Ethics and Banking: Do Banks Divest Their Kind?

Abstract

A growing group of institutional investors use divestment strategically to deter misconducts that are harmful for the climate and society. Based on Kantian ethics, we propose that divestment represents investors’ universal and absolute moral commitment to socially responsible investing (SRI). Following categorical and hypothetical imperatives and reciprocity as a norm, we hypothesize how institutional investors’ commit to SRI through a divestment strategy against ethically reprehensible behaviour of banks, especially when these investors represent banks themselves. Using a hand-collected database of the revelation dates of enforcement actions on banks, we find evidence that banks are less likely to divest equity holding on banks with misconduct (fined banks) than their non-bank institutional investors peers. Banks that commit to invest responsibly by signing for the Principles for Responsible Investment (PRI) are not significantly more likely to divest on fined banks stocks than non-signatory banks. Moreover, divestment of fined banks whose own legitimacy to operate is in question is not significantly different from non-fined banks divestment. We find that European banks are more inclined to sell their holdings permanently on fined banks than their United States peers. Therefore, bank’s moral commitment to SRI via divestments is influenced more by cultural and reciprocity norms than their moral commitment to participate in the PRI.

Introduction

Global banks are committing to invest responsibly, with a special focus on climate change and cutting carbon emissions. The Institute for Energy Economics and Financial Analysis (IEEFA) keeps a record of claims made by financial institutions (including banks with at least $10Bn in assets) regarding Oil and Gas¹ and Coal² divestiture. As of December 2021, IEEFA lists pledges from 37 banks regarding Oil and Gas and 90 banks regarding Coal financing. These claims mainly involve winding down of financing for new coal plants or mines, oil sands extraction or, more generally, reducing exposure to fossil fuels and aligning with the Paris Agreement to limit global temperature increase below 2 degrees above pre-industrial levels. Along with individual claims, banks have issued collective commitments such as the United Nations-backed Net-Zero Banking Alliance,³ established in April 2021. Ninety-seven banks, which account for 40% of all banking assets, are part of this alliance, meaning that they commit to aligning their portfolios with net-zero carbon emission objectives by 2050.

Our purpose is to examine banks’ commitment to SRI via divestment of the stockholdings of their peer banks that have been found to violate ethical rules (fined bank peers). Consistent with the literature, we argue that divestment represents the absolute moral commitment to SRI compared to the relative approach of active engagements (Dawkins 2018). Banks from around the world have been under scrutiny since the 2008 Global Financial Crisis due to multiple cases of misconduct. US regulators imposed penalties on both US and non-US banks of $17Bn in 2008, 123 times more than in 2007.⁴ Fig. 1 shows that 2008 was not an outlier; US regulators have imposed penalties of similar aggregate quantities from 2012 to 2015, with enforcement activity reducing since then, at least in terms of amounts paid in fines. Hence, examining banks’ absolute moral commitment to SRI via divestment is important since banks are essential in establishing public trust in the entire financial markets (Berger et al. 2017; Diamond and Dybvig 1983).

Fig. 1

Total Amount paid in regulatory penalties by global banks to US regulators

Misconduct cases involving banks often involve serious ethical matters that adversely affect individuals and society. For example, multiple fines (44) were sanctioned by US agencies on banks for mortgage abuses, which involved misconduct referring to mortgage servicing, foreclosure abuses or racial discrimination,⁵.⁶. The highest penalty for this matter, the $16.65Bn penalty imposed by the US Department of Justice (US DOJ) in 2014 on Bank of America, represents the highest penalty imposed by the US regulatory agency. Regulators have also penalized banks for the misselling of residential mortgage-backed securities (CDOs) that adversely affected investors and the economy and it is considered a “focal feature” (Griffin 2021) of the global financial crisis. Other recent misconduct cases involve rate manipulation, including the well-known LIBOR and FOREX scandals, involving penalties from US, UK and European Union institutions for 11 and 7 banks, respectively. Regulators have also sanctioned banks for anti-money laundering deficiencies and violating economic sanctions, e.g., the US DOJ sanction on BNP Paribas in 2015,⁷ the US DOJ fine on HSBC in 2012⁸ or the Federal Reserve Bank (FED) and Office of Foreign Assets Control penalty on Société Générale in 2018.⁹

Following Kantian ethics, we argue that divestment of bank equity that has been exposed to penalties due to misconduct (fined banks) represents an absolute universal moral commitment to SRI (Dawkins 2018). Based on the norm of reciprocity (Gouldner 1960; Mauss 1967), we also argue that banks as institutional investors are less likely to divest their fined bank peers’ stocks compared to non-bank institutional investors due to the reciprocity norm established through interbank lending and interconnected business relationships among banks (Bianchi et al. 2020; Li and Rowley 2002; Eccles and Crane 1987; Powell 1990). We also examine whether the absolute universal moral commitment to SRI measured by divestment is higher when banks are committed to the United Nations-backed Principles of Responsible Investment (PRI) and when banks’ legitimacy to operate is challenged, i.e., when they received penalties (fined banks). Furthermore, we examine whether differences in cultural norms and stakeholder orientation between European banks and US banks play a significant role in banks’ divestment of their fined bank peers.

We empirically investigate institutional investors’ (banks and non-banks) divestment to the revelation of cases of misconduct sanctioned by US, UK and European Commission authorities from 2005 to 2019. First, we compare the divestment of the equity investment side of banks (Investor Bank) with non-bank institutional investors (IIs) to examine bank divestment behaviour relative to non-bank institutional investors. We focus on the divestment of banks’ stocks since the banking industry is essential in establishing public trust in financial markets and credibility of capital flows in our society. As a result, banks have been under significant regulatory scrutiny due to the multiple cases of misconduct as explained above. We use data from the Violator Tracker, a database of sanctions on corporations imposed by US regulators and agencies along with sanctions resulting from private litigation. We complement this dataset by manually collecting enforcement actions sanctioned by the Financial Conduct Authority in the UK and the European Commission, filtering for fines with penalty amounts of at least $50 M to keep the most relevant cases. Furthermore, we manually collect the “revelation date”, i.e., the date in which it is publicly announced that a bank is under regulatory scrutiny for misconduct. Flore et al. (2021) note that US regulators usually start by filing a civil lawsuit that is eventually settled. Thus, the misconduct is revealed before the settlement. Given that our interest is to examine divestment when misconduct is revealed rather than in the reaction to the announced monetary penalty, we manually collect news reports to find the revelation date for each misconduct announcement. Then, we investigate institutional investors’ trading activities around the revelation date.

We find that institutional investors are more likely to divest fined bank stocks post-revelation dates. However, the equity division of banking groups (Investor Bank) is less likely to permanently divest fined banks than non-bank institutional investors (IIs). We find that Investor Banks are less likely to divest fined banks that have more significant interconnections and are considered systemically important, as measured by banks listed in the Global Systematically Important Banks (G-SIBs). Divestment by banks that have indicated their commitment to invest responsibly by signing the Principles of Responsible Investment (PRI) is not significantly different from divestment of their non-signatory peers. We also do not find differences in divestment between the subset of fined banks and their non-fined bank peers. We find that European banks tend to divest fined banks permanently more than US banks. Our results are robust to matching, changing penalty quantity thresholds, time estimation windows, and controlling for geographic locations of banks and fined banks.

