 38 �0.53 0.03 �0.31 0.09OWNERSHIP 38 �2.44 3.34 �1.95 2.42LTA 39 0.83 1.85 0.19 0.51Tobin’s Q 39 0.20 1.01 �0.07 0.41DEBT 39 �0.97 �0.33 �0.27 �0.02OWNERSHIP 39 0.59 5.43 0.42 2.63LTA 40 1.11 2.27 0.39 0.74Tobin’s Q 40 0.21 1.51 0.00 0.55DEBT 40 �1.84 �0.36 �0.74 �0.09OWNERSHIP 40 0.37 7.40 1.30 3.80LTA 41 0.44 1.00 0.02 0.49Tobin’s Q 41 0.13 0.71 �0.08 0.36DEBT 41 �0.69 �0.10 �0.29 0.05OWNERSHIP 41 �0.94 3.02 �1.34 2.45LTA 42 1.09 2.34 0.33 0.77Tobin’s Q 42 0.06 1.65 0.00 0.56DEBT 42 �2.02 �0.33 �0.77 �0.08OWNERSHIP 42 �0.57 7.76 1.35 3.82LTA 43 0.26 0.82 �0.05 0.43Tobin’s Q 43 0.05 0.54 �0.15 0.35DEBT 43 �0.34 �0.04 �0.12 0.10OWNERSHIP 43 �1.76 2.61 �1.72 2.58LTA 44 1.10 2.32 0.30 0.70Tobin’s Q 44 0.16 1.54 0.00 0.51DEBT 44 �1.84 �0.40 �0.64 �0.08OWNERSHIP 44 0.18 7.46 1.30 3.44LTA 45 0.39 0.94 �0.02 0.46Tobin’s Q 45 0.11 0.63 �0.12 0.35DEBT 45 �0.51 �0.09 �0.20 0.07OWNERSHIP 45 �1.13 3.02 �1.55 2.51LTA 46 1.06 2.21 0.36 0.68Tobin’s Q 46 0.24 1.38 �0.01 0.51DEBT 46 �1.65 �0.37 �0.62 �0.08OWNERSHIP 46 0.67 7.08 1.23 3.34LTA 47 �0.04 0.42 �0.13 0.35Tobin’s Q 47 �0.17 0.36 �0.27 0.35DEBT 47 �0.02 0.25 0.03 0.24OWNERSHIP 47 �3.13 1.61 �2.24 3.05LTA 48 0.89 1.44 0.11 0.54Tobin’s Q 48 0.22 0.98 �0.04 0.39DEBT 48 �1.05 �0.22 �0.41 0.01OWNERSHIP 48 1.35 4.20 �0.64 2.53LTA 49 0.43 0.82 0.02 0.42Tobin’s Q 49 0.08 0.57 �0.14 0.36DEBT 49 �0.39 �0.12 �0.10 0.07OWNERSHIP 49 �0.91 2.06 �1.05 2.67LTA 50 1.05 1.80 0.17 0.62Tobin’s Q 50 0.24 1.23 �0.02 0.45DEBT 50 �1.38 �0.26 �0.57 �0.01OWNERSHIP 50 2.04 5.43 �0.01 2.73LTA 51 0.43 1.59 0.11 0.76





