Introduction

When an organization takes an action that has the potential to be perceived as morally questionable, does the public’s perception of the ethics of the action depend upon whether the action is a direct result of the organization’s own behavior, or is instead implemented indirectly through an agent of the organization? For example, if a multinational’s product is produced under ‘sweatshop’ conditions abroad, does it matter to consumers whether or not the multinational owns the offending factory? Furthermore, if the action is taken by an indirect agent, does it make a difference to outside observers as to whether the organization could foresee the agent’s actions? These issues feed into our understanding of the ethical judgment of individuals, the reactions of policymakers and regulators to such actions, and for the public relations of organizations that are faced with ethical dilemmas. Our interest is in how these concerns affect the perception of the U.S. government’s indirect bailout of the bank counterparties of American International Group (AIG) during the 2008 financial crisis and what this implies for business ethics education.

Specifically, we consider the lingering controversy surrounding the U.S. federal government’s 2008 bailout of American International Group (AIG) from a moral perspective. In particular, we examine the ethics of the government assuming 79.9 % of shareholder equity as compensation for providing AIG with an $85 billion two-year loan in conjunction with the government’s decision to have AIG pay its investment bank counterparties 100 cents on the dollar for credit default swap agreements the banks held with AIG’s Financial Products division. Initially, the Federal Reserve Bank of New York (hereafter, FRBNY) prohibited AIG from releasing the terms under which it paid the investment bank counterparties and the identities of the counterparties themselves. When these details became known almost four months after the fact, the FRBNY’s actions were (i) labeled as a ‘backdoor bailout’ by the Special Inspector General for the Troubled Asset Relief Program (hereafter, SIGTARP 2009); (ii) decried in the popular press (see Krugman 2009; Jenkins 2013; Taibbi 2013); and (iii) found to be misguided and serving no legitimate purpose by a federal judge (Wheeler 2015).

As Boddy (2011, p. 255) observes, the global financial crisis has raised many ethical issues concerning who pays for the damage inflicted and who is responsible for causing the crisis. In this paper we focus on the U.S. taxpayers, AIG’s shareholders, and bond insurance counterparties, and show that answering the question of who benefitted from and paid for the AIG bailout is not straightforward from a moral perspective. Consequently, we employ behavioral ethics to examine the AIG bailout.Footnote 1 We assess the ethics of the government’s actions in the AIG bailout in a laboratory setting by comparing subjects’ evaluation of the ethics of an indirect bailout of insured bondholders with a counterfactual where the insured bondholders are directly bailed out under similar terms. A separate set of subjects was asked to ethically compare both scenarios jointly and to judge whether the government had a dubious motive for the backdoor bailout of bondholders. We find a form of preference reversal in that when the two scenarios are compared jointly, subjects found the (actual) indirect bailout to be more unethical, whereas in separate and independent assessments subjects found the direct bailout to be more unethical. Preference reversals of this type are of concern because they indicate that eliciting ethical judgments may be sensitive to the mode (joint versus separate) in which equivalent moral issues are evaluated by decision makers. Yet, if ethical judgment is not invariant to the mode of evaluation, then the ability of decision makers to consistently undertake ethical judgments may vary with the way in which a decision is framed.

The paper is organized as follows. We begin with a literature review of the use of the joint/separate and direct/indirect mode of evaluation to elicit moral preferences. This is followed by a brief history of the government’s intervention with AIG. This sets up our description of the behavioral experiment and associated findings. We then focus on the pedagogical, policy, and business implications of our findings.

Related Literature

Preference reversals occur when an individual’s preferences vary with the mode of elicitation. A simple example is the consistent finding that subjects seek to sell an item at a price higher than the price they would be willing to pay to buy the same item. In this example, the mode of elicitation varies with the role of buyer or seller. Preference reversals are extensively studied within the social sciences because they indicate that the basis for making judgments—a subject’s preferences—may not be stable. Within the context of business ethics, Elm and Radin (2012) are particularly concerned with contradictions such as ethical preference reversals because they challenge the fundamental assumption that ethical decision making is meaningfully different from other decision-making processes that exhibit similar contradictions. To this we add that ethical preference reversals are especially troubling because they can foster misleading practices, whereby control over the evaluation mode facilitates a deceptive strategy for hiding morally dubious actions.

For example, in 2005, pharmaceutical giant Merck sold the rights to manufacture and market one of its cancer drugs to a much smaller firm with a lower public profile, Ovation Pharmaceutical. After the sale, Ovation raised the price of the drug tenfold on cancer patients (Berenson 2006). In examining subjects’ perception of the ethics of Merck’s actions, Paharia et al. (2009) allowed subjects to morally rate the Ovation-as-intermediary price gouging scenario versus a counterfactual where Merck directly raised the price of the drug on cancer patients. They found evidence of an ethical preference reversal. Specifically, when subjects from the same population were separately presented with either the scenario of indirectly attaining the revenue associated with a price increase by selling the marketing rights to another firm versus directly increasing the price itself, subjects found the direct price increase to be significantly more unethical. In contrast to these separate evaluations, when subjects were presented with both scenarios jointly, they found the indirect price increase through the sale of marketing rights to be more unethical. Joint evaluation makes the intent of indirect agency transparent. Furthermore, if the indirect action is articulated in terms of a dubious motive (e.g., avoiding the negative reputation that might result from a large increase in drug price), this increases the transparency of indirect agency, thereby establishing causal responsibility for unethical behavior.

