Introduction

In the night of Friday, 12 September 2008, the Federal Reserve received a late phone call. The CEO of the world’s largest insurance company, the American International Group, gave the central bankers startling news: his institution would need to raise $30 billion by Monday morning. Otherwise, it could face a downgrade and a bankruptcy that could precipitate the meltdown of the entire financial system (Dash and Sorkin 2008). This and many other episodes from the 2008 crisis illustrate the enormous danger that an unchecked financial sector poses to society through the imposition of systemic risks.

In the present paper, I will argue that some of the major theories in contemporary business ethics—including stakeholder ethics and Integrative Social Contracts Theory—have a serious flaw: they fail to provide an adequate ethical analysis of systemic financial risk. I will illustrate this point with a case study on credit default swaps trading and then develop three challenges that these theories struggle to address: first, my inquiry will find that they fall short of indicating a reasonable overall level of risk taking in the financial system (the problem of risk imposition). Second, they have difficulties in accounting for the fact that systemic risk in finance is mostly created collectively, while the individual contributions to it, considered in isolation, seem to be harmless (the problem of unstructured collective harm). Finally, these business ethics theories do not encompass a sufficiently sophisticated ethical account of knowledge acquisition. They do not tell finance professionals how much effort they need to invest into knowing more about the impact of their business practices (the problem of limited knowledge). Building on these criticisms, in the constructive part of this paper, I will suggest that, to address the problem of moral responsibility for systemic financial risk, we need to shift our attention to the pluralist connection models of moral responsibility used in political philosophy. This allows us to identify five “areas of risk responsibility” in which banks, funds and insurers incur additional obligations.

Systemic risk is the risk of a substantial disruption to the flow of financial services with the potential to have serious negative consequences for the real economy (cf. IMF et al. 2009). Importantly, systemic risk may engender considerable harm. The Bank of England estimates that worldwide GDP losses caused by the 2008 financial crisis will lie between $60,000 billion and $200,000 billion (Haldane 2010). However, many of its effects cannot be measured in monetary terms and do not affect the industrialised world alone. For instance, the World Bank calculates that the 2008 crisis and the subsequent recession threw 2 million Bangladeshis and 1.4 million Filipinos from a lower-middle-class life into poverty (Habib et al. 2010). Sociologists estimate that the crisis prompted 10,000 additional suicides in Europe and North America just in its first 3 years (Reeves et al. 2014). Note that this type of risk is not a force of nature. The harm it produces is different from the damage of an earthquake or hurricane. It is man-made and in principle avoidable.

Yet, despite its devastating impact, business ethicists have written remarkably little about the moral implications of systemic risk creation.Footnote 1 In fact, when writing about the recent global financial crisis, many of them do not mention systemic risk at all. Instead they focus on manifestly immoral business practices that also happened to be among the causes of the crisis. An extensive part of the literature in business ethics, but also in popular economics, highlights that finance professionals were driven by greed (e.g. Smith 2010; Madrick 2011). Others criticise banks, funds and insurers for lying, cheating, and breaches of trust (e.g. Fassin and Gosselin 2011; Boatright 2013). Some management scholars even discuss whether bankers suffer from psychological disorders and qualify as “corporate psychopaths” (Boddy 2011).

To be clear, it is important to expose such moral shortcomings. Yet, the ethical inquiry must not end here for two reasons: (1) the narrow focus on greediness, lying and cheating risks to convey the impression that the financial crisis was exclusively caused by crooks, spivs and psychopaths. It suggests that further crises can be avoided if these people simply “behaved well”. This empirical claim, however, is simply false. Financial economics has uncovered a multitude of factors contributing to systemic risk (cf. Danielsson 2013; Bisias et al. 2012). Most of them do not involve lying, cheating or other manifestly blameworthy actions. Therefore, chastising bankers’ alleged depravity may be justified in some instances but tends to encourage policy responses that are unhelpful for preventing further crises. (2) Showing how manifestly immoral business practices happened to contribute to the financial crisis obscures a second, far more important aspect: Some business practices were only immoral in virtue of contributing to the financial crisis. Much of the harm produced by the collapsing financial system was not caused by people with malicious intents. Rather it was the result of many individuals interacting according to what they deemed normal and socially accepted ways of doing business. Yet, given the disastrous effects of the crisis, it is sensible to go back and reassess the moral status of these business practices. This is why we need a better understanding of moral responsibility for systemic risk creation.

In order to lay the foundations for such an inquiry, the present paper will proceed as follows: part 1 investigates the emergence of systemic risk by examining the case of credit default swaps trading and highlighting its ethically relevant features. Part 2 draws on these characteristics to develop three challenges for contemporary business ethics: the moral problems of risk imposition, of unstructured collective harm and of limited knowledge. In part 3, I will criticise some of the most popular theories in business ethics for failing to address these challenges adequately. Before concluding, I will briefly discuss two possible solutions and endorse the second of them, pluralist social connection models of moral responsibility.