Our study provides several contributions to the streams of literature on ESG investing and SRI through divestment, especially for banks. First, the business ethics literature has indicated the importance of divestment as the absolute and ultimate ethical commitment to uphold inviolable norms and ethical standards (Dawkins 2018; Nath 2021). Researchers have studied the conditions that drive banks to consider ESG issues (e.g., high competitiveness, quality of management (Chih et al. 2010), country-level norms (Hoepner et al. 2016), or different ESG profiles between bank-borrower (Houston and Shan 2022). Our study contributes to this area by theorizing and providing empirical evidence on banks’ absolute moral commitment to SRI through divestment of stocks of their peers who have been fined due to their misconduct cases. Second, we add to the literature on enforcement actions on banks. While previous research has found negative abnormal returns after the announcement of an enforcement action (Jordan et al. 2000; Pereira et al. 2019) and has studied effects on the bond and CDS markets (Flore et al. 2021) and changes in lending behaviour (Roman 2020), we are the first to demonstrate that banks that have indicated their commitments to PRI and fined banks whose legitimacy to operate is in question do not seem to use divestment after an enforcement action on other fined banks compared to non-PRI banks and non-fined banks. Third, the literature has demonstrated that stakeholder salience, specifically legitimacy, plays an essential role in SRI among institutional investors (Gond and Piani 2013; Hebb et al. 2018; Majoch et al. 2017). Our findings indicate that even fined banks whose legitimacy is in question do not significantly divest more on their fined bank peers. Hence, we extend the divestment literature based on Kantian moral theory beyond the stakeholder salience and legitimacy theory. Finally, our work relates to the literature on how the equity asset management business works within banks, which has found that banks use their equity business to support their clients (e.g., Massa and Rehman 2008; Hao and Yan 2012), to the detriment of their investment fund participants’ social objectives. We find that cultural norms towards greater stakeholder orientation play a significant role in banks’ absolute moral commitment to SRI through divestment, i.e., European banks tend to hold a greater moral commitment to SRI through permanent divestment of their peer banks with violations compared to US banks.

The remaining sections of this paper are structured as follows: Section “Background and Hypothesis Development” provides a background and hypotheses development on ESG and SRI tendencies within the banking sector. This section also reviews the literature on the effect of divestment and enforcement actions on banks and behaviours on the equity investment side within banking groups. Section “Data and Methods” describes the database, explaining the details behind the construction of the final set of enforcement actions, along with a description of the empirical models. In Section “Results”, we present the results of the analysis and Section “Conclusion” concludes the paper.

Background and Hypothesis Development

Socially responsible investing (SRI) is represented by two non-mutually exclusive approaches: active stewardship engagement (best-in-class engagement) and divestment. In active stewardship engagement, investors engage in discussions and dialogue with firms’ top managers to influence firms’ behaviours (actions) that improve firms’ ESG performance. Divestment represents investors’ decision to exclude investing or selloff their ownership in firms with ESG issues. PRI indicates that divestment represents the final step in an escalation strategy of stewardship engagement.¹⁰ Some SRI funds use both divestment for ESG issues that they deem unacceptable, and engagement for other less critical ESG issues.

Divestment is inherently a controversial topic from an ethical point of view. Responsible investors that want to influence corporate behaviour have to consider divestment (negative screening or exclusion) and engagement, which both entail ethical components (Braungardt et al. 2019; Gosiger 1986; Kaempfer et al. 1987). A counter-argument to divestment from the moral standpoint, also discussed in Braungardt et al. (2019), is that the reliance of other businesses on fossil fuels may not be relevant in a bank divestment setting. It is argued that there is an inconsistency in considering fossil fuel production as an isolated issue while not considering the potential reinvestment into other sectors, such as banking (Ritchie and Dowlatabadi 2014) or sectors with heavy consumption of fossil fuels (Moss 2017). In our study this inconsistency does not arise because regulatory enforcement mainly highlights issues only concerning a bank itself (or the banking industry) and not external organizations.

According to Nath (2021), divestment has been a strategy used by Responsible Investors since the US civil rights protests in the 1960s, when responsible investment was not yet institutionalized. The literature on the topic suggests that equity divestment attempts to influence firms’ behaviour towards better ESG practices. For example, Rohleder et al. (2022) find that polluting firms reduce emissions when there is selling pressure.

Dawkins (2018) argues that engagement and divestment approaches have strengths and weaknesses. However, he emphasizes that without divestment, SRI through stewardship (best-in-class) engagement is ethically and strategically fraught because certain ESG violations (e.g., misleading investors to invest in risky assets, discrimination, etc.) are universally considered morally wrong based on the categorical imperative from the lens of Kantian ethics. He explains that best-in-class engagement strategy without a divestment is ethically timid by rewarding unethical behaviour that is relatively “best” compared to the worst unethical behaviour committed by others. Dawkins (2018) illustrates that the Norway Government Pension fund adopts explicit absolute inviolable standards and imperative guidelines that list activities and ESG violations will lead to exclusion and divestment decisions from the fund.

ESG in Banking

The literature shows that banks have reasons to incorporate ESG considerations into their decision-making. Chih et al. (2010) explore why the financial industry behaves in a socially responsible way. They find that, among other reasons, external pressures like the level of competitiveness that the firm is facing, a higher country-level legal enforcement and quality of management at the country level are determinants of better CSR practices. In terms of lending deals, Hoepner et al. (2016) find that banks are willing to provide financing at lower costs to firms located in countries with strong social and, especially, environmental institutional frameworks. Based on industry composition of bank loans and investment portfolios, Bernardelli et al. (2022) find that the impact of bank loans on banks’ ESG scores is a potential instrument for limiting banks’ exposure to the fossil fuel sector. Houston and Shan (2022) find that banks tend to provide funding to firms with similar ESG performance, and positively influence ESG practices of those borrowers with lower scores than the bank. This influence is not only determined by financial reasons, but also by reputational concerns, where banks are apprehensive about human rights abuses, social discrimination, and climate change. Kacperczyk and Peydró (2021) show that banks that commit to the Science Based Targets Initiative change their lending practices and that changes are also observed in the borrower firms. They find that these banks shift lending from firms with very high emissions¹¹ to firms with lower emissions. More importantly, they find that banks’ preferences towards greener firms explain the reduction (increase) in lending for brown (green) firms.

Financial reasons also influence banks’ behaviour towards ESG issues. Wu and Shen (2013) and Shen et al. (2016) have found a link between the financial performance of banks with high CSR scores. Similar results were found by Belasri et al. (2020), who measured bank efficiency as a function of bank inputs and outputs and found that efficiency increased as CSR scores improved. However, this only applied to banks in developed countries and countries with high levels of investor protection that are tilted towards stakeholders.

Banks are under scrutiny for financing fossil fuels despite public pressure and their pledges. Banks are still raising funds for fossil fuels projects, with the 60 largest banks in the world still securing similar amounts of capital as in 2016 for firms involved in fossil fuel activities.¹² Urban and Wójcik (2019) show that 124 of the 153 firms that were excluded by the Government Pension Fund Global of Norway received underwriting services from investment banks. However, just five banks (of almost 400 in their sample) generated 40% of the revenue from these underwriting services. Public pressure effectively reduces bank lending to fossil fuels activities, but with the side effect of increasing fossil fuel lending in other countries with less scrutiny (Cojoianu et al. 2021). Therefore, if banks do not shy away from fossil fuels despite the social pressure, they would be even more hesitant to divest stocks of other banks that engage in misconduct, given the lack of public pressure to do so.

The literature on banking misconduct shows that institutional investors and banks may also be concerned about financial and reputational losses due to misconduct and potential enforcement actions (EAs). Jordan et al. (2000) and Pereira et al. (2019) find that the announcement of regulatory investigations that eventually lead to a regulatory fine is associated with post-announcement negative returns. Flore et al. (2021) examine the market reaction to settlement announcement, finding the opposite result, a slightly positive return on the date on which the resolution occurs and improved conditions in the bond and CDS markets. They argue that these results are driven by how US regulators impose penalties. Misconduct is usually revealed before the announcement of the outcome (settlement) of the enforcement action. Therefore the positive reactions are likely explained by eliminating the uncertainty relative to the enforcement process.¹³ EAs can impact banks’ activities and decision-making. For example, Roman (2020) shows that penalized banks offset the adverse reputational effects of the EAs by offering better borrowing conditions to large businesses while tightening them for SMEs.