Coffman (2011) experimentally tests for the formation of a deceptive strategy that capitalizes on the difference between joint and separate ethical judgments. In Coffman’s study, ethical opinions are not elicited; instead, an outside observer is given the costless opportunity to punish direct and indirect behavior. Coffman uses a version of the dictator game, which is a two-player situation in which the first mover (the dictator) determines the split of a surplus (usually some cash amount) and the second player (the recipient) receives the split determined by the first mover. Prior experimental results indicate that outcomes where the first mover takes 70 % or more of the surplus are regarded by others as unfair or unjust (Camerer 2003). In Coffman’s extension, the first mover can either directly determine the division of the surplus in the usual way that a dictator game is played, or the first mover can sell the right to be the dictator to an intermediary. In this way, the intermediary takes on a role akin to that of an indirect actor. Selling the right to be the dictator obscures the first mover’s deceptive strategy to secure an unfair payoff, as the intermediary is required to cover the cost of becoming the dictator when determining the division of surplus. The recipient is harmed either directly or indirectly by the actions of the first mover. Finally, an outside observer is given the task of specifying whether and how the first mover is to be punished for all possible plays of the game [e.g., the punishment is elicited with a strategy method (Selten 1967)]. Coffman finds that the degree and frequency of punishment of the first mover significantly decreases when an intermediary is used. Of particular concern is that it appears that first movers understand the tempering effect of intermediaries on punishment and so first movers frequently employ the deceptive strategy of using intermediaries to avoid the punishment associated with unethical behavior.

The potential, therefore, exists for the joint evaluation of (in)direct actions within a morally charged situation to yield different ethical judgments than separate evaluations of the same actions. Indeed, Bazerman et al. (2011) opine that joint evaluations of morality are less emotionally charged than separate evaluations, and that emotions play too strong a role in separate decision making. Joint evaluation prompts people to consider the primary agent’s motives by drawing attention to the more straightforward alternative of direct action (Paharia et al. 2009). Heightened awareness of the sometimes dubious motivations for acting indirectly may, therefore, be a useful safeguard against the abuse of power. Bazerman and Gino (2012) contend such shifting of modes of thought can lead to profound differences in how individuals and societies make ethical decisions. For example, they conjecture that joint evaluation provides the means for best understanding the actions of those who engaged in the clearly illegal and unethical behavior involved in the 2008 financial crisis. In this way, behavioral ethics can be used to gauge the actions of those who indirectly cause great harm. We test this via an examination of the AIG intervention during the 2008 financial crisis. Indeed, in contrast to the cancer drug scenario examined in Paharia et al. (2009), where one can only make conjectures about Merck’s motivation for indirect versus direct action, it is shown below that AIG received explicit instructions from policymakers to hide its indirect action. Furthermore, indirect action was taken in lieu of direct action that was readily available to policymakers at the time.

Background to the AIG Intervention

Founded in 1987, AIG’s Financial Products division (hereafter, AIGFP) was originally in the business of taking on the risk of commercial transactions (e.g., interest or exchange rate fluctuations). Much later, AIGFP engaged in credit default swaps (CDSs) where, for a fee, AIGFP would agree to pay bondholders whatever portion of the underlying debt obligation remained in the event that the bonds defaulted. In addition to this ‘credit risk’ of a default, writing CDSs also exposed AIGFP to ‘collateral risk,’ in that AIGFP would be required to post collateral with its counterparty should AIG or the associated pool of bonds suffer a downgrade in its credit rating.

With the advent of the subprime mortgage crisis in 2007, AIGFP was hit by numerous collateral calls by its CDS counterparties. In 2008, through September alone, it posted $32.8 billion in collateral to counterparties (SIGTARP 2009, p. 8). AIGFP was not the only bond insurer experiencing liquidity difficulties associated with collateral calls stemming from downgraded mortgage-backed securities. Under severe credit market conditions, it is normal for the bank counterparty and the insurer to settle differences about bond valuations via a compromise in which the insurer pays a discount from the face value of the contracts, something less than 100 cents on the dollar. The amount below 100 cents on the dollar that the counterparty receives is known as a ‘haircut.’ The bank has to write off the haircut as a loss.

The details of government intervention with AIG are shown in the timeline in Fig. 1.

Fig. 1
figure 1

AIG Intervention Timeline

On September 16, 2008, the day after the Lehman bankruptcy, the U.S. government intervened in AIG, providing it with an $85 billion dollar two-year loan at a punitive annual interest rate of 14 percent.Footnote 2 In exchange, the government took 79.9 percent ownership of AIG.Footnote 3 Given the nearly 80 % loss of shareholder equity, AIG was effectively nationalized.

The AIG intervention was the most money ever expended by the U.S. government to save a private company. But was saving AIG the actual purpose of the intervention? A report of the Office of the Special Inspector General for the Troubled Asset Relief Program (SIGTARP 2009) unequivocally labeled the AIG intervention as a “backdoor bailout” of AIGFP’s bank counterparties to the CDSs (Barofsky 2012, p. 187). In fact, the government eventually used $62 billion to pay off AIGFP’s entire CDS obligations at 100 cents on the dollar (i.e., no haircuts) by having AIG make up the difference between the collateral already given to the counterparties and the par value of the CDSs (see Table 1). The structured investment vehicle (SIV) used to make the payments was called Maiden Lane III. Moreover, the FRBNY prohibited AIG from reporting the identities of the bank counterparties and the amounts of payments made to each of them. It was only after Congressional insistence that the names and amounts were finally released, approximately four months after the payments were made.