Part 1: A Case Study on Credit Default Swaps

How does financial systemic risk come about? Who is responsible for its creation? Financial economics has identified many different factors that may cause systemic risk increases (Danielsson 2013). Drawing on recent events and the relevant economics literature, the following section will analyse how one of these factors, the widespread use of credit default swaps (CDS), contributes to its emergence. This will enable us to identify some morally relevant patterns that are typical for systemic risk creation.Footnote 2

How do CDS work? A CDS is a financial product that functions very much like an insurance against unexpected loan defaults. Here is an example: SitSafe produces seat belts for DriveFast, a large automobile manufacturer, which is its only client. Now, it may be unlikely that DriveFast goes bankrupt and defaults on what it owes SitSafe. However, if this happened, it would endanger SitSafe’s survival as well. For this reason, the seat belt manufacturer will reach out to one of its banks. The bank and SitSafe agree on a contract: similar to an insurance premium, SitSafe regularly pays the bank a fee (also called “spread”) until the CDS’s maturity date. In return, if DriveFast defaults on its liabilities towards SitSafe during the CDS’s term, the bank will pick up the bill and pay SitSafe a predetermined sum of money minus the post-default value of the underlying outstanding credits (i.e. the part of the credit that SitSafe is still able to obtain from DriveFast). Such a contract is called a credit default swap because the bank and SitSafe “swap” the exposure to the credit risk for a predetermined premium. Yet, contrary to what their name suggests, most CDS are not used to insure credits, but for speculative purposes (Cont 2010, p. 35). This happens if investors “insure” themselves against risks that they are not exposed to. For instance, someone could buy CDS on DriveFast’s corporate bonds simply because she wants to place a bet on this company’s default. This is called a “naked CDS”.

For two main reasons, CDS trading might be beneficial for the involved transaction partners as well as society in general. First, it allows for a separation of investing and risk taking. Investors can decide much more flexibly which risks they actually want to be exposed to and transfer the unwanted rest to an insurer. Second, on a social level, CDS may help to allocate risks more efficiently. Assuming perfect information, risk exposures will be bought by the market participants who are most capable and willing of shouldering them. If this happens, the financial system becomes more robust (Cont 2010, p. 36).

However, the widespread use of CDS also has some important downsides. It creates a web of insurance contracts that makes the financial system more interconnected. As long as the losses are limited to a small number of institutions and are relatively small, this interconnectedness is an asset. It makes the entire system more resilient because the impact of the losses is spread across several market participants. However, if the shocks get larger or more frequent, the same interconnectedness becomes a calamitous pitfall. Then, CDS and other derivatives may act as fatal transmission mechanisms that trigger one bankruptcy after another like a chain of dominoes (Acemoglu et al. 2015).

In addition to the transmission of actual losses, the excessive use of CDS also triggers a psychological feedback loop fuelled by high levels of uncertainty (Landau 2009). A key reason for this is that CDS are sold in so-called “over-the-counter” markets: they are not traded on public exchanges, but directly between the involved parties. This creates a transparency problem. If firms plan to do business with one another, they do not really know what risks their transaction partner has agreed to shoulder. Moreover, they know even less about what risks their transaction partners’ transaction partners bear and how vulnerable they are to these potential losses. This intricate web of intertwined counterparts makes the default risks very hard to assess. As a consequence, this uncertainty in itself may deter financial institutions from providing short-term loans to each other, thereby exacerbate credit crunches and produce systemic risk. This is why CDS were one of the main reasons why banks ran into “the difficulty of identifying where risks and losses lay in the financial system, which in turn created some of the uncertainty behind the liquidity crisis of 2007” (Davies 2010, p. 74).

Given this characterisation of the causal link between CDS and systemic risk, what are the implications for an ethical analysis of this business practice? For the ethical evaluation in the next sections, I will assume that the CDS trading is often characterised by the presence of the following six morally relevant features:

  1. (1)

    CDS deals have the potential to be harmless or even beneficial for society.

  2. (2)

    At the same time, if they create a critical level of interconnectedness between the market participants, they become detrimental for society at large by augmenting systemic risk.Footnote 3

  3. (3)

    They do not necessarily entail actual negative consequences, but sometimes only raise the likelihood of negative outcomes.

  4. (4)

    The emergence of systemic risk is caused by the interaction of many market participants.

  5. (5)

    Often, the causal contribution of a single individual to an increase in systemic risk is negligible. Then, only many individual actions taken together have an appreciable impact.

  6. (6)

    The decisions of individual market participants involve epistemic uncertainty regarding the impact of their actions for the creation of systemic risk.

Part 2: Three Challenges for Contemporary Business Ethics

The Moral Problem of Risk Imposition

One morally relevant property of systemic risk is already contained in its name: it is a risk, i.e. “the probability of an unwanted event that may or may not occur” (Hansson 2014). For ethical theories, accommodating risk and uncertainty can be challenging. For instance, we find certain death from cyanide poisoning inacceptable. But many of us are disposed to take a one-in-a-billion risk of dying from cyanide poisoning whenever they swallow a headache pill (Otsuka 1991). This and many other examples raise the question: where is the threshold? How low does the probability of harm need to be to make it permissible? Some aspects of this question will be explored in the following section.

The first relevant aspect is that the risk of harm needs to be taken seriously in itself, not just the harm that may actually occur. When the resilience of the financial system diminishes through CDS trading, negative outcomes will not necessarily ensue. A crisis may materialise—but importantly: it also may not. What should we make of risk impositions that remain without consequences? Judith Jarvis Thomson (1986, p. 181) invites us to envisage the following thought experiment: Person A plays Russian roulette on person B. In the end, nothing happens. The chamber hit by the firing pin was empty. Does B have a complaint against A? Intuitively, this clearly seems to be the case. One reason for objecting might be B’s negative experience of fear, but Thomson shows that our intuition persists even without the disutility from being afraid. The fear can be stipulated away if we imagine B to be soundly asleep and that triggering the weapon does not wake or disturb her in any way. Still, the grounds for B’s complaint persist. The risk imposed on us is thus morally problematic in itself. Ethical actualism, the approach that only considers the actual consequences of an action, is hence inadequate in dealing with risky decision situations.

For rights-based ethics, this raises the following question: “Imposing how slight a probability of a harm that violates someone’s rights also violates his rights?” (Nozick 1974, p. 74). A possible response is to ban actions that put other people at risk completely; but such a ban is clearly too strict since many beneficial social practices involve the imposition of unwanted risks. Should we, for instance, completely prohibit air travel just because there is a small chance of a plane crashing into someone’s house? Should we stop using electricity because some people may risk electrocution? A categorical ban, i.e. a right of non-imposition of risks seems normatively undesirable and paralyse human civilisation. Thus, the right of not being exposed to risks must be a defeasible one. Hayenhjelm and Wolff (2011, p. 1) label this problem of deontological approaches the “problem of paralysis”.