Tobin’s Q 51 0.07 0.86 �0.13 0.51DEBT 51 �0.79 �0.07 �0.37 0.13OWNERSHIP 51 �0.05 4.66 0.40 4.03LTA 52 1.11 2.27 0.33 0.68Tobin’s Q 52 0.26 1.41 0.00 0.50DEBT 52 �1.66 �0.37 �0.63 �0.08OWNERSHIP 52 1.34 7.28 1.24 3.09LTA 53 0.70 1.24 0.20 0.53Tobin’s Q 53 0.17 0.80 �0.07 0.43DEBT 53 �0.88 �0.19 �0.36 0.02OWNERSHIP 53 0.32 3.63 0.50 2.71LTA 54 1.22 2.41 0.34 0.75Tobin’s Q 54 0.18 1.63 0.01 0.56DEBT 54 �1.88 �0.37 �0.77 �0.09OWNERSHIP 54 1.62 8.04 1.29 3.55LTA 55 1.06 2.08 0.19 0.64Tobin’s Q 55 0.29 1.26 0.00 0.44DEBT 55 �1.47 �0.28 �0.60 �0.03OWNERSHIP 55 2.21 6.52 0.34 2.85LTA 56 1.12 2.22 0.37 0.76Tobin’s Q 56 0.16 1.54 �0.01 0.59DEBT 56 �1.84 �0.33 �0.75 �0.08OWNERSHIP 56 0.66 7.39 1.11 3.57LTA 57 0.97 1.54 0.17 0.57Tobin’s Q 57 0.23 1.07 �0.03 0.41DEBT 57 �1.20 �0.23 �0.49 �0.01OWNERSHIP 57 1.58 4.53 �0.19 2.58LTA 58 1.11 2.32 0.22 0.65Tobin’s Q 58 0.26 1.40 �0.01 0.47DEBT 58 �1.58 �0.36 �0.58 �0.05OWNERSHIP 58 1.97 7.44 0.90 2.90LTA 59 0.77 1.28 0.21 0.50Tobin’s Q 59 0.17 0.84 �0.07 0.42DEBT 59 �0.84 �0.23 �0.29 �0.01OWNERSHIP 59 0.45 3.68 0.45 2.60LTA 60 0.95 1.46 0.21 0.56Tobin’s Q 60 0.23 1.00 �0.03 0.41DEBT 60 �1.17 �0.22 �0.48 �0.02OWNERSHIP 60 1.28 4.28 0.09 2.56LTA 61 0.24 0.80 0.00 0.46Tobin’s Q 61 0.06 0.55 �0.13 0.36DEBT 61 �0.48 �0.02 �0.23 0.08OWNERSHIP 61 �1.89 2.13 �1.33 2.59LTA 62 0.12 0.68 �0.07 0.41Tobin’s Q 62 �0.02 0.46 �0.17 0.36DEBT 62 �0.22 0.05 �0.06 0.14OWNERSHIP 62 �2.40 1.76 �1.78 2.74LTA 63 0.90 1.50 0.30 0.58Tobin’s Q 63 0.22 1.01 �0.03 0.45DEBT 63 �1.21 �0.23 �0.51 �0.04OWNERSHIP 63 0.72 4.65 0.77 2.72LTA 64 1.05 1.86 0.21 0.62Tobin’s Q 64 0.26 1.22 �0.01 0.44DEBT 64 �1.39 �0.28 �0.56 �0.03OWNERSHIP 64 1.93 5.59 0.35 2.73LTA 65 0.68 1.30 0.07 0.54Tobin’s Q 65 0.17 0.95 �0.05 0.40DEBT 65 �1.00 �0.15 �0.42 0.03OWNERSHIP 65 0.17 3.88 �0.97 2.50LTA 66 1.05 2.16 0.31 0.64Tobin’s Q 66 0.20 1.37 �0.02 0.49DEBT 66 �1.54 �0.38 �0.53 �0.07OWNERSHIP 66 0.72 6.88 1.09 2.95LTA 67 1.10 2.03 0.33 0.69Tobin’s Q 67 0.24 1.38 0.00 0.52DEBT 67 �1.63 �0.31 �0.68 �0.06OWNERSHIP 67 1.57 6.49 1.06 3.05