Table 1 Total Payments to AIG Credit Default Swap (CDS) Counterparties (in $U.S. billions)

The consensus is that most of the CDSs insured by AIGFP had a market value of less than 50 cents on the dollar (Boyd 2011, p. 293) and AIG had already provided a significant amount of collateral to its bank counterparties. When pressed by Congress as to why payment was made at full value, Treasury Secretary Timothy Geithner—who was the President of the FRBNY at the time—stated that, “the financial condition of the counterparties was not a relevant factor” [in the decision to bail out AIG] (SIGTARP 2009, p. 15). By contrast, in his memoire, Stress Test, Geithner (2014, pp. 215, 219, 246, 409) called the payment of AIG’s counterparties at 100 cents on the dollar a “no brainer” and part of what was to become a ‘no-haircuts-in-a-panic doctrine.’ The rationale was that haircuts would send a destabilizing signal that more haircuts were coming, encouraging a run on financial firms, thereby acting as a panic accelerant for the financial system as a whole.Footnote 4

The loss in shareholder equity and subsequent payments to the counterparties formed the basis for a lawsuit filed by Starr International (2013), a charity that was one of the largest shareholders of AIG.Footnote 5 Starr is run by former AIG CEO Maurice “Hank” Greenberg. Presiding Judge Thomas C. Wheeler (2015, p. 6) noted that, “since most of the other financial institutions experiencing a liquidity crisis were counterparties to AIG transactions, the Government was able to minimize the ripple effect of an AIG failure by using AIG’s assets to make sure the counterparties were paid in full on these transactions.” In his ruling, Judge Wheeler wrote (p. 7), “The Government’s unduly harsh treatment of AIG in comparison to other institutions seemingly was misguided and had no legitimate purpose, even considering concerns about ‘moral hazard.’ Having considered the entire record, the Court finds in Starr’s favor on the illegal exaction claim.” At the same time, Judge Wheeler awarded no damages to Starr, concluding that AIG’s shareholders’ benefit was to avoid bankruptcy, and to “live to fight another day.”

The SIGTARP characterization of the AIG ‘backdoor bailout,’ various media reports (e.g., Krugman 2009; Jenkins 2013; Taibbi 2013), and the judge’s subsequent ruling on the Starr International lawsuit all raise the ethical issues of why did the government choose to (a) make AIGFP’s counterparties whole at 100 cents on the dollar, (b) do so indirectly via AIG rather than directly, and (c) destroy AIG shareholder equity by this indirect means? Moreover, a direct counterfactual exists to the government’s indirect bailout of counterparties via its intervention in AIG. Specifically, the government could have bailed out AIGPF’s counterparties directly, thereby ending the collateral calls that were siphoning off AIG’s reserves. After all, AIG’s domestic bank counterparties also received capital injections directly from the government through TARP (see Table 1), and the Fed eventually purchased mortgage-backed securities directly from banks through its quantitative easing program. Given the existence of this direct bailout alternative, in what follows we use the direct/indirect and joint/separate methodology to assess the ethics of the government during the AIG intervention.

Experiments

We conducted two experiments to examine the ethics of the government’s AIG intervention during the 2008 financial crisis. We used a joint/separate evaluation design to assess subjects’ ethical evaluation of the government’s actual indirect (backdoor) bailout of investment banks through a large bond insurance firm versus a counterfactual where the government directly bails out investment banks. In either scenario, the government imposes hardship on the shareholders of the bond insurer’s parent company who, as a consequence, lose a percentage of their ownership stake in the parent company as compensation to the government for the bailout.

In the first experiment, we evaluated the effect of indirect agency while keeping the magnitude of harmful consequences on shareholders constant, by testing whether

  1. (1)

    when evaluating the facts separately, individuals judge the indirect bailout of investment banks through a large bond insurer more leniently than a direct bailout of the investment banks;

  2. (2)

    when evaluating the facts jointly, the direct versus indirect effect is eliminated and the indirect bailout of investment banks through the bond insurer is considered more unethical.

In the second experiment, we evaluated the effect of indirect agency and changed the magnitude of harmful consequences on shareholders in the direct versus indirect case.

In both experiments, subjects in the separate treatments were presented with either a description of a direct or indirect intervention and were asked to rate the ethics of the government’s actions on a ten-point scale ranging from 1 (the government’s behavior was not at all unethical) to 10 (the government’s behavior was very unethical). This rating scale is shown in Fig. 2. Differences between the two treatments were then tested statistically by comparing the 1–10 ratings of the indirect bailout with the 1–10 ratings of the direct bailout. In the joint treatment, subjects were instead asked to simultaneously compare the indirect bailout with the direct counterfactual. A bipolar scale, also shown in Fig. 2, was used for the joint treatment, with one unit rating increments from −5 to 0 to 5. This scale is used because subjects themselves were asked to directly make the comparison, whereas in the separate treatment the comparison is made by statistical means. In the joint rating, negative scores indicated that subjects found the direct bailout to be more unethical, with a score of −5 corresponding to the direct bailout being ethically much worse that the indirect one. A score of 0 corresponds to the rating that the two cases are equally ethical/unethical. Positive scores reflected a rating that the indirect bailout was more unethical, with a score of 5 indicating that the indirect bailout is ethically much worse than the direct one. Hence, a statistically positive score indicates that in the joint comparison the indirect bailout is rated to be more unethical than the direct bailout (a one-sample t test was used with 0—the scale midpoint—as test value).