Consequently, to fix the problem of paralysis, a more balanced approach to risk restriction is needed. The challenge is to fix a reasonable threshold for permissibility. One solution is to resort to cost–benefit analysis and say that risky activities should be permitted whenever they maximise aggregate welfare. However, utilitarianism has risk-related problems of its own. Most importantly, it is vulnerable to a probabilistic version of Rawls’s “separateness of persons” critique (Rawls 1999, p. 164). In the classic variant of this argument, Rawls denounces that utilitarianism is only focused on maximising aggregate utility and therefore neglects distributional issues. For a classic utilitarian, the great suffering of one person could be outweighed by small utility gains for many people. When the consequences of an action are indeterminate, this raises an analogous question: is it better to expose a single person to a substantial harm with a probability bordering on certainty? Or is it better to impose a vanishingly small probability of the same harm on many people? Since the latter seems undoubtedly preferable, utilitarianism must appear overly permissive. Risk ethicists call this twin problem of the problem of paralysis the “problem of permitted unfairness” (Hayenhjelm and Wolff 2011, p. 37). A credible theory of risk imposition must be capable of indicating a reasonable middle ground between paralysis and permitted unfairness.

The Problem of Unstructured Collective Harm

Let us turn to the second challenge for business ethics theories. One of the main features of systemic risk in finance is that it is an instance of unstructured collective harm. A collective harm is unstructured when it is not the result of a concerted effort, but instead a by-product of the uncoordinated interactions of the involved agents (Kutz 2000, p. 166). In the following, I will use the term to designate instances of harm that satisfy the following three conditions: first, none of the contributing agents makes a perceptible difference to the aggregate harm. Indeed, if they occurred in isolation, their actions would be deemed harmless. Second, the combined effect of all individual causal contributions is harmful. And third, the agents do not intend to produce the combined effect (although they may foresee it). Our case study has shown that this characterisation also applies to CDS trading. The moral status of CDS deals crucially depends on the market environment that surrounds them. The independent actions of numerous other market participants can completely alter the upshots of one firm’s CDS business. Hence, theories attempting to provide ethical guidance in financial markets must be able to deal with the moral challenges of uncoordinated interactions.

To begin with, a central issue related to unstructured collective harm is the imperceptibility of individual contributions. In financial markets, individual acts often seem negligible to the overall levels of systemic risk. This is because the negative upshots of the combined actions—a full-blown system collapse in the worst case—are causally overdetermined: they would occur with or without a particular individual contribution. Now, how can an action be wrong if its execution or omission does not make any difference for anyone? Many ethical theories run into serious problems when attempting to answer this question. Kutz (2000, p. 116) explains that this is because they are implicitly committed to the so-called Individual Difference Principle: “I am accountable for a harm only if what I have done made a difference to the harm’s occurrence”.

In his seminal discussion of the issue, Parfit (1984, pp. 75ff) claims that imperceptible harms are morally relevant, too. In his view, the Individual Difference Principle cannot be upheld because its implications are untenable. Parfit illustrates this with the following thought experiment: a thousand torturers each flick a switch that turns on a device. Each of the devices is connected to the same victim and causes her to feel a very small, but imperceptible increase in pain. In aggregate, however, these marginal contributions inflict an excruciating pain on the victim (Parfit 1984, p. 80). Has an individual torturer done anything wrong? Intuitively, this clearly seems to be the case. After all, the victim is suffering terrible pain through her torturers’ actions. It appears that individual torturers must be held accountable for their causal contribution to the aggregate harm, but this move is not warranted by any moral theory committed to the Individual Difference Principle. Since any individual torturer does not make any palpable difference, he cannot be held accountable on these grounds.Footnote 4

Instead of adopting the consequential Individual Difference Principle, an ethical analysis of systemic risk could adopt a deontological strategy and shift its focus to the intentions of the involved agents as basis of accountability. In the past few decades, group agency, complicity and collective moral responsibility have been rapidly growing fields of research in moral theory (e.g. Kutz 2000; Gilbert 2006; List and Pettit 2011). Nevertheless, even this strategy for connecting the collective harm and the individual contribution is bound to fail in the case of systemic risk. Importantly, the market participants do not intend to bring about a system collapse. Instead they pursue a variety of many different goals, like hedging their risk exposure to Spanish government bonds or bet on a falling oil price. In fact, they would probably all fare better if market functioning was robust. Since individual market participants do not intend the collective harm, they cannot be held accountable on these grounds.

To sum up, to a significant extent, systemic risk is a case of unstructured collective harm. Considered in isolation, the individual contributions to it seem innocuous because they often do not make any perceptible difference. However, the combined effect of many CDS deals may be harmful. But since the agents do not intend this harm, i.e. do not intend to cause a financial crisis or increase its probability, they cannot be held responsible on the grounds of collective intentionality.

The Problem of Limited Knowledge

Finally, when investigating the moral responsibility of individual actors or firms for systemic risk, the accessibility of knowledge is of central importance. If someone knows little or nothing about the impact of her actions, we generally take this to limit her moral responsibility for these consequences.Footnote 5 This raises the question: what did bankers, insurers and fund managers actually know about the impact of CDS deals?