Regarding the effects of divestment, Heinkel et al. (2001) theorize lower stock prices and an increased cost of capital for polluting firms, providing the firm an incentive to reform if the cost of reforming is lower than the cost of capital associated with the divestment. Dawkins (2018) argues that the threat of divestment is necessary for engagement, as the investor will gain negotiation power if there is a credible threat of withdrawing, implying that the “two approaches are symbiotic rather than exclusive”. Turning to empirical studies, little effect was found for the divestment movement that targeted South Africa (Gosiger 1986; Kaempfer et al. 1987). Divestment or exclusion does not necessarily negatively impact portfolio performance, as shown by Hoepner and Schopohl (2018), Trinks et al. (2018), and Plantinga and Scholtens (2021). Hunt and Weber (2019) find that divestment of carbon intensive stocks results in investment strategies with higher risk-adjusted returns and lower carbon intensity. Rohleder et al. (2022) find that firms with high carbon emissions that are under “decarbonization selling pressure” face lower stock market returns than other firms with high carbon emissions that are not under selling pressure. These decarbonization selling pressures are influenced by investor herding behaviour, as shown by Benz et al. (2020). Similarly, Dordi and Weber (2019) look at the effect of divestment announcements on 200 oil, gas and coal companies, finding short- and long-term negative abnormal returns for the divested firms.

Banking Relationships

Banking groups’ equity investment and asset management side are influenced by other businesses within the group. Massa and Rehman (2008) find that mutual funds offered by the asset management side of commercial banks overweight their lending client’s stock. Ferreira et al. (2018) further show that overweighting clients’ shares implies lower performance than other mutual funds not managed by banks. Hao and Yan (2012) and Bodnaruk et al. (2009) find that investment banks overweight the shares of their IPO and M &A clients, respectively, although the former find negative abnormal returns, while the latter points to positive abnormal returns. These results suggest that banks use their asset management business to support their clients and build business relationships with them. Golez and Marin (2015) find that bank-affiliated equity funds invest more in parent banks around seasoned equity offerings and during stock price drops due to bad news. Literature has also documented that interbank lending markets allow banks to maintain liquidity to withstand idiosyncratic negative shocks (Freixas et al. 2000; Alfonso et al. 2013; Craig et al. 2015; Chiu et al. 2020). Therefore, banks could overweight or not to divest other banks’ stock to maintain good relationships that potentially bring payoffs in other markets. Another example is interbank lending, where maintaining relationships with lenders is important for banks to gain access to cheaper liquidity (Cocco et al. 2009; Temizsoy et al. 2015; Bräuning and Fecht 2017). In contrast, divestment tends to be more pronounced among non-banks institutional investors, especially large public pension funds. Egli et al. (2022) find that divestment is an approach more likely used by large public pension funds.

Hypothesis Development

We develop our divestment hypotheses based on Immanuel Kant’s Groundwork for the Metaphysics of Morals or Grundlegung zur Metaphysik der Sitten (Kant 1785). We recognize the ongoing debate among business ethicists and moral philosophers on whether corporations can be considered moral agents (corporate personhood) and, therefore, be subject to Kantian ethics (e.g., Altman 2007; Barry et al. 2019; Bowie 2017; Mansell et al. 2019). Following studies that apply Kantian ethics to firm-level actions as a group of maxims and collective wills of corporate leaders (Van de Vijver 2022; Lenz 2020; MacArthur 2019; Robinson and Shah 2019), we argue that institutional investors can be considered as collective moral agents as shown in misconduct studies (e.g., Lilly et al. (2021); Stevens (2013)). Hence, institutional investors are considered collective moral agents responsible for making investment decisions at the institutional level and are subject to Kantian ethics.

Kant’s moral view is considered as the absolute deontological approach that focuses on the universal law of ethical decisions and is based on the pure ethical intention of action itself, regardless of its results or consequences. Kant argued that everything in nature works according to laws, and a rational being has the power to act according to the conception of universal laws and therefore, one has a will (Kant (1785), [401]). The will is a faculty of choosing based on a reason, but the reason itself does not sufficiently determine the will if the will submits to subjective conditions (incentives). Following Kant’s universal law of moral ethics, we argue that institutional investors’ decision to divest stocks that violate inviolable moral ethics represents the absolute universal law of moral and ethical values to SRI (Dawkins 2018).

According to Kant, the doctrine of morals or ethics (what ought to be) based on an objective principle of reason necessitates the will as a command of reason and the formula of the command is called an imperative (Kant (1785), [413]). All imperatives are expressed by ought but could be driven by internal motivation (subjective conditions). Since imperatives could be driven by objective and subjective conditions, Kant classified imperatives as categorical (kategorien) and hypothetical (hypothetisch) imperatives, as individuals’ moral actions are expected under rational and autonomous circumstances. Kant’s categorical imperative implies that we are not acting morally out of a sense of duty but because of our own moral maxims as the universal law while treating people as a means to an end. Doing the right thing (righteousness) is morally and ethically the right thing to do within itself and is not driven or motivated by anything else (Kant (1785), [413–414]). In contrast, Kant’s hypothetical imperative implies that we are motivated by the practical necessity of a possible action that is considered ethical as a means for attaining something that is desirable or expected by others (Kant (1785), [415]). The hypothetical imperative indicates that the action is only good for some possible or actual subjective conditions (Ellington 1993).

Since divestment is considered the absolute approach of SRI (Dawkins 2018), based on Kant’s categorical imperative, we argue that if divestment represents the absolute universal law of socially responsible investing, then divestment should only be driven by the objective law of reasons because violations of ethics are deemed as breaches of inviolable moral and ethical standards. In this case, we expect investors tendency to divest to increase when violations occur regardless of the outcomes or consequences to themselves from divestment because violations against inviolable moral and ethical norms are not morally acceptable and cannot be tolerated. However, divestment can also be influenced by hypothetical imperatives, where institutional investors’ divestment decision is expected to be influenced by the perceived subjective consequences to themselves.

Divestment has been proven to have a significant negative effect on equity prices, as discussed in section “ESG in Banking”. The literature on enforcement actions involving ethical violations has found negative stock market returns for firms targeted by enforcement actions (EAs) once the misconduct is known, as indicated in section “ESG in Banking”. Furthermore, EAs for banking misconduct often involve matters regarded as inviolable from a moral and ethical standpoint, such as misinformation about the risk of securities sales, anti-money laundering deficiencies, racial discrimination, rate manipulation, etc., as indicated in the introduction. Banks incorporate, to some extent, ESG policies due to social pressures, interrelationships among banks and financial reasons. Banks’ investment exposure may affect their ESG ratings (Bernardelli et al. 2022). However, as indicated in section “ESG in Banking”, social pressures may not be strong enough to stop banks from investing in other banks that have been involved in misconduct (fined banks).

We further explain that banks’ divestment behaviour is influenced by reciprocity as a norm in the banking industry. Mauss (1967) developed the theory of reciprocity based on society’s the gift-giving norm. He argued that giving and receiving gifts represents a moral obligation (norm) where those who do not comply are punished. Gouldner (1960) suggests that the norm of reciprocity within a group may become a universally accepted principle. He argued that people should help and not injure those who have helped them. Cropanzano and Mitchell (2005) postulate that one is obliged to reciprocate if one receives something from another person. Section “Hypothesis Development” reviews how banks use their asset management side to build business relationships. Eccles and Crane (1987) show evidence that reciprocity through relationships, internal competition, and compensation based on services provided is the norm in investment banking. Li and Rowley (2002) find evidence of a reciprocity relationship among the co-leads of initial public offerings (IPOs) syndication in US investment banks. Bianchi et al. (2020) find that the European interbank market demonstrates reciprocal relationships that are critical to build longitudinal financial networks. Hence, we argue that banks’ use of divestment as a tool for SRI is subject to the norm of reciprocity. Banks face potentially harmful consequences if they do not follow the reciprocity norm in their industry. Due to these subjective conditions and considerations, we expect banks may be less likely to engage in the categorical imperative of SRI, as measured by divestment. Therefore, banks may be less likely to divest stocks of their own fined bank peers relative to non-bank institutional investors. We state our first hypothesis (H1) as the following:

Hypothesis 1

(H1) Banks are less likely to divest stocks of fined banks than non-bank institutional investors.