Fig. 2
figure 2

Rating Scales

The experiment headings below are denoted by two numbers, the first being the shareholders’ loss of equity in the direct intervention (the counterfactual) and the second number corresponding to the (actual) shareholder loss of equity in the indirect intervention. A total of 541 subjects participated in the experiment, 225 in the 80–80 treatment and 316 in the 50–80 treatment. Student participants were asked to complete the survey in exchange for extra points toward their class grade. Experiments were conducted at Virginia Commonwealth University, with students recruited mostly from upper level business and economics courses. After completing the survey, students answered a questionnaire about their personal characteristics. The subjects’ sample was about 60 percent male. As is quite common in behavioral studies of ethics (e.g., Bateman et al. 2002; O’Fallon and Butterfield 2005), we found no statistically significant differences for our findings when the sample was subdivided according to gender. For this reason, the results below are presented for the sample as a whole. The sample mean age was 21.56 (with a standard deviation of 4.0). The subjects’ sample was diverse and close to the U.S. population composition, as 52 % of the subjects identified themselves as White, 18 percent Afro-American, about 13 % as Asian, and 6 percent Hispanic with the remainder being either multiracial or giving no response.Footnote 6

Experiment 80–80

Procedure

225 individuals (79 females, 146 male, Mean age = 22.57, SD = 4.02) were randomly assigned to one of five different conditions:

  • A: Separate evaluation—direct

  • B: Separate evaluation—indirect

  • B′: Separate evaluation—indirect + foreknowledge

  • AB: Joint evaluation

  • AB′: Joint evaluation

When facing scenario A or B separately or both scenarios jointly, participants read the following background information:

Investment banks keep bonds on their books that are risky, but can be extremely profitable. The investment banks design and create these bonds themselves. Because investment banks cannot be insured by the government, they purchase insurance on their assets from private bond insurance firms. Many of the shareholders of the parent companies of these bond insurance firms do not have a good understanding of the type of bonds their firms insure, owing to the bonds’ complexity and that bond insurance is not regarded as the primary business activity of the parent company.

When the bonds of several investment banks go bad at the same time, it is unlikely that a bond insurance firm can reimburse all of the investment banks at the insured value (100 cents on the dollar). Recently a bond insurance firm (Company X, in the Table below) and an investment bank entered into negotiations because $45 billion of the bank’s bonds had gone bad. These negotiations resulted in the insurance firm paying the investment bank 14 cents on the dollar for bonds that were originally insured for the total value of $45 billion. As a result of this, shareholders of the bond insurer’s parent company X lost approximately 50 % of the value of their ownership stake in the insurer’s parent company. This privately negotiated resolution and two others like it, for parent companies Y and Z, are summarized in the table below:

Bond insurer parent company

Amount insured by investment banks

Amount received by investment banks from the bond insurer

X

$45 billion

14 cents on the dollar

Y

$3.5 billion

29 cents on the dollar

Z

$1.4 billion

60 cents on the dollar

Two months later a much larger bond insurance firm, Firm W, is having trouble covering its $63 billion in insurance obligations to investment banks. The government decides to pay $63 billion to bail out the investment banks, rather than Firm W, the bond insurer. That is, the government pays the investment banks 100 cents on the dollar for the $63 billion in bonds that were insured by Firm W. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government.

When facing scenario B, on the other hand, participants read (the text in italics was replaced with the following text):

The government decides to pay $63 billion to bail out the large bond insurer, Firm W, and then instructs Firm W to conduct a back-door bailout of the investment banks, in which Firm W pays the investment banks 100 cents on the dollar for the $63 billion insured. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government.

When facing scenario B′, which casts the government’s action through a dubious motive, participants read (the text in italics was replaced with the following text):

The government decides to pay $63 billion to bail out the large bond insurer, Firm W, and then instructs Firm W to conduct a back-door bailout of the investment banks, in which Firm W pays the investment banks 100 cents on the dollar for the $63 billion insured. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government. One possible motive for the government using taxpayer money to bail out Firm W and having it pay 100 cents on the dollar to the investment banks is that the government is worried by the negative press that a bailout of the investment banks would cause.

All participants were asked “On a scale of 1 (not at all unethical) to 10 (very unethical), how unethical do you think the government’s behavior was in this decision?”

In scenario AB (AB′), the text in italics was replaced by the following:

Consider the following two cases:

  1. A.

    The government decides to pay $63 billion to bail out the investment banks, rather than Firm W, the bond insurer. That is, the government pays the investment banks 100 cents on the dollar for the $63 billion in bonds that were insured by Firm W. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government.

  2. B.

    The government decides to pay $63 billion to bail out the large bond insurer, Firm W, and then instructs Firm W to conduct a backdoor bailout of the investment banks, in which Firm W pays the investment banks 100 cents on the dollar for the $63 billion insured. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government.

  3. B′.

    The government decides to pay $63 billion to bail out the large bond insurer, Firm W, and then instructs Firm W to conduct a backdoor bailout of the investment banks, in which Firm W pays the investment banks 100 cents on the dollar for the $63 billion insured. As a result of this, shareholders lost 80 % of the value of their ownership stake in bond insurer firm W’s parent company as compensation to the government. One possible motive for the government using taxpayer money to bail out Firm W and having it pay 100 cents on the dollar to the investment banks is that the government is worried by the negative press that a bailout of the investment banks would cause.

In scenarios AB and AB′, participants were asked “In which case would the behavior of the government have been more unethical?” Possible ratings ranged from −5 “In terms of ethics, Case A is much worse than Case B (B′)” to 0 “In terms of ethics, the two cases are equally ethical/unethical,” to +5 “In terms of ethics, Case B (B′) is much worse than Case A.”