In this section, I will first show that three epistemic obstacles can make it quite hard to work out the systemic-risk-related upshots of CDS dealings and explain why this raises the need for a moral theory of knowledge acquisition in finance. The first epistemic obstacle is that CDS markets are opaque, i.e. individuals and firms only have partial access to the information they need to fully appraise the systemic-risk-related consequences of their CDS deals. As mentioned above, the main reason for this is that credit derivatives are largely traded “over the counter”, i.e. not through a third-party public exchange but directly between the involved parties. A lot of the morally relevant information about transactions is covered by banking secrecy. This opacity creates information asymmetries between the transaction partners (or rather: lacks of information on both sides). The patchy picture that a financial institution has of the risk exposures of its transaction partners to other parties, prevent it from adequately accounting for it in its own risk management. Then again, there are some financial information services like MarkIt and DTCC that do gather aggregate data about the scale of derivatives markets including the volume of outstanding credit default swaps. Nevertheless, the opacity of CDS market raises the question if all participants have enough information to assess the consequences of their business operations. The second epistemic obstacle is the complexity of CDS markets. The large number of different players and contract types, of different underlyings, the difficulties of predicting the reactions of other players make it hard for large firms, not to mention smaller market participants, to estimate what impact their dealings have on financial stability. That means that, even if the individuals and firms possessed a sufficient amount of relevant information, two further questions would need to be answered: (a) Do they have the analytical capacities to distil the ethically relevant features from the information they possess? (b) How onerous are these deliberations? Is it even reasonable to ask the agents to perform these analytical tasks? A third epistemic obstacle is the low level of moral intensity of finance, i.e. the low psychological salience of the moral dimension of an action (Jones 1991). If a financial firm overlooks or underestimates the moral dimension behind a business operation, it is also more likely to put less effort into overcoming the opacity and complexity of the situation and ascertaining that it does not harm third parties.

Yet, moral responsibility is not limited to the consequences an agent actually knows about. The agent is also responsible for the process of knowledge acquisition itself. In other words, it does not only matter what finance professionals do know, but also what they should know about the social impact of CDS. There must be an ethical evaluation of whether the search for morally relevant facts has been undertaken with due diligence. This idea is also the starting point of contemporary virtue epistemology. Its proponents argue that there is a set of virtues governing responsible knowledge acquisition (e.g. Fairweather and Zagzebski 2001). De Bruin (2015) has applied this approach to the realm of finance. In a recent monograph with the programmatic subtitle, “Why incompetence is worse than greed”, he argues that the key moral failure behind the 2007 financial crisis was a lack of epistemic virtues, like love of knowledge or epistemic humility. De Bruin advocates a “virtue-to-function matching” (p. 112) and maintains that different professional functions and job descriptions in finance are linked to distinct sets of epistemic virtues. For instance, a firm’s chief risk officer needs to display a different level of “epistemic temperance” in uncovering dangerous risk taking from the CEO, whose function it is to explore new—and possibly bold—plans for the future. Yet, epistemic duties do not only apply to individuals, but also to firms as a whole. To prove this, De Bruin cites the example of a board of directors, members of which are all willing to diligently monitor their firm, but their efforts are thwarted by the company’s executives (p. 118). This shows that virtuous individuals do not necessarily form a virtuous firm; epistemic virtues must also be incorporated into the company structure.

Yet, the acquisition of morally relevant knowledge is also to a large extent a cooperative project (Buchanan 2009). Much of what we know depends on the epistemic efforts of others—and the methods that banks, funds and insurers use to measure and predict systemic risk are no exception to this rule. The recent financial crisis was also a crisis for mathematical modelling. Some remarks suggest that some banks’ quantitative forecasts dramatically fell short of grasping the new levels that the systemic risk had reached. On 13 August 2007, with the worst shocks of the crisis still ahead, Goldman Sachs CFO David Viniar, told the Financial Times about the past months: “We were seeing things that were 25-standard deviation moves, several days in a row”. For comparison: 20-sigma events should statistically occur only once in a number of years that is ten times higher than the number of particles in the universe (Dowd et al. 2008). In other words, some of Goldman’s models had become completely useless for assessing the then-prevailing market risks. While we should partly blame individual risk managers for the blind trust they put into their quantitative tools, their actions also depended on the current state of academic research as well as customs and industry standards among practitioners. Hence, the social dimension of knowledge acquisition should also be taken into account when assessing the responsibility of businesses for systemic financial risk.

As the links between individual actions and collective harms become more complex, it becomes more important for contemporary ethics to comprise an ethical theory of knowledge acquisition. The complexity, opacity, and low moral intensity of CDS trading bolster this requirement. But individual knowledge acquisition also depends on the social context. Therefore, the liability for some cases of ignorance cannot fully be attributed to the individuals.

The implications from our three challenges—the risk challenge, the collective action challenge and the knowledge acquisition challenge—are summarised in Table 1.

Table 1 A six-point checklist for business ethics theories of systemic risk imposition
Full size table

Part 3: Applying the Challenges to Major Business Ethics Theories

Blind Spots in Stakeholder Theories

This section will be devoted to a discussion of our three challenges in connection with what is probably the most popular cluster of business ethics theories: stakeholder theories. Their main idea is that companies should not be managed in the exclusive interest of their shareholders but also in the interest of their stakeholders, i.e. “groups or individuals who benefit from or are harmed by, and whose rights are violated or respected by, corporate actions” (Freeman 2004, pp. 55f). Usually, this includes shareholders, employees, consumers, suppliers and civil society organisations (Crane and Matten 2004, p. 62). Sometimes the list is longer and also comprises other players like the media, governments, and local communities. Nevertheless, in many cases such a concise list seems reasonably comprehensive. But is it enough in cases involving systemic risk?

To begin with, due to the indeterminate nature of systemic risk, Freeman’s stakeholder definitions must be amended to account for the limitations of actualism. As argued above, a surge in risk is morally problematic in itself, that is to say: even if the negative outcome does not materialise. In fact, some stakeholder theories have recognised the limitations of actualism. For example, Clarkson (1994, p. 5) defines a stakeholder as someone who is “placed at risk as a result of a firm’s activities”.