We further examine whether banks that have publicly declared their commitments to SRI are more likely to engage their categorical imperative and are therefore more likely to divest of fined banks than their peers that have not declared their SRI commitments publicly. We measure banks’ strong public commitments to SRI by examining banks’ pledges to invest responsibly through their participation in the UN-backed Principles for Responsible Investment (PRI). PRI is a network of investors launched in 2006 that pledge to consider ESG issues in their decision-making process. The 2021 PRI annual report¹⁴ showed that, as of March 2021, over 3800 investors had signed, covering $121Tr of assets under management. These investors commit to observe the six PRI principles,¹⁵ which involve considering ESG issues when making investment decisions and engaging with their portfolio firms, reporting ESG implementation and disclosing problems found in portfolio firms, improving the implementation of the principles and promoting the PRI network within the investment community. The network may exclude signatories for failing to meet the requirements; PRI reported in 2021 that five signatories were delisted for not meeting the minimum requirements, while another ten were excluded for failing to report.

Recent empirical research reports mixed results regarding whether PRI investors better embrace ESG practices than other investors and whether institutions improve their sustainability scores after becoming PRI signatories. Gibson Brandon et al. (2022) find that PRI signatory institutions have significantly higher ESG scores than other investors, although only in the Social and Governance dimensions, and improving scores after signing. US PRI signatories are the exception, as their scores are lower than non-signatories from outside the US. Kim and Yoon (2022) find similar results for US mutual funds; they do not hold better ESG scores before signing and do not improve after signing. However, mutual funds receive more positive cash flows after joining, suggesting a marketing incentive to sign up for the PRI that does not necessarily embrace the PRI principles. Similarly, in the analysis of hedge funds by Liang et al. (2022), it is found that PRI signatories perform worse than non-signatory hedge funds. However, signatories receive higher cash flows and charge higher fees. The performance gap is explained by signatories with agency problems (misalignment of manager-investor objectives) that have very low ESG scores, while signatories with high ESG scores perform similarly to those that did not sign.

We posit that if a banks’ commitment to invest responsibly by becoming a Principles for Responsible Investment (PRI) signatory aligns with their universal moral and ethical commitment to SRI, then we expect bank PRI signatories to follow Kant’s categorical imperative. Thus, these PRI banks are more likely to divest fined banks than those not participating in PRI. However, if a banks’ intention to participate in PRI is driven by certain subjective conditions (i.e., financial reasons, reputation, responding to social pressure, etc.), then their intention to participate in PRI is driven by Kant’s hypothetical imperative. Therefore, their divestment may not be significantly different from those that do not participate in PRI because their motivations to participate in PRI do not reflect their absolute universal law to SRI. Our second hypothesis (H2) is stated as:

Hypothesis 2

(H2) If the motivation of banks to participate in PRI is driven by the categorical imperative, then banks that participate in PRI are more likely to divest fined banks than non-PRI bank peers.

To the best of our knowledge, none of the existing studies has examined SRI behaviour, in terms of divestment, for the subset of banks that have been found to engage in misconduct and have received penalties due to their misconduct (fined banks). Sullivan (1989) argues that Kant’s moral theory depends on the assumption that moral reasons are also driven by the conative view (perception that motivates us to act). In this case, the conative view explains what motivates banks to divest. Fined banks may perceive that their misconduct is not morally reprehensible, hence they may not perceive similar misconduct committed by other banks to be morally reprehensible. Hence, fined banks may not be motivated to divest equity of other fined banks based on their subjective perceived moral views. Furthermore, the individuals within the bank involved in the misconduct case may not be the same as those making equity investment decisions. Therefore, divestment by fined banks of other fined banks’ equity may be no different from divestment by non-fined banks.

Extant literature generally examines divestment from the lens of stakeholder theory and stakeholder salience (Mitchell et al. 1997). Based on the stakeholder salience and collective action theory, Gond and Piani (2013) argue that institutional investors can use their increased power, urgency, and legitimacy, along with collective actions facilitated by enabling organizations such as PRI, to increase the effectiveness of deliberate negotiation over the legitimacy of ESG issues with corporate managers. Majoch et al. (2017) indicate that pragmatic legitimacy, normative power, management values, and coalition building contribute most to the effectiveness of PRI as a stakeholder in SRI. Based on Davis (1973) Iron Law of Responsibility, Wood (1991) argues the principle of legitimacy as the society grants legitimacy and power to businesses, and those who do not use power in a manner society considers responsible will tend to lose their social licence to operate. Gifford (2010) and Hebb et al. (2018) defend legitimacy as the most crucial aspect of shareholders’ salience. Since we examine institutional investors’ divestment of equity on fined banks, consistent with these studies, we argue that banks whose own legitimacy to operate is in question, as measured by violation penalties (i.e., fined banks), are more likely to divest other fined banks to avoid losing their legitimacy and licence to operate.

Based on legitimacy theory, we argue that banks involved in some form of misconduct or violations (fined banks) are more likely to penalize via divestment their own fined bank peers that have also been found to have cases of misconduct than their peers that are not found to have violations or misconducts (non-fined banks) to keep their own legitimacy to operate. Our third hypothesis (H3) is stated as:

Hypothesis 3

(H3) To maintain their own legitimacy to operate, fined banks that are also part of the enforcement actions sample are more likely to divest misconducting banks than non-fined banks.

Another important stream of ESG literature relevant to this study is the legal environment and the impact normative and cross-country cultural differences in banks as institutional investors have on attitudes to ESG issues. North (1990, 1991) proposes that institutional settings are formed by both formal and informal constraints. The informal constraints are shaped based on norms and mental models of individuals who may have different cultural heritage, religious or political beliefs, or reside in different geographic areas. In interpreting Kantian ethics, Wood (2002) describes individuals as free agents who act under normative principles with a capacity to choose between alternatives according to one’s judgement about which alternative is permitted or required by a norm. In other words, the absolute universal law of morality and ethics is also bound by the norm. Hence, we expect that the universal absolute moral and ethical commitment to SRI in terms of divestment is influenced by informal constraints, i.e., local norms and mental models of individuals in certain geographic areas.

In a cross-country empirical study of banks’ ESG performance, Wu and Shen (2013) classify the 162 banks in their sample into four categories by their CSR scores, finding that 71% of their observations in the group with higher scores were from banks in Western Europe. One explanatory factor of this difference may be the tendency of financial industry firms located in countries with German or French legal systems to incorporate better ESG practices (Chih et al. 2010). Looking at a worldwide sample of firms from any sector, Liang and Renneboog (2017) found a similar result; firms from civil law countries, that have a “stakeholder view”, exhibit better CSR scores than firms from common law countries. Döring et al. (2023) find that foreign institutional investors with civil law origin, which have a higher preference for environmental responsibility, improve the scope and quality of GHG emissions reporting. Zheng et al. (2013) find that national culture and norms affect banks’ ethical behaviour. Moufty et al. (2022) find that European banks provide more sustainability disclosure than US banks due to the normative pressure of stakeholder-oriented, as opposed to shareholder-centric, models.

Regarding empirical studies on institutional investors’ commitment to ESG and SRI across countries, Amel-Zadeh and Serafeim (2018) conducted a survey in which European investors were more likely to respond that incorporating ESG criteria is an “ethical responsibility”, while 22% of US investors believed that ESG information is not useful and that incorporating it would go against their fiduciary duties, as opposed to 4% among European respondents. Hoepner et al. (2021) argue that institutional investor’s attitudes to SRI are more influenced by the broader social and cultural environment in which they operate (Scott and Christensen 1995). Following Scott (1995) institutional theory of regulative, normative, and cultural-cognitive framework, they find that normative and cultural pillars have stronger impacts than regulatory pillars on the likelihood of asset owners signing up for the Principles of Responsible Investment (PRI). Egli et al. (2022) indicate that banks as institutional investors may need to adjust their investment strategies beyond what is legally required to maintain their social licence to operate because of their funding reliance on deposits. They argue that divestment is more likely for institutional investors based in countries with stronger societal preferences.

Given these existing theories and empirical evidence, we argue that European banks tend to be bound deeply by the normative culture towards a greater commitment to SRI than US banks because European culture and norms focus more on broarder stakeholders than the US culture and norms. Hence, we expect that European banks tend to follow the categorical imperative and therefore, they are more likely to divest the stocks of their fined bank peers than US banks. Thus, our fourth hypothesis (H4) is stated as the following:

Hypothesis 4

(H4) European banks are more likely to follow the categorical imperative and therefore, they are more likely to divest fined banks than United States banks.