Notice that in this experiment, the magnitude of harm (80 % of ownership value) does not change in the cases A versus B (B′). Only the direct versus indirect action by the government is different in the two scenarios. The statistics for each treatment are given in Table 2 and the associated results are presented and discussed below.

Table 2 Results for the 80–80 Treatment 80 % loss of shareholder equity in the direct treatment 80 % loss of shareholder equity in the indirect treatment

Results

When judging separately case A and B, the results from the survey reveal a significant effect of directness, as t(84) = 2.89, p < 0.05, indicating that the direct bailout in case A was rated as significantly more unethical (M = 6.8, SD = 2.24) than the indirect bailout in case B (M = 5.31, SD = 2.53). This holds even though the government’s actions in case B are labeled as a “backdoor bailout.” By contrast, the results from the survey do not support a significant effect of directness when combined with foreknowledge (A and B′), as t(88) = 1.09, p = 0.27, indicating that case A was not rated as more unethical than case B′. Foreknowledge of a dubious motive reduces the effect of indirect agency.

The results for the joint evaluation of A and B were analyzed with a one-sample t test using the midpoint of the scale (zero) as the test value. The results—t(46) = 2.75, p < 0.01—indicated that indirect bailout scenario B was rated as significantly more unethical (M = 1.0, SD = 2.49) than direct bailout scenario A, when the two scenarios were judged together. Using the same test for A and B’, the results—t(45) = 4.38, p < 0.0001—indicated that the indirect foreknowledge scenario B′ was rated as extremely significantly more unethical (M = 1.35, SD = 2.09) than the direct case A, when the two scenarios were judged together.

Discussion

When the scenarios were evaluated separately, participants considered the government’s direct bailout of the investment banks more unethical than the indirect bailout through bond insurer firm W. This means that participants were more lenient when separately judging harm caused by an indirect action. On the other hand, when the scenarios were evaluated jointly, a reversal of judgment was observed; implying that the unethical nature of an indirect action becomes more transparent. Notice that when adding the foreknowledge of a dubious motive, the difference in separate judgments of directness and indirectness disappears, as the participants did not distinguish between a direct bailout of investment banks and an indirect bailout with foreknowledge of a dubious motive. This fact is reinforced by the result that when judged together, scenario B′ is considered extremely more unethical than A.

Experiment 50–80

In this experiment, the two cases A and B (B′) differ in two ways, whereas previously the difference was only in terms of direct versus indirect action (and foreknowledge of a dubious motive for the B′ case). Now the magnitude of harm changes to a 50 % loss of ownership value in the direct case, A, but remains 80 % in the indirect case, B and B′. The 50 % loss in equity is closer to the 30 % loss that shareholders experienced in the Citigroup bailout. In the previous 80–80 experiment, we kept the magnitude of harm on shareholders constant at 80 % and examined only the effect of changing the indirectness of the government action. From a policy perspective, it would be useful to know if moral sentiment about the bailout would change if the consequences for shareholders had been more in line with other bailouts.

Procedure

316 individuals (133 females, 183 male, Mean age = 21.23 SD = 2.09) were randomly assigned to one of five different conditions:

  • A: Separate evaluation—direct + 50 % loss of shareholder equity

  • B: Separate evaluation—indirect + 80 % loss of shareholder equity

  • B′: Separate evaluation—indirect + foreknowledge + 80 % loss of shareholder equity

  • AB: Joint evaluation

  • AB′: Joint evaluation

Subjects in the separate treatments A versus B (and A vs. B′) were presented with the same ten-point scale as in the 80–80 experiment, and those in the joint treatment (AB or AB′) were again presented with the bipolar −5 to 0 to 5 scale (see Fig. 2). The statistics corresponding to these comparisons are given in Table 3.

Table 3 Results for the 50–80 treatment 50 % loss of shareholder equity in the direct treatment 80 % loss of shareholder equity in the indirect treatment

Results

In the separate evaluation of A and B, when the loss in shareholder equity is reduced from 80 to 50 % for the direct bailout, the results from the survey do not support a significant effect of directness, as t(131) = 0.39, p = 0.69, indicating that case A was not rated as more unethical than case B. Similarly, the results from the survey do not support a significant effect of directness when combined with foreknowledge (A and B′), as t(125) = 0.30, p = 0.76, indicating that case A was not rated as more unethical than case B′. Once again, foreknowledge of a dubious motive reduces the effect of indirect agency.

Given a 50 % loss in shareholder equity for case A, the results now indicate that the indirect scenario B was rated as extremely significantly more unethical (M = 2.08, SD = 2.49) than case A, when the two scenarios were judged jointly—t(61) = 6.57, p < 0.0001. The results were again analyzed with a one-sample t test using the midpoint of the scale (zero) as the test value. Moreover, indirect scenario B′ was rated as extremely significantly more unethical (M = 2.48, SD = 1.79) than case A—t(59) = 10.73, p < 0.0001—when the two scenarios were judged together.

Discussion

When separately evaluating the direct versus indirect bailout of banks through bond insurer firm W under greater harm in the indirect case (80 %) as compared to the direct case (50 %), participants were not able to distinguish the two in terms of ethicality. The increase in harm done to shareholders in the indirect case appears to offset the lesser harm done to shareholders in the direct case. On the other hand, when the two scenarios were evaluated together, a clear judgment was observed, as scenarios B and B′ are considered extremely more unethical than A, implying that the indirectness of the bailout is judged as less important than the magnitude of its harmful consequences.

Implications

We thank an anonymous reviewer for helping us to flesh out this section.