Second, collective harming also calls for a collective mode of stakeholder identification. In our CDS example, the impact of a single firm’s derivatives dealings on taxpayers is often undetectable and does not seem to make any significant difference to the overall costs they incur. This would suggest that taxpayers simply do not have a stakeholder status for this company and their interests do not matter when the management makes decisions on what volume and types of derivative contracts to enter. However, this is unconvincing because our case study shows that, together with other firms’ dealings, precisely these decisions may inflict substantial losses on current and future taxpayers. For stakeholder ethics, this means introducing a second class of stakeholders and stakeholder duties: those that spring from a firm’s contribution to unstructured collective harm and accrue to the contributors collectively.

Thirdly, in many cases involving systemic risk, it is not even possible to know in advance who the stakeholders are. Of course, some of these costs will be borne by the “classic” stakeholder groups of financial firms—their employees, shareholders, clients, and local communities. But these groups are hardly the only ones. For instance, a banking crisis and the subsequent recession may affect people who lost their homes, the newly unemployed, investors suffering stock market losses, pensioners whose retirement savings no longer suffice, taxpayers held to pay for the bailouts, artists and scientists whose public funding is cut, medical patients and schoolchildren who could receive a better education if the bailout money was still available, and so forth.

In other words, there is (1) considerable epistemic uncertainty about the potential consequences of increases in systemic risk, which leads to (2) an unmanageably large number of risk-exposed stakeholders, of whom (3) many are impossible to identify in advance. This is also a problem because a key idea behind stakeholder ethics is to carry out a qualitative analysis of the relationship between firm and stakeholder and thereby identify special obligations like fiduciary duties or duties of due care. By contrast, in cases involving systemic risk, much of the stakeholder list would simply feature strangers. The only relationship between them and the firm is that they happen to be harmed by it. To summarise this in slightly polemic terms, in a setting where the perpetrator is an unstructured collective and the stakeholders are anonymous, stakeholder ethics frameworks become practically unserviceable.

Blind Spots in Integrative Social Contracts Theory

Let us now turn to a second candidate, Integrative Social Contracts Theory (ISCT) (Donaldson and Dunfee 1994). Donaldson and Dunfee argue that the ethical obligations of businesses do not only stem from universal moral values, but also from context-dependent norms within smaller communities of agents. Thus, they propose a two-part structure consisting of a single “macrosocial contract” and many lower-level “microsocial contracts”. Since systemic risk is a truly global phenomenon and may substantially violate the rights of member of other microcommunities, in this section, I will only focus on the macrosocial contract. While the finer details may be left to microregulation, it is the macrosocial contract that must put a ceiling on legitimate systemic-risk imposition. For the negotiation of the macro-contract, Donaldson and Dunfee adopt a social-contract model based on hypothetical consent. They expressly opt against a Rawlsian procedure including a veil of ignorance and intend to secure fairness by “including among the contractors all persons whose interests are affected and by requiring consensus in the adoption of the terms of the contract—without the additional device of a veil of ignorance” (p. 260). The final macro-contract will contain a list of “hypernorms”, universal values that are “not microcommunity relative” (p. 264).

For a start, the ISCT framework is not vulnerable to the limitations of actualism. A social contract does not necessarily regulate actual outcomes but may equally well cover the permissibility of risk impositions. Nonetheless, it is prone to the problem of paralysis. This is because Donaldson and Dunfee have committed to (a) including among the contractors “all persons whose interests are affected”, and to (b) adopt it “by requiring [informed] consent” (ibid.). This gives rise to the following problem: what if there is at least a single person who is harmed by the risk imposition, but does not benefit enough to agree to it? As a consequence, under the terms of ISCT, this risk imposition should not be permitted. Yet, almost any risk-creating business practice in finance would most likely fail the consent criterion because it is extremely likely that some people derive no or too little benefit from it compared to the costs they incur. Munoz-Dardé illustrates this problem with the example of a totally self-sufficient Amish farmer who refuses to consent to air travel because of the infinitesimally small probability of a plane crashing into his home (cf. Hayenhjelm and Wolff 2011, p. 40). How to overcome this impasse? Hansson (2013) suggests that the solution may be risk exchanges. Everyone benefits from some kind of risk taking: some people use nuclear energy, some smoke in public places, and others benefit from a financial sector that heavily trades in derivatives. Thus, Hansson proposes that a person’s exposure to a risk “is acceptable if this exposure is part of a social practice of risk-taking that works to her own advantage” (Hansson 2013, p. 104), i.e. if the risk imposition is part of an arrangement resulting from fair exchanges of risk. Although this approach considerably reduces the number of “Amish farmer cases”, in principle, this objection remains valid.

Introducing a Rawlsian veil of ignorance, so that individuals do not know whether they turn out to be an “Amish farmer” or somebody else, could help ISCT to avoid the problem of paralysis. However, making use of a veil of ignorance entails a series of other problems. Just to give an example, behind Rawls’s version of the veil of ignorance contractors do not know anything about their level of risk aversion (Rawls 1999, p. 118). Under these circumstances, it is questionable if the resulting social contract will contain any useful information on the permissibility of various risk levels at all. I will not discuss these intricate issues in detail. For our purposes, I take this section to have shown that ISCT may be turned into an adequate framework for regulating risk imposition, but to do this would require substantial specifications.