Data and Methods

Data

The main part of our dataset is equity ownership data retrieved from FactSet Ownership database. Our sample period runs from 2005 to 2019 and it contains all the holdings of institutional investors (banks and non-bank institutional investors) that hold any bank equity in their portfolio holdings at any point in time. FactSet ownership is composed of three groups of sources; 13F filings, “stakes-based sources”, e.g., the UK Share Register or 13D filings and “sum of funds”, i.e., data collected at the fund level that is aggregated to the entity level. Following Massa and Rehman (2008), we keep all institutional investors except for those with Asset Under Management below $1 M and those that are less than one-year old to ensure the sample is representative of institutional investors who can charge fees to their clients according to the SEC Rule 205 (Investment Advisers Act of 1940).

Enforcement Actions Data

Our main source of enforcement data is the Violation Tracker search engine, which is offered by the Non-Profit Organization Good Jobs First. This dataset¹⁶ contains sanctions applied by US federal agencies, the Department of Justice and state attorney generals. We complement this data by collecting the fines imposed by the Financial Conduct Authority (FCA) in the UK.

From the Violator Tracker, we collect penalties imposed on financial institutions beginning in 2005, with a total of 616 penalties above $10 M on both listed and unlisted firms. We drop the unlisted firms and the non-bank financial institutions, filtering the list of enforcement actions (EAs) down to 467. Then the FCA fines that meet the same conditions are incorporated into the sample, raising the number of observations to approximately 500. Next, we group fines imposed on the same bank on a given day, as in some occasions, different regulators perform enforcement actions for the same reason. Then fines are filtered to keep those where the penalty was of a significant amount, using a threshold of $50 M. The final list compromises 209 enforcements.

Misconduct events are generally known before the EA, especially in the US. Flore et al. (2021) note that US regulators like the SEC file lawsuits with misconducting entities that typically are settled, while for other regulators, including the UK’s FCA, the first public announcement also includes the regulatory fine.

Our analysis focuses on the ethical considerations of investors regarding banks misconduct. Therefore, we replace the enforcement date with the revelation date, i.e., the time at which the misconduct became public knowledge. We collected news articles from the Wall Street Journal, Financial Times and the BBC, using Factiva, to identify the revelation dates. These were found for 82% of the fines, with an average time difference between the revelation and penalty date of 622 days. There were a number of cases where different monetary penalties for the same bank had the same revelation date, reducing the number of fines to 183.

We investigate divestment behaviour on 736 banks (Investor Bank) and 5,163 non-bank institutional investors (IIs) to test for H1. Then, we examine the divestment of subsamples of 126 fined banks as institutional investors versus 610 non-fined banks as institutional investors to test for H3 and 372 European banks and 234 US banks to test for H4. The subsamples of PRI and non-PRI signatory banks are described in the next section.

Table 1 shows the results of a standard event study on the market returns of the sanctioned banks. Betas are calculated for each bank using 1-year rolling windows, excluding the month prior to the fine. Then we obtain the Cumulative Abnormal Returns (CARs) at different windows around the time of the fine or the revelation date. While, on the penalty date, no negative returns are found, the announcement of a possible future fine implies a drop in banks share prices. The result is robust to raising the fine quantity threshold, with returns becoming more negative as the threshold increases. Flore et al. (2021) argue that pinpointing the exact revelation date is often difficult. However, our results on revelation date stock returns show a significant reaction in capital markets. Thus, our revelation dates are strongly associated with dates in which the knowledge of potential future regulatory sanctions became public. Furthermore, given the quarterly frequency of our ownership data, finding the exact date is not extremely important.

Table 1 Event study results on banks CARs around penalty and revelation dates

PRI Signatories Match

The Principles for Responsible Investment (PRI) provides an open access list of the institutions that have signed the principles.¹⁷ As of May 2020, 3109 entities had signed up to PRI. To integrate the signature data with FactSet Ownership data, we proceed with a name-matching methodology similar to that used by Gibson Brandon et al. (2022), as PRI does not provide entity identifiers. The name match is composed of three elements, an exact match, two rounds of matching using Jaro-Winkler distance and a manual check.

The PRI list is matched to FactSet using the 63,000 entities contained in the FactSet ownership database. A few pre-processing steps are conducted on the names in these two lists. All the names are converted to lowercase, punctuation marks are removed, and special characters are normalized to Unicode NFKD e.g., “N” is replaced by “N”, “Ç” by “C” and so on.

Then, the Jaro-Winkler distance (JWD) is calculated using the PRI list and the FactSet ownership list in the first iteration. The closest match for each signatory is found and the results are then divided into three: exact matches (controlling by country), matches where JWD is different from zero, but the country is the same, and matches in which the country is different. The first two lists are manually checked to ensure that they are right and, if not, manual searches are done using the FactSet database to find the right match. The third list compromises approximately 1200 signatories and it is assumed that these matches are wrong, as the matches are not robust to controlling by country. So, the JWD is calculated a second time, but instead of searching for matches in the ownership database entities list, we use the whole list of FactSet entities, which contains around 8 million entities. The closest match for each signatory is selected and all matches are again manually checked and, in case it is incorrect, the FactSet database is used again to find an appropriate match. Finally, out of the 3109 signatories, a match is found for 2583 of them.

The final step is to consider the signatory status across the corporate hierarchy of each signatory, where we classify as signatories all the subsidiary entities of the actual signatories. Our final sample consists of 152 banks as PRI signatories and 584 banks as non-PRI signatories.

Control Variables

Investment decisions are determined by a series of factors relating both to the investor and to the universe of firms to invest in. Therefore, we include both bank and investment controls. We collect bank-level data from Datastream and Fitch Connect, obtaining returns from the former and balance sheet and P &L data from the latter. Investors may be reacting to negative financial market performance of the banks in their portfolios, therefore we control for lagged bank returns and, as a measure of risk, for the lagged second-order Lower Partial Moment of the bank return. Our equity holdings data are measured at the entity level; hence, we cannot observe whether a portfolio holding is the result of an active or passive strategy. To account for this issue, we control for the Market Value of the portfolio bank, as passive strategies allocate higher weights in larger firms. Investors may consider a bank’s financial situation when assessing investment decisions, thus, we control for a set of bank-level variables that measure size (total assets), liquidity (Net Loans/Total Assets), capital adequacy (TCE ratio) and profitability (Net Interest Margin and Return Over Average Equity). The last control variable regarding bank-level data is the Book-to-market ratio, which is included to control for investor preferences towards value or growth stocks.

Regarding investors, we control for the size of their portfolios, measured in Asset Under Management and the number of individual stocks held. We also control for their past performance and the average market value of their portfolio holdings. We account for investment horizon preferences by controlling for portfolio turnover, calculated as the 4-quarter moving average Churn Ratio consistent with Gaspar et al. (2005). Finally, we add a “Home Bias” dummy variable that equals 1 if both the fined bank and the institutional investor are based in the same country (same state or county in our untabulated robustness test).

Methods

Difference-in-Differences (Diff-in-Diff) models are estimated around the time of the fines to identify the impact of the misconduct revelation events on institutional investors' holdings of banks. However, setting a Diff-in-Diff window for every event would lead to overlapping windows and possible confounding effects between events. Therefore, we further filter the list of revelation dates by keeping the first fine by investors and topic. For example, Citigroup received a penalty from the SEC in 2010 for giving investors misleading information about mortgage CDOs. For this same issue, Citigroup also settled with the National Credit Union in 2011, Freddie Mac & Fannie Mae in 2013 and the Department of Justice in 2014. However, the misconduct was known at the revelation date of the SEC enforcement, so we keep SEC enforcement and ignore the subsequent fines. The last step regarding enforcement events is to aggregate them by bank and quarter, reducing the list of events to 82.