LTA 68 1.15 2.43 0.20 0.66Tobin’s Q 68 0.27 1.46 0.00 0.47DEBT 68 �1.64 �0.36 �0.62 �0.04OWNERSHIP 68 2.28 7.88 0.90 3.01LTA 69 1.14 2.39 0.27 0.67Tobin’s Q 69 0.24 1.49 0.00 0.49DEBT 69 �1.70 �0.39 �0.63 �0.07OWNERSHIP 69 1.68 7.66 1.23 3.08LTA 70 1.08 2.21 0.35 0.68Tobin’s Q 70 0.28 1.34 0.00 0.49DEBT 70 �1.62 �0.34 �0.65 �0.07OWNERSHIP 70 1.29 7.06 1.22 3.13LTA 71 0.84 1.70 0.25 0.53Tobin’s Q 71 0.19 1.00 �0.06 0.43DEBT 71 �1.01 �0.31 �0.32 �0.04OWNERSHIP 71 0.54 5.02 0.69 2.66LTA 72 0.68 1.90 0.18 0.58Tobin’s Q 72 0.16 0.99 �0.09 0.45DEBT 72 �0.94 �0.30 �0.29 0.04OWNERSHIP 72 0.25 5.76 0.56 3.09LTA 73 1.16 2.33 0.30 0.75Tobin’s Q 73 0.10 1.65 0.01 0.55DEBT 73 �1.96 �0.35 �0.75 �0.08OWNERSHIP 73 0.05 7.74 1.34 3.68LTA 74 1.21 2.34 0.32 0.72Tobin’s Q 74 0.21 1.60 0.01 0.54DEBT 74 �1.84 �0.39 �0.73 �0.08OWNERSHIP 74 1.47 7.70 1.35 3.47LTA 75 0.90 2.07 0.29 0.60Tobin’s Q 75 0.24 1.14 �0.04 0.46DEBT 75 �1.27 �0.34 �0.43 �0.05OWNERSHIP 75 0.47 6.48 0.96 3.02LTA 76 1.13 2.23 0.37 0.72Tobin’s Q 76 0.26 1.45 0.00 0.54DEBT 76 �1.73 �0.35 �0.72 �0.08OWNERSHIP 76 1.38 7.30 1.25 3.35LTA 77 0.99 1.88 0.32 0.60Tobin’s Q 77 0.25 1.15 �0.02 0.45DEBT 77 �1.36 �0.31 �0.51 �0.06OWNERSHIP 77 0.99 5.86 1.02 2.74LTA 78 0.98 1.94 0.36 0.66Tobin’s Q 78 0.23 1.23 �0.02 0.50DEBT 78 �1.46 �0.32 �0.58 �0.07OWNERSHIP 78 0.56 6.25 1.09 3.20LTA 79 0.70 1.21 0.10 0.52Tobin’s Q 79 0.16 0.90 �0.06 0.39DEBT 79 �0.92 �0.16 �0.35 0.01OWNERSHIP 79 0.10 3.44 �0.70 2.48LTA 80 0.97 1.57 0.14 0.57Tobin’s Q 80 0.23 1.06 �0.03 0.41DEBT 80 �1.18 �0.23 �0.47 0.00OWNERSHIP 80 1.64 4.58 �0.32 2.58LTA 81 0.73 1.21 0.10 0.52Tobin’s Q 81 0.19 0.83 �0.05 0.38DEBT 81 �0.93 �0.16 �0.39 0.01OWNERSHIP 81 0.37 3.54 �0.73 2.48LTA 82 �0.03 0.47 �0.13 0.35Tobin’s Q 82 �0.17 0.38 �0.27 0.35DEBT 82 �0.06 0.25 0.02 0.24OWNERSHIP 82 �3.06 1.49 �2.14 3.07LTA 83 1.09 1.95 0.31 0.68Tobin’s Q 83 0.24 1.35 �0.01 0.51DEBT 83 �1.58 �0.30 �0.66 �0.06OWNERSHIP 83 1.60 6.18 0.92 2.96LTA 84 0.94 1.61 0.23 0.54Tobin’s Q 84 0.24 1.00 �0.04 0.41DEBT 84 �1.13 �0.27 �0.41 �0.03