A similar verdict applies to the problem of unstructured collective harm. To begin with, ISCT has an important advantage over the other two theories I discussed earlier. Its focus is not limited to the individual level (i.e. on a single firm’s impact on stakeholders or the universalisability of a single firm’s business dealings) but comprises an evaluation of the social situation. For instance, Donaldson and Dunfee approvingly cite Gauthier’s contractarian account of morality (Gauthier 1986), which claims that it is often individually rational to renounce to one’s short-term interests—provided that others do the same. If all contractors cooperate by showing restraint, this will allow them to attain individually and socially advantageous outcomes. It is this susceptibility to aggregate social outcomes that enables ISCT to examine the collective harm as a whole, rather than just the seemingly harmless individual contributions to it. Therefore, ISCT—unlike stakeholder ethics—is at least able to detect that there is something morally objectionable in collective risk creation under causal overdetermination. ISCT will judge certain levels of systemic risk to be socially suboptimal and bringing them about will be deemed a violation of the social contract.

Nonetheless, this is only a partial solution to our problem. To begin with, contractarianism does not answer the relevant question that finance ethics needs an answer to when evaluating systemic risk. Contractarian theories like Gauthier’s do have a convincing reply to “Why should I restrain my immediate self-interest of excessively dealing in CDS provided that all others restrain their self-interest in the same way?” (Because it is individually advantageous and rational.) However, the problem of unstructured collective harm raises a different question. The question that financial institutions contributing to systemic risk want answered is: “Why should I not pursue my immediate self-interest of excessively dealing in CDS given that (a) all others deal in CDS and (b) my self-restriction would not make any difference?” Under these circumstances, unilateral self-restraint is not advantageous.

In addition, the appeal to rational self-interest is problematic for a second reason: the systemic risk created by CDS is an externality, it affects a considerable number of agents who are not contracting parties in the CDS market. While the profits from CDS trades are mostly realised by a small group of sophisticated investors involved in those deals, the incidence of systemic risk is much more widespread. Although CDS trading is only one among many causes of the 2008 financial crisis and it is not possible to disentangle its consequences from the repercussions of the crisis in general, it seems plausible to assume that CDS trading indirectly contributed to the increases in poverty in developing countries and the wave of suicides cited in the introduction. Furthermore, in this case, the capacity of the victims to retaliate—a foundational assumption of many contractarian theories—does not obtain here.

In other words, ISCT is in the confusing situation of recognising that there is something unjust and “suboptimal” in systemic risk creation, but it does not provide any guidance on what, if anything, the individual risk creators should do about it. What it still lacks is a convincing account of individual responsibility for unstructured collective harm. I do not take these to be knock-down arguments. A contractarian may be able to fend these criticisms off by invoking that agents always benefit from signalling a stable disposition to cooperate or by adding requirements for the initial bargaining situation. However, none of these rebuttals are obvious and all of them would require additional elaboration from ISCT proponents.

Concerning the problem of limited knowledge, ISCT should in principle be suitable for establishing what finance professionals should have known. The macrosocial contract would set a lower boundary for required epistemic efforts based on principles of due care that everybody agrees on. In addition, the various microsocial contracts could add further requirements about what responsible knowledge acquisition in finance means to different business communities. Hence, this last problem seems solvable for ISCT, but yet unsolved.

Part 4: Pluralist Connection Models of Moral Responsibility

Given the identified problems, how should we proceed instead? The scope of this paper does not allow for developing a full solution that answers all of the cited criticisms, but I will use the remaining space to sketch the general lines of possible ways forward. I will propose two options and then reject the first to highlight the necessity and appeal of pursuing the second.

The first alternative is to ask policymakers to take care of systemic risk. Why even bother to develop theories that can provide private financial firms with ethical guidance when we can simply ask the state to regulate or ban overly risky business practices? Given the harm engendered by excessive systemic risk, I think that this reasoning has some merit and that better regulation is advisable. Yet, the history of financial crises shows that good regulation alone does not suffice. First, the new financial products and business practices that inflate systemic risk are complicated and time consuming to understand. Legislators and regulatory agencies are bound to remain one step behind the industry—in terms of timing as well as relevant know-how. Second, much of the regulation is limited by national or regional boundaries; the impact of large financial crises is not. States have a powerful incentive to engage in regulatory races to the bottom to attract financial firms to their countries. Third, even the best legal framework is bound to remain imperfect and leave room for avoidance. Indeed, many financial products—among them CDSFootnote 6—are deliberately designed to exploit these regulatory loopholes. Fourth, financial institutions command considerable resources, most importantly, economic know-how and political power. These resources can be employed to thwart better regulation, but they could also be put to work to promote it. The absence of a moral vision can immobilise change agents and perpetuate a state of political paralysis (Petrick 2011). This is why it is so important to develop ethical theories for banks, funds and insurers and enlist them as responsible moral agents.

What could such ethical theories look like? My proposal, sketched in the remainder of this article, is that business ethicists should draw on current research in political philosophy, in particular the debate on the responsibility for global injustices. The participants in this debate reject the assumption that culpable causation is the only link between a harm and an agent that gives rise to moral responsibility. Miller (2001), for example, advocates a pluralist “connection theory” of responsibility. He starts from the normative premise that remediable suffering and deprivation are intolerable, “in other words that where they exist we are morally bound to hold somebody (some person or some collective agent) responsible for relieving them” (p. 464). Consequently, ethicists should look beyond culpable causation and establish further grounds for assigning moral responsibility. According to Miller, three further connection types can also give rise to “remedial responsibility”, i.e. to a “special obligation to put a bad situation right” (p. 454): unintentional causal involvement, the capacity to rectify, and the fact of belonging to the same community as the victim or the perpetrator. In my view, it is useful to add a further connection type: benefitting from unjust harm. Arguably, benefit obtention is among the most powerful reasons for attributing remedial responsibility to financial institutions, since profiting was the entire purpose of the business practices that made systemic risk exceed all reasonable limits.