Even applying this further filter, there can be overlap between Diff-in-Diff (DiD) windows. To address this overlap, we classify the events in three groups given their time difference to the previous event and create a window accordingly. The list below describes how we obtain the estimation windows if using four-quarter periods:

• 1. Events at least one year after the end of the post-window of the previous event. No overlap is found here, so a new estimation window is created.

• 2. Events that occur in the post-window of another event. Even though there is overlap, this happens on the post-window of the previous event, so, we extend the window until the end of the 4-quarter window of the second event, setting a maximum window length of 8 quarters.

• 3. Events occurring between 1 and 2 years of the end of the post-window of the previous event. There is no time for a second window because it would create overlap between the previous post- and current pre-periods. Furthermore, the post-window cannot be extended as before, as there would be a “gap” between the end of the previous post-window and the new one. These events do not get a DiD window, but we still consider them, following the same rules, for the following events.

The same logic is used to generate estimation windows of a different length. Figure 2 contains a timeline of estimation windows for a given bank, where the orange bars identify estimation windows. Low bars indicate pre-period and high bars post-period. This bank had five events starting in 2011 Q1. As this is the first event, it gets a Diff-in-Diff window. The second event is an example of event type 2, as it just came one quarter later than event one. So, the post-window for Event 1 is “reset” and carried until covering 4 quarters after Event 2. Event 3 is an example of event type 3. An estimation window for it would create overlap while extending the previous post-window is not an option as it ended 2 quarters previously to Event 3. Thus, Event 3 is not included in the estimation. Event 4 is discarded, as Event 3 is considered when creating the windows even though it was discarded. Finally, Event 5 is suitable for estimation. Figure 2 contains an example of estimation windows generated for one bank. There are 5 revelation dates left after filtering for this bank; however, only two estimation windows and 3 cases of misconduct were included, while the other 2 were discarded to avoid overlapping estimation windows.

Fig. 2

Example DiD estimation windows for one bank

Once the estimation windows are defined, the following Diff-in-Diff models are estimated:

$$\begin{aligned} \begin{aligned} Y_{j,i,q}&= \beta _{1}InvestorBank_{i}\cdot Post_{j,i,q} \\&\quad + \beta _2 InvestorBank_{i} + \beta _3Post_{j,i,q} \\&\quad + \beta _{4}^{'}BankControls_{i,q} + \beta _{5}^{'}InvestorControls_{j,q} \\&\quad + \epsilon _{j,i,q} \end{aligned} \end{aligned}$$

(1)

where i indexes fined banks, j investors and q quarters. $\textit{Investor Bank}$ is a dummy that takes a value of one if the institutional investor is part of a bank (equity investment division of a bank) or zero for non-bank institutional investors (IIs). $\textit{Investor Bank}$ can be replaced with $\textit{PRI Investor Bank}$ that takes a value of one for PRI signatory banks for investor j at quarter q or zero for non-PRI banks. $\textit{Investor Bank}$ can be replaced with $\textit{Fined Investor Bank}$, a dummy variable with a value of one for banks that have been found to engage in misconduct or zero otherwise. $\textit{Investor Bank}$ can be replaced with $\textit{EU Investor Bank}$, a dummy variable equal to one if banks are European banks or zero for US banks. Post is a dummy variable with a value equal to one during the period after the misconduct is revealed or zero for pre-revelation dates. $Y_{j,i,q}$ is measured at the investor-firm-quarter level and it is either one of the following:

• $\textrm{Selloff}_{\textrm{j,i,q}}$. A dummy equal to 0 if investor j holds a position in bank i at time q and equal to 1 if the investor had a holding in bank i but has since sold their holding in bank i.

• $\textrm{PermSelloff}_{\textrm{j,i,q}}$. Same as $\textrm{Selloff}_{\textrm{j,i,q}}$ but only turns 1 if investor j never buys again (in our sample) stock of bank i.

We estimate these models using data from investors that had activity during the complete estimation window and that had a holding on the fined banks at the start of the window. Thus, we do not consider investors that made their initial investment in the bank during the estimation period. See Appendix A (Table 13) for variables description.

Matching Investors

Table 2 presents summary statistics by bank affiliation at the start of each one-year estimation window on the unmatched and matched samples. Both groups are well balanced in terms of their portfolio size, with similar asset under management (AUM) and the number of distinct shares held. Furthermore, the average market value of their portfolio stocks is similar, suggesting no difference in preferences for large or small caps. Banks have portfolios that tend to be more constant, as measured by the portfolio turnover. The highest differences are found in the location and type of entity variables, with institutions in the bank group mainly located in Europe, while the non-bank group is mainly composed of US investors. Furthermore, only 1% of bank-affiliated investors are asset owners, while 5% of the non-bank sample are, while PRI affiliation is higher within banks, as 18% are affiliated to PRI while only 8% of non-banking institutions are.

Table 2 Covariate balance pre- and post-matching

We account for these differences by matching exactly investors by country, asset owner dummy and PRI affiliation. We search for matches at the bank-investor level, i.e., among investors that also hold the same bank. Matching is timed at the start of the estimation window, e.g., if the estimation window length is 1 year for each period and misconduct is revealed in quarter q, investors are matched at quarter $q - 4$. We then follow with nearest neighbour Propensity Score Matching in two versions, generating 1-to-3 pairs. The matching variables are log-transformed AUM, the number of distinct shares held, the log-transformed average market value of portfolio constituents and 4-quarter rolling average portfolio turnover.

Matching results are displayed in Table 2, showing a more balanced dataset, especially regarding the categorical variables used in the exact matching.

Results

Summary Statistics

We start the discussion of the results by looking at descriptive statistics, presenting summary statistics in Table 3 and a correlation matrix in Table 4.¹⁸ The average position held by investors in each bank is $41 M and represents 1.3% of the average investor’s portfolio. Both measures have a heavy right tail in their distributions, as in both cases the mean is above the 75% percentile. This suggests that a few investors may have very significant positions while many others hold a portion of stock that is relatively low. However, the value of the holding highly correlates with Assets Under Management; thus, the heavy right tail may be explained by the differences in AUM between investors.

Table 3 Summary statistics

Table 4 Correlation matrix

Institutional investors' part of banking groups represents 19% of the positions observed in our sample, hence cross-ownership between banks is common. Correlations indicate that their positions are not higher than other investors in both absolute (holding value) or relative terms (portfolio weight); therefore, positions held by banks are not very different from their non-banking counterparts. The main difference observed in the descriptive analysis is the higher preference for foreign banks of Investor Bank.

Turning to our dependent variables, in 17% of our investor bank-level observations, the investors have sold the position, while this has been done permanently in 11% of the cases. The probability of selling or permanently selling at every given quarter is 3.7% and 2%, respectively. Thus, we find evidence of divestment on fined banks stock. These variables negatively correlate with the bank-affiliated investor dummy, indicating a lower probability of divesting for banks, although this correlation is low and needs to be empirically tested. Regarding correlations with control variables, the highest correlations are the positive correlation with lagged downside risk and turnover and the negative correlation with the number of individual shares that the investor holds.

Divestment Models

Figures 3 and 4 show the trend in the cumulative selloff and permanent selloff variables using one-year windows. In both cases, bank-affiliated institutions were less likely to divest than their peers at every point in time with a widening gap. One-quarter before the enforcement 9.3% of non-bank investors (IIs) and 8% of Investor Bank had already permanently divested, while 27.2% of IIs and 22.4% of banking investors permanently sold their holdings during the complete estimation window.

Fig. 3

Selloff trends pre- and post-enforcement in one-year time windows

Fig. 4

Permanent Selloff trends pre- and post-enforcement in one-year time windows

Table 5 reports the results from the model described in Eq. (1) using four-quarter estimation windows, allowing us to test the hypotheses that Investor Banks are as likely as other IIs to react negatively to their peers’ misconduct. The first two columns show model results without control variables or fixed effects, allowing us to check the percentage of investors that left or permanently left the bank by group and time. The coefficients in column (2) represent the average probability of permanent selloff at any given quarter before and after the fine revelation. The constant indicates that, on average, 6.4% of IIs permanently divested at any given quarter during the year prior to the fine. The post-coefficient captures the probability of divestment after the fine is estimated to be 9.3%, that is, a higher share of investors permanently divested the year after the fine. In the year before the enforcement, banks’ probability of divestment was 5.3% (6.4% minus 1.1%), and it increased to 8.1% (9.3% minus 1.2%) post-revelation date. Hence, we find evidence of Kant’s categorical imperative from permanent divestment of fined bank stocks. However, banks (Investor Bank $\times$ Post) divest 1.2% less than non-bank institutional investors in each quarter post-revelation date. The slope for banks (Investor Bank) also indicates that banks (Investor Bank) are 1.1% less likely to divest than their peer non-bank institutional investors (IIs). Overall, we find evidence to support our H1 that banks (Investor Bank) are less likely to divest fined banks than non-banks (IIs).

