Monetary policy, ownership structure, and risk-taking at financial intermediaries

Thispaperexamineshowownershipstructureinteractswith monetary policy in shaping financial intermediaries’ appetite for risk. By constructing a large panel of banks across Western Europe, we provide evidence that differences in bank ownership influence the transmission of monetary policy via the risk-taking channel. While shareholder banks actively adjust the riskiness of their portfolios to changes in interest rates, stakeholder banks appear to be less responsive to such changes. These findings call for greater attention to the nature of bank ownership when setting monetary policy.

little attention has been paid to how ownership structure interacts with monetary policy in shaping banks' appetite for risk.
The aim of this paper is to cast new light on the role of banks in monetary policy by examining how bank ownership affects the transmission of monetary policy via the risk-taking channel.We test whether banks that pursue social as well as financial objectives (i.e., stakeholder banks) are less responsive to changes in interest rates than are profit-maximizing banks (i.e., shareholder banks).The results, based on a large panel of commercial, cooperative, and savings banks from 17 Western European countries, provide robust evidence that banks with alternative ownership types respond differently to changes in monetary policy.While shareholder banks tend to actively adjust the riskiness of their portfolios to changes in interest rates, stakeholder banks are found to be less responsive to such changes.Evidence that stakeholder banks' risk-taking varies less with interest rates than does that of shareholder banks is also available for the crisis years, when monetary policy continued to be effective at influencing the share of risky assets held in stakeholder banks' (but not shareholder banks') portfolios.In addition, we show that the loose monetary environment in the aftermath of the crisis delayed loan loss provisioning by stakeholder banks, possibly reflecting a positive impact of the accommodative monetary policy on borrowers' creditworthiness.Overall, these results suggest that stakeholder banks generally follow fewer procyclical risk-taking policies compared with their shareholder counterparts.
The property rights (Alchian & Demsetz, 1972) and agency theory (Jensen & Meckling, 1976) literature views ownership type as a key determinant of firm risk-taking.Shareholder-owned commercial banks (commercial banks hereafter) are characterized by a separation between ownership and debtholding (Valnek, 1999).Stockholders have all the cash flow rights and control the board by virtue of their voting rights (Bøhren & Josefsen, 2013), with votes apportioned according to the amount of capital provided to the bank (Hansmann, 2000).The one-share-one-vote rule to allocate cash flow rights and voting rights is, together with publicly traded common stock and the market for corporate control, one of the main mechanisms for motivating managers to maximize owners' utility (Valnek, 1999).
By contrast, the distinguishing feature of cooperative banks, which together with savings banks belong to the group of stakeholder banks, is that they are owned by their members, who are households and businesses whose aim is not exclusively the maximization of their ownership stake in the bank (Ayadi et al., 2010).It is optimal for depositors to own the bank, that is, for the bank to be a mutual when the cost of capital to depositors is not markedly larger than that to shareholders (Habib, 2018).Members of cooperative banks are entitled to only one vote (one-member-one-vote principle), stakes are generally not marketable, and the distribution of profits is limited.Like financial cooperatives, savings banks are not strictly profit-oriented institutions and are characterized by a dual financial and social mission (Ayadi et al., 2009).However, savings banks differ from financial cooperatives in that they are owned either by an organization that belongs to the government or by a foundation.The lower incentives for stakeholder banks to use leverage in an attempt to increase the expected return on equity, along with the significant obstacles (at least for cooperative banks) in raising external capital, may make stakeholder banks less prone to risk-taking (Llewellyn, 2017).This paper contributes to the literature in several ways.In responding to recent calls for a better understanding of banks' incentives to increase their risk exposure (Bikker & Vervliet, 2018), it brings concepts from the property rights and agency theory perspectives into the analysis of the risk-taking channel.By estimating the differential effects of monetary intervention on bank risk-taking owing to ownership structure, our findings advance the literature concerned with the role of financial intermediaries as conduits for monetary policy transmission.These findings provide new evidence on the functioning of the risk-taking channel in Western European countries during periods of financial distress while contributing to a greater understanding of the implications that the ownership composition of the banking sector has for financial stability.
The remainder of the paper is organized as follows.Section 2 reviews the related literature and advances the key theoretical predictions.Section 3 describes the sample and key variables used in the analysis, alongside the econometric model to be estimated.Section 4 illustrates the main empirical results and summarizes a series of robustness checks.Section 5 discusses the implications of the findings, along with some suggestions for future research, and concludes.