OWNERSHIP 84 1.24 4.61 0.43 2.55LTA 85 0.66 1.26 0.11 0.45Tobin’s Q 85 0.14 0.77 �0.10 0.38DEBT 85 �0.63 �0.25 �0.15 0.03OWNERSHIP 85 0.29 3.21 �0.21 2.65LTA 86 0.89 1.56 0.11 0.60Tobin’s Q 86 0.21 1.12 �0.03 0.43DEBT 86 �1.24 �0.19 �0.55 0.01OWNERSHIP 86 1.26 4.90 �0.69 2.64LTA 87 0.76 1.17 0.11 0.49Tobin’s Q 87 0.20 0.77 �0.07 0.37DEBT 87 �0.80 �0.19 �0.29 0.01OWNERSHIP 87 0.61 3.22 �0.55 2.49LTA 88 0.57 1.10 0.06 0.51Tobin’s Q 88 0.16 0.78 �0.07 0.37DEBT 88 �0.82 �0.12 �0.34 0.03OWNERSHIP 88 �0.36 3.22 �1.03 2.46LTA 89 0.84 1.36 0.18 0.51Tobin’s Q 89 0.22 0.85 �0.05 0.39DEBT 89 �0.96 �0.22 �0.36 �0.01OWNERSHIP 89 1.00 3.77 �0.01 2.51LTA 90 0.78 1.39 0.09 0.56Tobin’s Q 90 0.19 1.01 �0.04 0.41DEBT 90 �1.08 �0.18 �0.46 0.02OWNERSHIP 90 0.64 4.12 �0.80 2.53LTA 91 0.49 1.34 0.01 0.58Tobin’s Q 91 0.16 0.97 �0.05 0.41DEBT 91 �1.05 �0.10 �0.51 0.05OWNERSHIP 91 �0.62 4.99 �1.46 2.55LTA 92 0.34 1.10 0.02 0.44Tobin’s Q 92 0.03 0.61 �0.17 0.39DEBT 92 �0.39 �0.02 �0.12 0.15OWNERSHIP 92 �0.69 2.33 �0.48 2.93LTA 93 0.50 0.90 0.04 0.42Tobin’s Q 93 0.10 0.60 �0.13 0.36DEBT 93 �0.44 �0.15 �0.10 0.06OWNERSHIP 93 �0.49 2.21 �0.83 2.66LTA 94 1.13 2.29 0.39 0.73Tobin’s Q 94 0.22 1.52 0.00 0.55DEBT 94 �1.83 �0.38 �0.73 �0.09OWNERSHIP 94 0.63 7.46 1.31 3.66LTA 95 0.96 1.53 0.18 0.59Tobin’s Q 95 0.22 1.09 �0.03 0.43DEBT 95 �1.24 �0.22 �0.52 �0.01OWNERSHIP 95 1.36 4.53 �0.14 2.60LTA 96 0.69 1.34 0.13 0.52Tobin’s Q 96 0.11 0.98 �0.09 0.43DEBT 96 �0.88 �0.22 �0.29 0.02OWNERSHIP 96 0.07 3.85 �0.14 2.64LTA 97 0.77 1.68 0.27 0.70Tobin’s Q 97 0.20 1.02 �0.05 0.50DEBT 97 �1.23 �0.23 �0.54 �0.01OWNERSHIP 97 0.33 5.37 0.86 3.60LTA 98 0.86 1.32 0.20 0.55Tobin’s Q 98 0.21 0.90 �0.04 0.40DEBT 98 �1.06 �0.18 �0.46 �0.01OWNERSHIP 98 0.80 3.88 �0.03 2.53LTA 99 0.16 0.70 �0.08 0.39Tobin’s Q 99 �0.02 0.46 �0.20 0.34DEBT 99 �0.18 0.06 �0.05 0.17OWNERSHIP 99 �2.14 2.06 �1.80 2.78LTA 100 0.19 0.75 �0.08 0.38Tobin’s Q 100 �0.03 0.49 �0.21 0.35DEBT 100 �0.21 0.07 �0.03 0.18OWNERSHIP 100 �1.81 1.78 �1.56 2.89LTA 101 0.24 0.81 �0.04 0.44Tobin’s Q 101 0.06 0.54 �0.14 0.35