Iris Young’s (2011) “social connection model” adds a proactive and political dimension to the attribution of moral responsibilities. Whereas Miller’s proposal focuses on providing remedy, i.e. on intervening after the damage is already done, Young argues that assignments of responsibility should also be “forward-looking”, i.e. aim at averting future, systematically recurring harms. Many unjust harms, she claims, are not the intentional deed of a single person or collective. Instead, the problem lies at the level of social structures—patterns of social norms, practices and institutions—that channel seemingly harmless individual contributions into constellations that systematically generate harm for third parties. Hence, Young concludes that a proactive, “forward-looking” model of responsibility attribution “finds that all those who contribute by their actions to structural process with some unjust outcomes share responsibility for the injustice” (ibid., p. 96). The involved agents are not necessarily blameworthy for their individual causal contributions, but their involvement puts them under a political obligation to work towards a reform of the harmful social structures. Young identifies four groups of agents that are particularly well placed to do so: agents with political and social power; agents privileged by the current social structures; the victims of structural injustice (because of their special epistemic vantage point); as well as members of powerful collectives (e.g. political parties or rich nation states).

In my view, Miller’s and Young’s models should not be seen as rivals. Instead, they should be read as emphasising different but complementary foundations of moral responsibility. While Miller’s “connection theory” focuses on the remedy of past harms, Young’s “social connection model” shifts the attention to the prevention of future ones. In addition, both theories share a similar structural feature that one may call the “backwards-engineering” of moral responsibility: they do not start with an individual action and then try to evaluate its normative status by looking at its consequences. Instead, they first identify (actual past or potential future) harms at the social level and then assign moral responsibilities to the agents that can effectively mitigate the diagnosed problems. Moreover, the fact that Young’s proposal is forward-looking and focuses on averting potential future harms makes it particularly suitable for cases that involve the impositions of risks.

I am not alone in thinking that proposals like Miller’s and Young’s can also be applied to private firms. Recently, business ethics has tried to incorporate these findings by developing accounts of global corporate citizenship (Scherer and Palazzo 2008). Two theories explicitly introduce responsibility models like Miller’s and Young’s into business ethics seem particularly promising. Drawing on Rawls’s Law of the Peoples, Hsieh (2009) argues that the obligation to promote well-ordered political institutions also applies to corporations. Its normative basis is the duty not to harm. Similarly, Wettstein (2012) explicitly refers to social connection models and claims that multinationals must undertake collaborative efforts in addressing injustices by thwarting human rights violations.Footnote 7

So much for the normative grounds for attributing moral responsibilities for systemic financial risk. What can be said about the content of these responsibilities? Miller’s and Young’s approaches both suggest that agents with a link to an unjust (actual past or potential future) harm have an obligation to take the steps in their power that effectively mitigate this harm. While philosophers cannot, and should not, come up single-handedly with detailed lists of practical measures for each of the agents in question—the expertise for doing this mainly lies with financial economists, management scholars and, most importantly, the practitioners themselves—, business ethicists are well placed to point out broad domains in which financial firms arguably neglect their moral co-responsibilities. More concretely: I submit that banks, funds and insurers should reflect on their moral obligations linked to the creation of systemic financial risk in terms of five areas of risk responsibility.

First, there is the “classic” causal responsibility for reducing the magnitude of the risk externality imposed on third parties. In many real-world cases, the externality imposed were clearly too large. In the summer of 2008, the “non-trivial exotic derivatives book” of AIG, consisting of CDS and similar products, amounted to $2.7 trillion (FCIC 2011, p. 348). For comparison: that was more than the double of India’s GDP in the same year. Such conduct is manifestly irresponsible and AIG would have had a moral obligation to reduce its volume of risky liabilities. Arguably, similar levels of recklessness were rather the rule than the exception in most of finance in the run-up to the 2007/08 financial crisis. Claassen (2015) argues that such cases of reckless risk taking should be analysed with reference to an implicit social contract between the financial sector and the rest of society: citizens allow financial institutions to operate (often very profitable) business models, create systemic risks and impose them on society—but they may only do so in return for, and to the extent that, their operations contribute to providing citizens with an adequate standard of living. Moreover, the right to impose risks on society also presupposes that there is no other “morally preferable way to safeguard each citizen’s right to an adequate standard of living” (p. 532) than higher systemic risk in the financial sector. Finance professionals can translate Claassen’s insights into at least two ethical rules of thumb: first, a business practice that imposes a risk on society, but is only beneficial to the involved parties, is impermissible. Arguably, some of the speculation with “naked CDS” falls into this category. Second, if another less risky business model can produce the same social benefits as a high-risk option, the high-risk business model is not morally justifiable and needs to be revised. This puts finance professional under an obligation to search for and devise alternatives to current high-risk business practices.

A second class of obligations concerns the provision of remedy once some harms to society have materialised. As we have seen above, it is typically impossible to identify direct links between the risky conduct of an individual financial institution and concrete harms incurred by a third party. This mostly precludes to possibility of direct compensation. However, due to their contribution to the risk-generating structural processes that caused the harm, I submit that individual firms incur a broader obligation to consider how their business operations can help the victims that were hit hardest by the last financial crisis. Banks could, for instance, discharge such an obligation by prioritising lending to businesses in austerity-ridden countries like Greece and Spain, which still struggle to gain access to new credits. Such obligations particularly apply to financial institutions that profited from betting on a sovereign debt default and crashing housing markets in these countries.