Table 5 Investor reaction to fines on banks on one-year estimation windows

In columns (3) and (4), we add controls for quarter and investor country fixed effects to test whether the differences found in the average probabilities are explained by these variables. Initial differences in permanent divestment are insignificant, while the post-difference is still negative, indicating that after controlling for diverse factors, banks (Investor Bank) had the same probability of permanently selling as non-banks (IIs) prior to the revelation date, but they do not follow their non-bank peers once the misconduct has been revealed. Columns (5) and (6) add investor fixed effects that, although it results in dropping the coefficient for the bank dummy, allow us to confirm the magnitude of the divestment difference after the fines, which is slightly lower than in the previous specification.

Table 6 repeats the same analysis using the 1-to-3 propensity score matched sample that accounts for differences in investor types (asset owner or asset manager), geographic location (country) and quantitative variables like portfolio size or turnover and, more importantly, matches investors with holdings in the same bank one year before the revelation date. Results are similar to the unmatched sample. Selloff probability is still higher previous to the fine, although the difference is lower than in the unmatched sample. Regarding permanent selloff, it indicates again that bank (Investor Bank) divestment is relatively lower than non-banks (IIs) after the fine in the three specifications examined, while it seems that differences are explained by the control variables in the pre-revelation window.

Table 6 Investor reaction to fines on banks on one-year estimation windows matching investors 1-to-3

In Table 7 we assess our main estimates by raising the minimum amount required to consider a penalty when constructing estimation windows. We estimate again the divestment models in 1-year pre-and-post-windows around the revelation date using only fines that eventually involved a penalty equal to or above $100 M, $200 M or $500 M. Matching is performed again for every subsample and then we re-estimate the models obtaining, in all cases, that banks (Investor Bank) are still less likely to divest their fined bank peers after the fine than their non-bank peers (IIs), with similar post-coefficients as obtained with the $50 M threshold in the $100 M samples with even larger differences in Investor Bank and II divestment when considering only fines above $200 M or $500 M.¹⁹

Table 7 Investor reaction to fines above $100 M, $200 M or $500 M on one-year estimation windows matching 1-to-3

We further examine the impact of bank interconnectedness on bank divestment relative to non-banks using ratings from G-SIBs (Global Systematically Important Banks), established by the Financial Stability Board (FSB) and the Basel Committee of Bank of International Settlements (BIS). We collect the list of G-SIB banks from the BIS website.²⁰ Banks that are listed in G-SIBs are considered to be more important to the financial system because of their size and their interconnectedness with other banks. BIS has assigned banks to buckets, numbered from one to five, where five represents higher importance to the financial system, since late 2011. Since our sample starts in 2005, we backfill the ranking for the invested banks. To avoid assigning a very different rank to a bank compared to hypothetical BIS ranks prior to 2011, we categorize banks in two different ways. First, we test the results with a simple dummy that takes one if the invested bank is on the G-SIBs ranking. Second, we categorize banks into three buckets, non-G-SIB, Low-Rank G-SIB and High-Rank G-SIB, with the threshold for splitting between Low- and High-Rank set to three, and where banks with a level of three or higher are assigned to the High-Rank bucket. We choose this threshold for two reasons. First, it evenly splits the buckets, given that any bank has a maximum rank of five. Secondly, the ranking is very stable at every level but particularly around three; in the period 2011–2019, only three banks out of thirty-eight permanently crossed our imposed threshold.

The three-way interaction variable (Investor Bank $\times$ G-SIB Invested Bank $\times$ Post) in column (1) of Table 8 shows that Investor Bank are less likely to temporarily divest fined banks that are G-SIBs banks (G-SIB Invested Banks) than non-G-SIBs fined banks during the post-revelation dates of misconducts (Post), while column (3) shows the likelihood of permanent divestment is negative but not significant. However, when we interact the low- and high-rank variables (Investor Bank $\times$ G-SIB Invested High-Rank $\times$ Post and Investor Bank $\times$ G-SIB Invested Low-Rank $\times$ Post, being no-rank the baseline) in columns (2) and (4) of Table 8 we see that Investor Bank divest less of fined banks that are high-rank G-SIBs (G-SIB Invested High Rank) during the post-revelation period. Importantly, low-rank G-SIB banks were more likely to be permanently divested in the post-period (G-SIB Invested Bank Low-Rank $\times$ Post), while no difference was found for high-ranked G-SIB banks (G-SIB Invested Bank High-Rank $\times$ Post). This indicates that fined G-SIBs banks with higher G-SIBs rankings (i.e., banks that are considered systematically important for the entire banking industry due to their size, interconnections, etc.) are not more likely to be divested after the revelation day. This holds even for non-bank investors (IIs) and, in this case, Investor Bank are still less likely to divest than non-bank investors (IIs).

Table 8 Investor reaction to misconduct of banks matching accounting for the invested bank G-SIB Invested Bank status

Overall, our results support our H1 that bank-affiliated investors (Investor Bank) do not show differences in selloff decisions prior to the revelation date after controlling for confounding variables, but bank-affiliated investors (Investor Bank) are less likely to divest compared to non-bank investors (IIs) after the misconduct is revealed. We find evidence that banks divest less for fined banks that are more important to the financial system’s stability (i.e., banks listed in the G-SIBs).

We test Hypothesis 2 (H2) in Table 9, estimating the same divestment models in a subsample containing only banks that allow us to evaluate whether PRI banks are more likely to divest than non-PRI banks. We would expect signatories to have a higher probability of divestment; however, the observed result is the opposite, as signatory banks were more reluctant to divest in the post-period than non-PRI signatories. In Table 10, we match banks 1-to-3 by PRI status in the same fashion as banks are matched to non-banks in Table 6. Results show no difference in divestment behaviour, nevertheless, these results indicate that PRI signatory bank divestment is not significantly different from non-PRI signatory banks. Our findings here indicate that banks’ participation in PRI is driven by certain subjective conditions (i.e., reputation building, ability to collect higher fees, etc.), as shown in Liang et al. (2022). Hence, we do not find support for our H2, with our results suggesting that the motive to participate in PRI is driven by Kant’s hypothetical imperative as opposed to the categorical imperative. Our findings are consistent with the claim that PRI tends to downplay the importance of divestment as an engagement strategy for SRI (Dawkins 2018). Our findings are also consistent with Bae et al. (2021) who find that the stock performance of 181 US companies that participated in the Business Roundtables is not different from those who did not participate in the Business Roundtable during COVID19. Hence, such commitment represents rhetorical public relations as opposed to representing meaningful actions towards stakeholder orientation.

Table 9 PRI vs non-PRI banks reaction to fines on banks on one-year estimation windows

Table 10 PRI vs non-PRI banks reaction to fines on banks on one-year estimation windows matching investors 1-to-3

We test our third hypothesis (H3) by examining whether fined banks divest their fined bank peers with the same intensity as their non-fined bank peers. There are very few observations involving own-ownership, i.e., holding, thus we are only able to test bank reactions to similar misconduct to their own.²¹ Table 11 shows the reaction of fined banks compared to non-fined banks. Banks were matched to account for differences between fined and non-fined banks, choosing a 1-to-1 match as it achieved a better covariate balance than the 1-to-3 match. We find no significant differences in divestment behaviour between fined banks and non-fined banks. Overall, we do not find evidence to support H3 with our results indicating that even banks that received large penalties and, therefore, whose legitimacy is in question do not show their moral commitment to divest their fined bank peers. Thus, divestment behaviour for fined banks does not seem to follow the stakeholder salience and the legitimacy theory (Majoch et al. 2017; Gond and Piani 2013) and stigmatization (Hunt and Weber 2019). We interpret this finding as evidence to support Kant’s moral reasons, where banks’ divestment is driven by low conative moral reasons (Kleingeld 1998; Sullivan 1989), i.e., low perceptions that motivate fined banks to divest stocks from other fined banks peers. The reciprocity norms in the banking industry strongly influence bank’s perception of their peer fined banks to such an extent that even banks that have been found to engage in misconduct do not divest their fined bank peers.