DEBT 101 �0.39 �0.03 �0.16 0.08OWNERSHIP 101 �1.83 2.48 �1.63 2.54LTA 102 0.30 0.81 0.02 0.45Tobin’s Q 102 0.07 0.56 �0.13 0.36DEBT 102 �0.49 �0.03 �0.21 0.08OWNERSHIP 102 �1.65 2.09 �1.11 2.64LTA 103 0.36 0.79 0.03 0.44Tobin’s Q 103 0.08 0.55 �0.13 0.36DEBT 103 �0.45 �0.06 �0.16 0.08OWNERSHIP 103 �1.36 2.04 �1.03 2.66LTA 104 0.78 1.15 0.17 0.50Tobin’s Q 104 0.19 0.78 �0.06 0.38DEBT 104 �0.86 �0.17 �0.35 0.00OWNERSHIP 104 0.42 3.25 �0.15 2.51LTA 105 0.75 1.36 0.10 0.56Tobin’s Q 105 0.18 1.00 �0.04 0.41DEBT 105 �1.09 �0.16 �0.47 0.02OWNERSHIP 105 0.42 4.05 �0.76 2.52LTA 106 0.70 1.56 0.18 0.51Tobin’s Q 106 0.15 0.90 �0.09 0.43DEBT 106 �0.79 �0.28 �0.24 0.02OWNERSHIP 106 0.31 4.43 0.44 2.76LTA 107 0.37 0.88 0.00 0.40Tobin’s Q 107 0.05 0.56 �0.16 0.36DEBT 107 �0.33 �0.08 �0.07 0.12OWNERSHIP 107 �0.87 1.97 �0.92 2.82LTA 108 0.42 1.16 0.05 0.45Tobin’s Q 108 0.06 0.66 �0.15 0.39DEBT 108 �0.45 �0.10 �0.13 0.12OWNERSHIP 108 �0.34 2.67 �0.27 2.87LTA 109 0.89 1.40 0.17 0.55Tobin’s Q 109 0.23 0.93 �0.04 0.40DEBT 109 �1.10 �0.19 �0.47 �0.01OWNERSHIP 109 1.11 4.08 �0.20 2.54LTA 110 0.86 2.48 0.14 0.79Tobin’s Q 110 �0.16 1.80 0.00 0.59DEBT 110 �2.22 �0.18 �0.81 �0.05OWNERSHIP 110 �2.29 8.13 1.24 3.86LTA 111 0.25 0.81 �0.03 0.44Tobin’s Q 111 0.06 0.55 �0.13 0.35DEBT 111 �0.39 �0.03 �0.16 0.08OWNERSHIP 111 �1.83 2.41 �1.62 2.55


Our results provide justification for the apparent keenness of shipping companies in making more and more corporatesocially responsible behavior disclosed. Therefore, while we cannot unambiguously solve the issue of causality, the main pol-icy implication of our study for shipping management is that corporate performance may be promoted by enhancing corpo-rate socially responsible behavior. In fact, several of the items included in the construction of our CSR disclosure index areprerequisites for achieving cost minimization together with enhancing reputation through transparency, arguably settingthe ground for the long-run growth prospects of a shipping firm. The cross-sectional variation of CSR reporting practicespoints to firm-level characteristics, which shape a shipping firm’s effort to achieve sustainable legitimacy in the society,low cost of finance in the capital markets, and viability in the competitive arena of transportation. CSR disclosure should alsobe seen as being part of the broader institutional framework in the shipping industry, which aims to promote and safeguardstakeholders’ interests.

Our results have important implications for investors and corporate managers alike. For investors, our results have animpact on portfolio construction by suggesting that the inclusion of firms which improve their CSR disclosures may enhanceportfolio performance. A socially responsible shipping firm’s investor base may be enlarged by attracting more and sociallyresponsible investors, ultimately leading to better risk-sharing opportunities. At a firm level, our results should lead to moresocially responsible management, if managers pursue the primary goal of firm (shareholder) value maximization. Overall,implementing CSR disclosure practices, at least in the shipping industry, which is viewed as one of the ‘‘pillars’’ of globaltrade, is likely to have far-reaching implications.

A promising future research path could be to examine the degree of incorporating CSR behavior into individual firm strat-egies, while also analyzing the quality of CSR reporting in more detail. An analysis of how market conditions (bull and bearmarkets) impact the amount and quality of CSR disclosure could provide further interesting insights.


Appendix A

See Table A1.

References

Albertijn, S., Bessler, W., Drobetz, W., 2011. Financing shipping companies and shipping operations: a risk-management perspective. J. Appl. Corp. Financ.23, 70–82.

Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. J. Financ. Econ. 17, 223–249.Andrikopoulos, A., Merika, A., Merikas, A., Triantafyllou, A., 2013. Internet disclosure and corporate performance: a case study of the international shipping

industry. Transport. Res. Part A 47, 141–152.Banz, R.W., 1981. The relationship between return and market value of common stocks. J. Financ. Econ. 9, 3–18.Barnea, A., Heinker, R., Kraus, A., 2005. Green investor and corporate investment. Struct. Chance Econ. Dyn. 16, 332–346.Barnea, A., Rubin, A., 2010. Corporate social responsibility as a conflict between shareholders. J. Bus. Ethics 97, 71–86.Baron, D., Harjoto, M., Jo, H., 2011. The economics and politics of corporate social performance. Bus. Polit. 13, 1–46.Bauer, R., Derwall, J., Hann, D., 2009. Employee relations and credit risk. In: Working Paper. Maastricht University.Bauer, R., Koedijk, K.C.G., Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. J. Bank. Finance 29, 1751–1767.Blacconiere, W.G., Patten, D.M., 1994. Environmental disclosures, regulatory costs, and changes in firm value. J. Account. Econ. 18, 357–377.Blank, V., Broberg, P., Sven-Olof, C., 2009. What explains the extent and content of social and environmental disclosures on corporate websites: a study of

social and environmental reporting in Swedish listed corporations. J. Corp. Soc. Responsibility Environ. Manage. 16, 352–364.Bouslah, K., Kryzanowski, L., M’Zali, B., 2013. The impact of the dimensions of social performance on firm risk. J. Bank. Finance 37, 1258–1273.Brammer, S., Pavelin, S., 2006. Voluntary environmental disclosures by large UK companies. J. Bus. Finance Account. 33, 1168–1188.Brooks, A.C., 2012. The Road to Freedom: How to Win the Fight for Free Enterprise. Basic Books, New York, NY.Carroll, A.B., 1979. A three-dimensional conceptual model of corporate performance. Acad. Manage. Rev. 4, 497–505.Carroll, A.B., 1999. Corporate social responsibility: evolution of a definitional construct. Bus. Soc. 38, 268–295.Coady, L., Strandberg, C., Ota, Y., 2013. The role of corporate social responsibility (CSR) in the international shipping sector. In: Working Paper. University of

British Columbia.Cooke, T.E., 1989. Voluntary corporate disclosure by Swedish companies. J. Int. Financ. Manage. Account. 1, 171–195.Cremers, M., Ferrell, A., 2014. Thirty years of shareholder rights and firm value. J. Finance 69, 1167–1196.Debreceny, R., Grey, G., Rahman, A., 2002. The determinants of internet financial reporting. J. Account. Public Policy 21, 371–394.Det Norske Veritas, 2004. Corporate social responsibility and the shipping industry-project report. Report 2004-1535, rev. 1.Dhaliwal, D., Li, O., Tsang, A., Yang, Y., 2006. Voluntary non-financial disclosure and the cost of equity capital: the initiation of corporate social responsibility

reporting. Account. Rev. 86, 59–100.Dimson, E., Karakas, O., Li, X., 2013. Active ownership. In: Working Paper. London Business School.Drobetz, W., Richter, T., Wambach, M., 2012. Dynamics of time-varying volatility in the dry bulk and tanker freight markets. Appl. Financ. Econ. 22, 1367–

1384.Eichholtz, P., Kok, N., Quigley, J.M., 2010. Doing well by doing good? Green office buildings. Am. Econ. Rev., 2492–2509.Ettredge, M., Richardson, V.J., Scholz, S., 2002. Dissemination of information for investors at corporate web sites. J. Account. Public Policy 21, 357–369.Fama, E.F., French, K.R., 2007. Disagreement, tastes, and asset prices. J. Financ. Econ. 83, 667–689.Ferguson, M.J., Lam, K.C.K., Lee, G.M., 2002. Voluntary disclosure by state-owned enterprises listed on the stock exchange of Hong Kong. J. Int. Financ.