In both experimental treatments, subjects were told that the scenario they just evaluated was based on a real-world event. During the survey, subjects were also asked to state whether they had any knowledge of what real-world event that was. Even though the word bailout was used in the description of the scenario, only 4.25 % of the subjects correctly identified AIG as the real-world event. The majority were only able to generally connect the scenario to the 2008 financial crisis and the government interventions in the banking and/or financial system. Thus, the ethical preference reversal observed in the experiment cannot be attributed to prior knowledge by the subjects about the AIG bailout.

The AIG intervention explored here provides valuable lessons as to how unethical conduct results from cognitive restructuring. Bandura (2002) contends that moral agency has dual aspects: inhibitive—the power to refrain from behaving unethically, and proactive—the power to behave ethically. The mechanisms that prevent either of these include the cognitive restructuring of unethical conduct into a benign or worthy one by: moral justification, sanitizing language and exonerative social comparison, disavowal of personal agency in the harm one causes by diffusion or displacement of responsibility, disregarding or minimizing the injurious effects of one’s actions, and attribution of blame to, and dehumanization of, those who are victimized. Safeguards must be put into place that uphold ethical behavior and renounce unethical behavior.

In the AIG intervention, sanitizing language and euphemistic labeling were widely used. For example, in order to limit taxpayers’ ire, Secretary of the Treasury Hank Paulson (2010, pp. 233, 237, 240) asked all concerned to characterize the Fed’s actions toward AIG as rescues or interventions, but not bailouts. In addition, exonerating comparisons are often used to elicit approval of morally questionable actions. As Fed Chairman Ben Bernanke testified before Congress, “I share your concern. I share your anger. It’s a terrible situation….But we’re not doing this to bail out AIG or their shareholders, certainly. We’re doing this to protect our financial system and to avoid a much more severe crisis in our global economy.” Finally, disregarding or minimizing the injurious effects of one’s actions and the attribution of blame to those who are victimized also occurred. “There are a lot of things that have happened in the last eighteen months, but what has happened at AIG is the most outrageous …. No one cares about the shareholders of AIG. No one feels the slightest obligation to people who led us into these difficulties” (Obama administration economic advisor Larry Summers in Suskind (2011, p. 216).

Our results on direct versus indirect and joint versus separate moral evaluations of the AIG bailout also fall within the domain of the moral emergencies that are popular dilemmas in ethics classrooms (Appiah 2008), as exemplified by ‘trolley’ or ‘footbridge’ problems. In these exotic moral dilemmas, ethical judgment is influenced by perceived moral differences between harmful omission versus harmful action, even though the ultimate result is the same (Lapsley and Hill 2008). Appiah (2008, pp. 96–97) identifies moral emergencies as having the following four features: they involve (i) limited (i.e., instantaneous) decision time that disallows the opportunity to gather more information; (ii) a clear and simple set of options; (iii) high stakes that narrow the ranges of options to consider; and (iv) optimum placement in that the decision maker bears responsibility because no one else is in better position or more equipped to act. By pointing out these features, Appiah casts doubt as to whether much is learned from such highly unlikely and idiosyncratic trolley or footbridge problems, because they require the assumption that what is learned from imaginary scenarios mirror our responses to real ones.

By contrast, the AIG intervention is not a theoretical construct. It was instead a real-world phenomenon and the actions of the participants involved have subsequently been evaluated by a federal judge. The AIG bailout satisfies Appiah’s four features of a moral emergency. First, heightened sensitivities to adverse market reactions immediately following the ‘Lehman weekend’ meant that government officials needed to act almost instantaneously to address AIG’s situation, because it was felt that without an intervention AIG may not last out the week (Paulson 2010, p. 217). Second, the option considered was clearly identified by a similarly structured but failed attempt at a private resolution of AIG’s collateral crisis during the same weekend by Goldman Sachs and JPMorgan Chase that the government then adopted and augmented with an additional $10 billion in funding (Barofsky 2012; Bernanke 2015). Third, the stakes were extremely high for both AIG, its counterparties, and perhaps the world economy, given AIG’s role in greasing the wheels of world commerce via its non-AIGFP businesses involved in retirement savings and pension funds, transoceanic shipping insurance, airline insurance and aircraft leasing, employee healthcare/benefits, etc. Finally, by their own assertion, the intervention in AIG by Treasury and Fed officials satisfies the condition of optimal placement, because private attempts at rescuing AIG had failed. No other entities could possibly act in rescuing AIG.

Within the context of teaching business ethics, Elm and Radin (2012) argue that when ethical decision making exhibits contradictions—such as the preference reversals identified in this study—there may be no distinction between ethical and other types of decision-making processes. Elm and Radin (p. 325) further contend that this means that the field of ethical decision making as it exists among the social sciences may be impoverished by not being connected to research on decision making in general. We add that if this is the case, then it suggests that ethical decision making should be regularized within the social sciences and its extension to business. If ethical decision making is not all that different, then it should not be held apart and the ethical content of a decision should be considered as regularly as the accounting, economic, strategic, etc., dimensions of a business decision. Ethics is no more separatable from business decision making than are the aforementioned functional areas of business. Our study supports Elm and Radin’s thesis via the identification and consideration of three decision-making contradictions within an ethical context: preference reversals, framing effects, and dual processing. We briefly discuss each in turn.