Table 11 Fined Banks vs Non-Fined Banks reaction to misconduct of other banks

Finally, we test our fourth hypothesis (H4) by examining whether European banks are more likely to divest than United States banks. Table 12 presents results from the estimation of the divestment models on a European and United States Banks subsample. The table shows results on a matched sample, in which each European bank is matched with a US pair. The proportion of US-EU banks is close to fifty-fifty, therefore we choose a 1-to-1 match instead of 1-to-3 as used in the bank-affiliated and PRI models, where the proportion of bank/non-bank and PRI/non-PRI are more uneven.²² The results show evidence to support our H4 that European banks are more likely to permanently divest their fined bank peers than US-based banks. Therefore, we conclude that cultural and social norms for stakeholder orientation play an important role in banks’ divestment of their fined bank peers.

Table 12 European vs US Banks reaction to misconduct of other banks matching investors 1-to-1

Conclusion

Divestment and active stewardship (best-in-class) engagement are considered two non-mutually exclusive approaches to SRI. While active engagement has become increasingly popular among the ESG investment community, we argue that active engagement without divestment is ethically fraught (Dawkins 2018). Based on Kantian ethics (Kant 1785), we argue that divestment represents the categorical imperative based on the absolute universal ethical and moral commitment to SRI. Following recent studies (Van de Vijver 2022; Lenz 2020; MacArthur 2019; Robinson and Shah 2019), we argue that institutional investors can be considered as collective moral agents as shown in studies of misconduct cases (e.g., Lilly et al. (2021); Stevens (2013)). Hence, institutional investors as collective moral agents are subject to Kantian ethics. We focus on the banking industry given the growing regulatory concerns regarding ethical violations, malfeasance, and misconduct in the banking industry. Thus, we examine whether banks divest their equity holdings in their own peer banks who have engaged in misconduct cases (fined banks). However, due to their interconnected business operations through interbank lending, strategic alliances, portfolio diversification, and market expansion (access to market), the banking industry has been known to follow reciprocity norms. Therefore, bank divestment decisions, while possibly taking into account ethics and socially responsible investing, could also be influenced by reciprocity norms, financial considerations, and reputational reasons.

Our findings reveal that banks’ (the equity side of the banking business or Investor Bank) are less likely to divest than non-bank institutional investor peers (IIs). This result is robust to matching and more restricted subsets of fines with higher sanction monetary thresholds. Therefore, we find evidence that reciprocity norms among banks influence their divestment. We also find that banks that participate as PRI signatories, at best, are not more likely to divest fined banks stocks than non-PRI signatory banks. Hence, banks commitment to participate in PRI seems to be driven by the hypothetical imperative since it is subject to certain conditions or incentives (i.e., financial goal, reputation, responding to social pressure, etc.). We find that banks whose own legitimacy to operate are in question (received penalties or fined banks) also do not have a greater likelihood to divest than non-fined banks. We deduce that bank divestment decisions follow a low conative moral view due to a strong reciprocity norm within the banking industry. We also find that European banks are more likely to permanently divest fined bank peers than US-based banks. Thus, we conclude that cultural and social norms towards broader stakeholders among European banks influence banks’ universal moral and ethical commitments towards divestment of their own bank peers who have been found to engage in misconduct.

Our study provides several implications. Our findings bring awareness among ESG and SRI communities that, despite increasing regulatory and ethical concerns in the banking industry, banks are strongly bound by the reciprocity norm to the extent that they are less likely to divest their own bank peers who have violated moral and ethical standards, even when their own legitimacy is in question. Hence, we propose that the regulatory agencies and various stakeholders should continuously monitor reciprocity norms in banking practices to prevent potential ethical and moral violations that stem from banks’ reciprocity practices. Second, according to their divestment behaviour, banks’ participation in PRI seems to be driven by their own incentives (e.g., financial, reputation, response to public pressures, etc.). Therefore, similar to the Norwegian Government Pension Fund, PRI may need to establish explicit standards and imperatives for divestment over inviolable norms and ethical standards. Furthermore, PRI just started to delist a very small number (five) signatories that are not meeting PRI standards in 2020.²³ Therefore, the impact of PRI delisting to ensure that banks that participate as PRI signatories follow PRI standards is currently relatively weak. Finally, our finding of divestment in European banks compared to US banks implies that cultural and social norms towards stakeholder orientation significantly influence banks’ ethical and moral commitments to SRI through divestment.

Our study contributes to several streams of academic literature. First, we contribute to the literature on the incorporation of ESG issues within banks, which has focused on how banks incorporate responsible investing in the whole market (Chih et al. 2010; Hoepner et al. 2016; Houston and Shan 2022) and in environmental topics (Cojoianu et al. 2021), but not on how banks react to misconduct of their own kind. Second, we contribute to the literature on enforcement actions and, more specifically, enforcement actions (EAs) on banks. We use EAs as a proxy for bank misconduct and document that different types of institutional investors divest differently to EAs. Finally, we contribute to the literature that studies how the asset management side of banks is run. Researchers have found that this part of a bank’s business is conditioned by other parts of the business with the intention of improving their relationships with lending (Massa and Rehman 2008) or IPO (Hao and Yan 2012) clients at the expense of financial performance. In our case, this negative result could also imply investment decisions that are not aligned with their clients’ objectives and ethical preferences.

Our study has some limitations. First, the frequency of the ownership data is quarterly, while some banks have received multiple penalties within a year. Therefore, we have filtered out fines by using a quantity threshold and ignoring the latest fines that may be imposed on a bank during a short period of time to avoid overlapping estimation windows. Although this allows us to estimate the Diff-in-Diff models with the most relevant fines, we have discarded a small number of relevant fines. Previous research has shown that banks use equity investing to build the relationships with clients and interbank lending. However, access to interbank lending data is restricted to the regulators and banks (Alfonso et al. 2013; Cocco et al. 2009; Temizsoy et al. 2015; Bräuning and Fecht 2017). We attempt to capture interbank relationships by examining banks that are considered to be systemically important using the G-SIBs lists from the FSB and BIS. We find evidence that banks divest significantly less on fined banks when these fined banks are G-SIBs banks, in particular, when they are high-ranked G-SIBs. We also control for fined banks that have the same geographic location (state or county) as a proxy for “home bias” and our untabulated results are similar to the reported results. Nevertheless, given the importance of interconnected relationships from interbank lending, future work could focus on capturing this effect if the bank-level data become available.

Appendices

Appendix A Variables Appendix

Table 13.

Table 13 Variables description

Appendix B Additional Regressions

We re-estimate the divestiture models on different time windows. Table 14 uses two-, six- and eight-quarter windows in the matched sample. Note that given the Diff-in-Diff setup covered in section “Methods”, the number of included penalties is different as the estimation window changes, thus the two-quarter window will contain the maximum number of enforcement actions and the eight-quarter window the minimum. We also redo the matching for every alternative time window, finding matches at the start of each respective pre-period. Results do not find a difference in trading behaviours over two and eight quarters, being six-quarter results similar as displayed in the main results section. However, raising the penalty threshold finds again differences between banks and no-banks. In Table 15, we raise the penalty threshold to $200 M, finding relatively lower divestment from banks in all the specifications.

Table 14 Investor reaction to fines on banks on 2-, 6- and 8-quarter estimation windows matching investors 1-to-3

Table 15 Investor reaction to fines on banks above $200 M on 2-, 6- and 8-quarter estimation windows matching investors 1-to-3