Manage. Account. 13, 125–152.Freeman, R.E., 1984. Strategic Management: A Stakeholder Approach. Cambridge University Press, Boston, MA.Freedman, M., Stagliano, A.J., 1991. Differences in social-cost disclosures: a market test of investor reactions. Account., Audit. Account. J. 4, 68–82.Friedman, M., 1970. The social responsibility of business is to increase its profits. New York Times Mag., September 13.Geczy, C.C., Stambaugh, R.F., Levin, D., 2005. Investing in socially responsible mutual funds. In: Working Paper. University of Pennsylvania.Geweke, J., 1992. Evaluating the Accuracy of Sampling based Approaches to the Calculation of Posterior Moments. Oxford University Press, Oxford.Geweke, J., Keane, M., 2007. Smoothly mixing regressions. J. Econom. 138, 252–291.Godfrey, P.C., Merrill, C.B., Hansen, J.M., 2009. The relationship between corporate social responsibility and shareholder value: an empirical test of the risk

management hypothesis. Strateg. Manage. J. 30, 425–445.Graham, J.R., Harvey, C.R., 2001. The theory and practice of corporate finance: evidence from the field. J. Financ. Econ. 60, 187–243.Graham, J.R., Harvey, C.R., Rajgopal, S., 2005. The economic implications of corporate financial reporting. J. Account. Econ. 40, 3–73.Guenster, N., Bauer, R., Derwall, J., 2011. The economic value of corporate eco-efficiency. Eur. Financ. Manage. 17, 679–704.Hagel, J., Brown, J.S., 2010. The best way to measure corporate performance. Harvard Bus. Rev., March 4.Hamilton, S., Jo, H., Statman, M., 1993. Doing well while doing good? The investment performance of socially responsible mutual funds. Financ. Anal. J. 49,

62–66.Heinkel, R., Kraus, A., Zechner, J., 2001. The effect of green investment on corporate behavior. J. Financ. Quant. Anal. 36, 431–449.Hobson, J.L., Kachelmeier, S.J., 2005. Strategic disclosure of risky prospects: a laboratory experiment. Account. Rev. 80, 825–846.Hong, H., Kacperczyk, M., 2009. The price of sin: the effects of social norms on markets. J. Financ. Econ. 93, 15–36.Hong, H., Kubik, J.D., Scheinkman, J.A., 2012. Financial constraints on corporate goodness. In: Working Paper. National Bureau of Economic Research.Hughes, J.S., Liu, J., Liu, J., 2007. Information asymmetry, diversification, and cost of capital. Account. Rev. 82, 705–729.Ingram, R.W., Frazier, K.B., 1980. Environmental performance and corporate disclosure. J. Account. Res. 18, 614–622.Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. J. Financ. Econ. 3, 305–360.Kim, C., Pantzalis, C., Park, J.C., 2013. Do family owners use firm hedging policies to hedge personal undiversified wealth risk? Forthcoming in: Financ.

Manage.Kovats, L., 2006. How flag states lost the plot over shipping’s governance. does a ship need a sovereign? Marit. Policy Manage. 33, 75–81.Krueger, P., 2013. Corporate goodness and shareholder wealth. In: Working Paper. University of Geneva.Lambert, R., Leuz, C., Verrecchia, R.E., 2007. Accounting information, disclosure, and the cost of capital. J. Account. Res. 45, 385–420.Lee, D.D., Faff, R.W., 2009. Corporate sustainability performance and idiosyncratic risk: a global perspective. Financ. Rev. 44, 213–237.Leftwich, R., Watts, R.L., Zimmerman, J.L., 1981. Voluntary corporate disclosure:the case of interim reporting. J. Account. Res. 19, 50–77.Leuz, C., Verrecchia, R., 2000. The economic consequences of increased disclosure. J. Account. Res. 38 (Suppl.), 91–124.Luo, X., Bhattacharya, C.B., 2009. The debate over doing good: corporate social performance, strategic marketing levers, and firm-idiosyncratic risk. J. Mark.

73, 198–213.Mackey, A., Mackey, T.B., Barney, J.B., 2007. Corporate social responsibility and firm performance: investor preferences and corporate strategies. Acad.