First, ethical preference reversals raise the issue of the extent to which ethical judgment takes place with respect to subjects’ preexisting moral constructs, or if instead moral criteria are constructed within the context of eliciting ethical judgment. That is, preference reversals suggest that subjects’ moral reasoning may be unstable, depending upon intuitions that vary with the mode of evaluation. Moreover, moral judgments can only be made in separate or joint evaluation mode. If ethical judgments are not stable across these modes, then either decision makers apply different ethical principles in the separate and joint treatments, or they are unable to apply the same principle uniformly over the treatments. Ethical preference reversals, therefore, raise the uncomfortable reality that what is considered ethical may depend upon how a particular decision is presented. In other words, those who have the power to design how an ethical dilemma is approached may be able, in a very real way, to dictate what is viewed as ethical and what is not (Table 4).

Table 4 Summary of implications for pedagogy

Indeed, Dedeke (2015, p. 438) suggests that the framing of a moral issue by the decision maker deserves perhaps even more attention than moral awareness does, as the first stage of moral decision making, because moral decision making is influenced by how issues are framed. Dedeke, therefore, recommends that ethics training include examples of how different framing of issues lead to different outcomes. The sensitivity of direct versus indirect moral judgments to whether they are framed jointly or separately is such an example. As our study shows, framing can introduce an uncomfortable degree of relativism into the question of what is ethical or unethical. It also implies that meta-processes can be consciously manipulated in order to make a decision appear more ethical or more moral. Hence, while framing is a traditional subject in judgment and decision theory (e.g., Baron 2008), our results suggest that cases such as the AIG bailout are needed within business ethics courses so that the effects of framing are understood within an ethical context. Only then, as Dedeke suggests, can close attention be paid to the contribution of framing when unethical decisions occur. Business ethics education that includes the effects of framing can, therefore, lead employees to challenge framing practices.

Finally, this leads us to the application of dual processing to moral decision making. Following the terminology of Stanovich and West (2000), dual processing is characterized by System 1 and System 2 methods of reasoning. System 1 processing is embodied by associative, experiential, implicit, and tacit heuristics. By contrast, System 2 processing is analytical, conscious, deliberative, and ‘rational.’ In particular, a decision maker may use System 2 processing to override System 1 decisions via judgment upon reflection (Lapsley and Hill 2008). In our study, the limited information presented in the separate framework lends itself to System 1 processing, while the joint framework permits System 2 processing. Our results, therefore, illustrate the ability of System 2 processing to monitor System 1 judgments, as well as the limitations of moral heuristics in the absence of considering counterfactuals, whether or not they are explicitly provided. Provis (2015) calls this “hypothetical thinking.” System 2 processing facilitates connecting the dots, but it does so at a tradeoff of requiring much more cognitive effort than System 1 processing does. Including separate versus joint and direct versus indirect examples within business ethics education allows hypothetical reasoning to become more automatic, thereby reducing the associated cognitive effort.

Conclusion

Charles Ferguson, the 2011 Oscar-winning director of the financial crisis documentary, Inside Job, lamented during his acceptance speech that, “not a single financial executive has gone to jail” in relation to behavior that led to the financial crisis.Footnote 8 Ben Bernanke has similarly concluded that there should have been more accountability at the individual level because, “everything that went wrong or was illegal was done by some individual, not by an abstract firm” (Page 2015).Footnote 9 Yet, the statute of limitations ran out long ago for dubious financial practices undertaken prior to the crisis. However, no statute of limitations exists for judging the ethics of actions committed during the financial crisis. It is also well accepted among business ethicists that judgment pertaining to the legality of an action can be separate from judgment about the action’s morality. Given that the new millennium has already experienced two crises related to dubious business practices, the need to understand how business and government decision makers form ethical judgments to preclude such crises is greater than ever.

We use the direct/indirect and joint/separate evaluation mode to examine one of the most controversial decisions made by the U.S. government during the financial crisis; namely, the decision to bail out American International Group (AIG) in the fall of 2008. This action, taken collectively by the U.S. Department of Treasury, the Federal Reserve, and the Federal Reserve Bank of New York, was controversial for at least three reasons. First, it was the largest government bailout of a public corporation, eventually totaling some $182 billion. Second, shareholders lost 79.9 percent of their ownership stake in AIG as compensation for the intervention. Indeed, a large AIG shareholder, the Starr International charity, sued the U.S. government over these terms, and a federal judge found that the exaction of shareholder equity was illegal (Wheeler 2015). Third, the newly government-controlled AIG paid off bank counterparties at 100 cents on the dollar for collateralized debt swaps (bond insurance) that the banks held with AIG’s Financial Products division (AIGFP) and had a market value of less than 50 cents on the dollar. This final action was characterized as a ‘backdoor bailout’ of the bank counterparties by the Office of the Special Investigator General of the Troubled Asset Relief Program (SIGTARP 2009) and as misguided and serving no legitimate purpose by the judge in the Starr lawsuit (Wheeler 2015) .

Our results show that when the indirect bailout of banks is jointly compared with a counterfactual where the government directly bails out banks, subjects judge the indirect bailout to be far more unethical. Joint evaluations can establish causal responsibility for indirect unethical behavior. Moreover, Coffman (2011) shows that those responsible for an unethical action understand that they can avoid being associated with unethical behavior by employing a deceptive strategy that designates the action to an intermediary. Unless outside observers have the benefit of being presented with the possibility of direct action, it is difficult to trace the indirect action back to the party ultimately responsible for the questionable behavior. It appears that the government officials involved with the decision to pay 100 cents on the dollar to AIG’s counterparties understood this when they prohibited AIG from revealing the extent of the payments made to the counterparties or even to identify the counterparties. By contrast, our study suggests that government officials may have been better served by transparently bailing out AIG’s counterparties directly, if this was their ultimate concern, as has been claimed by SIGTARP, and to do so at terms more in line with other bailouts; e.g., the $99 billion bailout of Citigroup in which shareholders lost 30 % of their equity or the FDIC’s insistence that Washington Mutual’s bondholders receive 25 cents on the dollar. Indeed, the government eventually did buy distressed mortgage-backed securities directly from banks through its quantitative easing program. In the end, the government ultimately made $22 billion on the AIG bailout (Paletta and Scism 2014), so it could certainly have afforded to make less by taking a lower share of AIG.Footnote 10

Most importantly, what we find is that ethical judgments are not consistent or stable, and can, therefore, be manipulated by those with the power to frame ethical dilemmas. This is what our analysis of the AIG bailout illustrates and what our experiments corroborate. Moral decision making can exhibit inconsistencies that are commonly associated with decision making in the social sciences. Specifically, moral decision makers can exhibit preference reversals, framing effects, and dual decision processes.