The third domain in which banks, funds and insurers should take their own responsibilities more seriously—perhaps less obvious than the first two domains—is the area of epistemic support. They should internally foster organisational structures and corporate cultures, which enable their employees to identify and denounce instances of excessive risk taking more effectively. Virtue epistemological approaches like De Bruin (2015) already provide ethical guidance on how to approach such problems. In my view, encouraging the development and use of alternative, more cautious risk management methods will be the main epistemic co-responsibility of financial firms in the years to come. In post-crisis financial economics, there already is a growing consensus that risk analysis must overcome its blind trust in old-established mathematical models based on historical data. More recent textbooks recommend more qualitative approaches like: “Risk managers should be watching for situations where there is irrational exuberance” or for a “lack of transparency” (Hull 2015, p. 133). Stress testing, agent-based modelling or financial networks analysis are further, quantitative epistemic innovations that could contribute to better information about imminent systemic harms.

A fourth, often neglected area of risk responsibility consists in curbing the transmission and amplification of initial risk triggers. Financial contagion is a key reason why minor events can prompt full-blown financial crises. For example, the initial trigger of the 2008 financial crisis, the losses from subprime mortgages, was surprisingly small compared to the overall size of this market (Hellwig 2009; Gorton and Ordonez 2014). Yet, the involved banks, funds and insurers were strongly exposed to each other’s losses and almost all of them failed to set aside sufficiently liquid assets to back their enormous liabilities (Admati and Hellwig 2014). As a consequence, out of “13 of the most important financial institutions in the United States, 12 were at risk of failure within a week or two” (Bernanke in FCIC 2011, p. 254). On the one hand, the crisis would most likely not have been triggered without the reckless dealings in the mortgage markets. On the other hand, the rest of the financial sector should have been able to absorb these losses. This example highlights the need of attributing responsibilities for contributions to and the regulation of the mechanisms that transmit and amplify the initial risk trigger.

Finally, financial firms have the obligation to work towards political reforms and safer market designs. As discussed above, financial businesses have the power and expertise to thwart—but also to promote—better regulations as well as simpler and more transparent market designs. Consequently, their obligations in this area fall into two categories: negative obligations to refrain from obstructing socially beneficial regulatory reforms through lobbying; and positive obligations to use their expertise to support these reforms. In addition, financial firms can contribute to the abolition of harmful social practices through self-regulation. Often, their membership in business associations gives them the collective ability to set industry-wide standards and thereby co-influence the conduct of other firms (Herzog 2019). Third, financial firms can use their political clout and influence in the industry to promote lower-risk market designs. In the case of CDS trading, for instance, making it compulsory to standardise CDS contracts and clear them through a central market place could reduce uncertainty and contagion risks considerably.

This expanded set of responsibilities shows how the “connection model” approach differs from other existing proposals for the attribution of moral responsibility for systemic financial risk in the literature. De Bruin (2018) focuses on the creation of asset price bubbles and argues (correctly) that institutional investors cannot be blamed for their causal contribution to a price collapse, given the legal and market pressures that limit their scope of action. Yet, as we have seen, the responsibilities of institutional investors for systemic financial risk are not limited to direct causal contributions to asset price bubbles. For example, their room for manoeuvre is much less constrained when it comes to prudent risk management or socially-minded lobbying. James (2017) recognises that the obligations of financial institutions are more varied and wide ranging. He argues that they incur “collective responsibilities of due care” (p. 242) as well as political obligations to instigate the reforms of harmful financial business practices, “e.g., if only by declaring [their] readiness to move to the better arrangement” (p. 254). However, emphasising the structural origins of systemic financial risk and the limited powers of individual market participants, James argues that financial crises are, to a large extent, cases of “collective responsibility” and “individual innocence”.Footnote 8 “Many might be to blame for personally failing their political obligations. But if success in managing a system’s risk could only have been a shared political project, then it remains true, in a sense, that no one in particular is to blame, taken alone, when the group fails to take due care” (p. 254). In my view, the first part of James analysis is broadly correct: the key to assigning moral responsibility for systemic financial risk is to stress the need for structural reform. However, my conclusions deviate from James’ on two main points: first, the areas in which financial institutions can work towards structural change are more varied and numerous than James seems to assume and comprise responsibilities for epistemic improvements, for curbing transmission and amplification as well as for improvements in self-regulation and market design. Additionally, financial institutions are in a position to provide partial remedies for past harms. Second, some of these obligations can be discharged unilaterally and at affordable cost (most importantly, the implementation of more cautious risk management practices). Adopting a “social connection” approach, therefore, shows that, in many cases, the excuse of individual powerlessness and a continuation of the status quo is neither inescapable nor acceptable.

Conclusion

“Don’t lie, don’t cheat, and don’t be greedy”. While contemporary business ethics puts its main focus on manifestly immoral business practices when evaluating the shortcomings of financiers, these behaviours do not capture what lies at the heart of what makes the financial sector so harmful for society in times of crisis. This is why we need an ethical analysis of systemic risk. Such an analysis, however, faces three obstacles: the first is that some ethical theories cannot account for problematic risk imposition when the negative outcomes do not actually materialise. They struggle to provide criteria for acceptable probability ceilings on risk impositions. Second, systemic risk constitutes an instance of unstructured collective harm. It emerges from the interaction of a multitude of agents whose individual contributions are often imperceptible and morally unproblematic if they occurred in isolation. Third, large parts of business ethics neglect the epistemic prerequisites for some types of moral responsibility. They neglect the moral dimension of knowledge acquisition and the need for a financial virtue epistemology. These shortcomings apply to many popular business ethical theories, including stakeholder theories and Integrative Social Contracts Theory, with ISCT being the only analysed approach whose problems of which seem in principle resolvable. Moreover, systemic risk reduction cannot be achieved with tougher regulation alone. This is why business ethics needs to shoulder its part of the theoretical burden and provide businesses with better arguments for refraining from excessively contributing to the creation of systemic financial risk. In my view, a promising solution is to shift the focus to the pluralist connection models of remedial responsibility frequently used in political theory. Embracing helps practitioners to identify additional areas of risk responsibility and reject the alleged innocence of financial firms.