Manage. Rev. 32, 817–835.Margolis, J.D., Walsh, J.P., 2003. Misery loves companies: rethinking social initiatives by business. Adm. Sci. Q. 48, 268–305.Margolis, J.D., Elfenbein, H.A., Walsh, J.P., 2007. Does it pay to be good? A meta-analysis and redirection of research on the relationship between corporate

social and financial performance. In: Working Paper. Harvard Business School.Marston, C., Polei, A., 2002. Corporate reporting on the internet by German companies. Int. J. Account. Inf. Syst. 5, 285–311.

http://refhub.elsevier.com/S1366-5545(14)00146-X/h0005






































































Mattingly, J.E., Berman, S.L., 2006. Measurement of corporate social action: discovering taxonomy in the Kinder Lydenberg Domini ratings data. Bus. Soc. 45,20–46.

McGuire, J.B., Sundgren, A., Schneeweis, T., 1988. Corporate social responsibility and firm financial performance. Acad. Manage. J. 31, 854–872.Merton, R.C., 1987. A simple model of capital market equilibrium with incomplete information. J. Finance 42, 483–510.Orens, R., Aerts, W., Cormier, D., 2010. Web-based non-financial disclosure and cost of finance. J. Bus. Finance Account. 37, 1057–1093.Orlitzky, M., Schmidt, F.L., Rynes, S.L., 2003. Corporate social and financial performance: a meta-analysis. Org. Stud. 24, 403–441.Palazzo, G., Scherer, A., 2006. Corporative legitimacy as deliberation: a communicative framework. J. Bus. Ethics 66, 71–88.Park, B.J., Zhang, Y., Lord, D., 2010. Bayesian mixture modeling approach to account for heterogeneity in speed data. Transport. Res. Part B: Methodol. 44,

662–673.Pava, L., Krausz, J., 1996. The association between corporate social responsibility and financial performance. J. Bus. Ethics 15, 321–357.Preston, L., O’Bannon, D., 1997. The corporate social–financial performance relationship. Bus. Soc. 36, 5–31.Ruf, B.M., Muralidhar, K., Brown, R.M., Janney, J.J., Paul, K., 2001. An empirical investigation of the relationship between change in corporate social

performance and financial performance: a stakeholder theory perspective. J. Bus. Ethics 32, 143–156.Sharfman, M.P., Fernando, C.S., 2008. Environmental risk management and the cost of capital. Strateg. Manage. J. 29, 569–592.Stanwick, P.A., Stanwick, S.D., 1998. The relationship between corporate social performance and organizational size, financial performance and

environmental performance: an empirical examination. J. Bus. Ethics 17, 195–204.Stulz, R.M., 1996. Rethinking risk management. J. Appl. Corp. Finance 9, 8–25.Tagesson, T., Blank, V., Broberg, P., Sven-Olof, C., 2009. What explains the extent and content of social and environmental disclosures on corporate websites:

a study of social and environmental reporting in Swedish listed corporations. Corp. Soc. Responsib. Environ. Manage. 16, 352–364.Tenold, S., 2003. A most convenient flag-the basis for the expansion of the Singapore fleet, 1969–82. Marit. Policy Manage. 3, 255–268.Tierney, L., 1994. Markov chains for exploring posterior distributions. Ann. Stat. 22, 1701–1728.Tsionas, M., Merikas, A., Merika, A., 2012. Concentrated ownership and corporate performance revisited: the case of shipping companies. Transport. Res.

Part E 48, 843–852.Verrecchia, R.E., 2001. Essays on disclosure. J. Account. Econ. 32, 97–180.Verschoor, C.C., 1998. A study of the link between as corporation’s financial performance and its commitment to ethics. J. Bus. Ethics 17, 1509–1516.Villani, M., Kohn, R., Nott, D., 2012. Generalized smooth finite mixtures. J. Econom. 171, 121–133.Vilsted, P., 2004. Corporate Social Responsibility and the Shipping Industry. Technical Report No. 2004-1535. Norwegian Shipping Association.Waddock, S.A., Graves, S.B., 1997. The corporate social performance–financial performance link. Strateg. Manage. J. 18, 303–319.Wintoki, M.B., Linck, J., Netter, J.M., 2012. Endogeneity and the dynamics of internal corporate governance. J. Financ. Econ. 105, 581–606.






























Corporate social responsibility disclosure: The case of international shipping

Documents

Transcript of Corporate social responsibility disclosure: The case of international shipping