Demonstrating that moral decision making and decision making in the business arena or social sciences can exhibit similar inconsistencies is one thing. The societal implications of such inconsistences, however, define an entirely different issue. Individuals that seek a higher price when acting as sellers as compared to the price sought when acting as buyers are not violating the tenet of “conforming to the basic rules of the society, both those embodied in law and those embodied in ethical custom” (Friedman 1970, p. 33). On the other hand, seeking a higher price for cancer drugs or bond counterparties via purposeful indirect action adversely affects the health of patients and violates shareholders’ rights, respectively. Capitalizing on one’s ability to undertake morally dubious behavior by hiding it via indirect action is an attempt to violate social mores without incurring the consequences. It is, therefore, imperative that the effects of such strategies be addressed within the business ethics classroom, so that role of power in framing issues can be discussed and students can become practiced at hypothetical thinking and positing counterfactual alternatives as counterpoints to framing power.

Moral character, therefore, appears to have as much to do with what you do when you have the power to frame an action as you would like others to see it, as it does with the common assertion that moral character is what you do when no one else is watching. Framing affects the extent to which others are watching. This, in turn, raises the questions of how are moral issues framed within the classroom and what constitutes evidence of an ethical action. All education is, in some sense, socialization. Take, for example, voluntary product recalls, which are generally interpreted as evidence of ethical behavior within the business ethics classroom. Is this correct? Through a series of counterfactual alternatives, Freeman and Gilbert (1988) and Arce (2004) show that observing a firm voluntarily recalling a product is not evidence that the firm is, in fact, ethical or even socially concerned. Taking an action that favors a certain segment of society (e.g., consumers) does not imply that the segment in question is of primary concern. On the contrary, in the examples provided by Freeman and Gilbert (1988) and Arce (2004), the firm’s primary concern when considering a ‘voluntary’ recall is instead the reaction of the firm’s regulator. Context matters. As a tangible example, the same Johnson & Johnson Company that is celebrated for conducting its 1982 public recall of Tylenol also conducted a 2009 stealth recall of Motrin by hiring contractors to buy the Motrin back at points of sale as if they were customers. The fact that deceptive practices such as a stealth recall or indirect action can be used to manipulate ethical judgments implies a responsibility to make frames explicit in classroom teaching and to question and reflect on the frames employed within the classroom. Students should be encouraged to approach ethical questions from multiple frames of reference and to consciously reflect on their implications, particularly if they result in preference reversals.

One can also think of framing power in terms of the way in which an ethical problem is initially defined or the background assumptions implicit in raising ethical awareness are presented. For example, the AIG case examined here is framed in terms of shareholder rights, consistent with the primacy of shareholder interests in the U.S. version of capitalism. This is another form of socialization that occurs within business education. AIG shareholders lost their equity as part of the AIG bailout, the amount lost was deliberately punitive, and when representatives had their day in court, a federal judge ruled that they had been deprived of their legal rights and expectations, but no damages were ultimately awarded to shareholders. By contrast, Bernanke’s (2015, p. 261) framing of the bailout is well-intentioned but legalistic, emphasizing the Fed’s charter and the question of systemic risk. “Unlike Lehman … AIG appeared to have sufficiently valuable assets … to serve as collateral and to meet the legal requirement that the loan be ‘secured to the satisfaction’ of the lending Reserve Bank.” Continuing (p. 367), “Why had we not insisted that those [AIG’s] counterparties, which included companies like Goldman Sachs, bear some losses? … we had no legal means to force reductions.” Yet, when the totality of the Fed’s actions in the AIG intervention were laid side-by-side, as is the case in joint evaluation, they were judged to constitute a “backdoor bailout” of counterparties (SIGTARP 2009) and an illegal exaction of shareholder equity (Wheeler 2015).

In the end, we have confirmed Prentice’s (2004) concern that a simple reframing of a moral issue can produce a totally different ethical evaluation of the same action. In our example, shareholder rights are violated via indirect action without provoking moral outrage. By contrast, we have shown that joint evaluation, counterfactual construction, and hypothetical reasoning provide a moral counterpoint to indirectly subverting shareholders’ rights. This is a provocative observation as it should be of particular concern to those that espouse a positive “values-free” version of business education based on the primacy of shareholder rights. In the AIG case, shareholder rights were found to have been violated via a reframing of the issue and when an AIG shareholder pointed this out through the legal system there was widespread outrage arguing that the lawsuit was tantamount to claiming that the bailout was insufficiently generous. This change in narrative is another reframing of the issue. From a pedagogical perspective, favoring outcomes involving certain groups or segments of society over others may actually encourage the use of deceptive practices, as was the case in the AIG intervention. Our experiment and its associated results highlight the benefits of business ethics education as providing the tools for seeing through such framing effects.