Monetary policy and bank risk-taking
Fueled by the 2008 global financial crisis, an increasing number of studies have posited that changes in official rates affect banks' risk perception and tolerance through a risk-taking channel of monetary policy transmission.In a nutshell, this channel works via three primary mechanisms (Borio & Zhu, 2012): (1) the impact of interest rates on valuations, incomes, and cash flows; (2) the existence of "sticky" target rates of return; and (3) the reaction function and communication policies of the central bank.Among these mechanisms, particular attention has been devoted by researchers to the link between interest rates and the search-for-yield effect.A prolonged period of low interest rates may induce a degree of procyclical risk-taking into the financial system, eventually generating an equilibrium with deteriorated bank portfolios, lower and more volatile profits, and higher aggregate credit (Dell'Ariccia & Marquez, 2006).
Drawing on this, empirical research has started to explore the link between monetary policy and banks' appetite for risk.Evidence of a risk-taking channel is found for both the US and European financial systems.Paligorova and Santos (2017) collect data at the bank holding company (BHC) level and find that US banks charge riskier borrowers lower loan spreads in periods of monetary easing than in periods of monetary tightening.The relationship between policy rates and US banks' risk-taking appears to be more pronounced for domestic banks of smaller size (Buch et al., 2014) and for better capitalized banks (Dell'Ariccia et al., 2017).Evidence from commercial and savings banks in the United States also suggests that a relatively low interest rate environment might induce banks to alter their business models and expand their trading activities in order to reduce their reliance on the lending business (Bikker & Vervliet, 2018).
Within the European context, early support for a link between interest rates and bank risk is put forward by Delis and Kouretas (2011), who construct a sample of commercial, savings, and cooperative banks from 16 eurozone countries and show that the impact of loose monetary policy on risky assets is amplified for banks with less equity capital as well as more off-balance-sheet (OBS) items.Similarly, Jiménez et al. (2014) use a micro-level dataset for Spain and find support for a risk-taking channel operating through less-capitalized banks.More recent evidence corroborates the existence of a risk-taking channel in the euro area that works through the relaxation of lending standards for borrowers (Neuenkirch & Nöckel, 2018); the strength of this channel is reduced by means of more stringent prudential policy on either bank capital or the loan-to-value (LTV) ratio (Maddaloni & Peydró, 2013).
The negative relationship between interest rates and bank risk seems to hold even if one considers somewhat more heterogeneous samples (Altunbas et al., 2014).Based on a panel of commercial, savings, and cooperative banks from 10 Central and Eastern European (CEE) countries as well as Russia for the period 1997-2011, Drakos et al. (2016) find that foreign, well-capitalized banks from CEE countries behave more aggressively compared with other banks in the sample by increasing their risky investments in response to declining short-term rates.

Ownership structure and bank stability
Extensive evidence in the literature suggests that banks' ownership type has a bearing on their behavior, performance, and ultimate survival (e.g., Fama & Jensen, 1983;Hansmann, 2000;O'Hara, 1981).Empirical research points to a number of differences in the behavior of shareholder banks vis-à-vis stakeholder banks.
By employing euro-area data covering the global financial crisis, Ferri et al. (2014) offer evidence suggesting that stakeholder banks attempt to smooth financial conditions for their customers by adopting fewer procyclical lending policies than do shareholder banks.These findings are complemented by Meriläinen (2016), who finds that the lending growth of cooperative and publicly owned savings banks across 18 Western European countries was less affected by the global financial crisis and the subsequent sovereign debt crisis compared with commercial banks.
Importantly, the literature lends support to the view that ownership structure has an impact on bank stability.Early evidence from the United States indicates that stakeholder banks are generally less risk-inclined than their shareholder peers (O'Hara, 1981).Moreover, there is evidence that stock firms exhibit higher concentration in lines of business and geographic areas with the greatest risk (Lamm-Tennant & Starks, 1993), together with the adoption of high-risk strategies through an investment in risky assets and a mismatch between assets and liabilities (Esty, 1997).
Similar evidence is found in Europe.Ayadi et al. (2009Ayadi et al. ( , 2010) ) compute Z-scores for cooperative, savings, and commercial banks operating in six Western European countries and show that cooperative and savings banks are generally more financially stable than commercial banks.Likewise, García-Marco and Robles-Fernández (2008) focus on the Spanish context and submit that savings banks have a lower insolvency risk relative to their commercial counterparts.
Differences in default risk between shareholder-and stakeholder-oriented banks appear to be explained by the lower volatility of returns for stakeholder banks compared with their shareholder counterparts (Hesse & Cihák, 2007).In addition, empirical support is found for significant differences between the two ownership types in terms of loan quality, with stakeholder banks having lower nonperforming loans (Beck et al., 2009) and loan loss provisions (Iannotta et al., 2007) than shareholder banks do.A recent study by Meriläinen (2019) also shows that cooperative banks' loan loss provisions have a smaller cyclical component relative to other ownership types, as demonstrated by the limited effect of gross domestic product (GDP) growth on their provisioning.
Combined, the two major bodies of literature reviewed above suggest a key testable prediction: banks that seek to balance the interests of a multiplicity of stakeholders tend to be less responsive to changes in the monetary environment compared with banks that focus primarily on maximizing shareholder wealth.To disentangle the effects of ownership structure on bank risk-taking, our econometric specifications also include various other bank-level characteristics (e.g., size, capitalization, and profitability) along with industry-specific and macroeconomic factors.

Sample selection
The primary source of data is Bankscope, a global database of banks' financial statements and ownership structures maintained by Bureau van Dijk.We use annual report data for a panel of banks operating in 17 Western European countries, namely the 15 economies that joined the European Union before the 2004 accession (i.e., Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, and the United Kingdom) as well as Norway and Switzerland. 1 The sample period is from 1999 (the year in which the euro was officially launched) through 2011 (the last year for which consistent data on the regulatory indices is available). 2 This time window is interesting as it encompasses the global financial crisis as well as the eurozone sovereign debt crisis.The process that we adopted in selecting the sample is described in detail in the supplementary material available online.
Table 1 shows the composition of our sample of 5677 commercial, cooperative, and savings banks by country and ownership structure.Whereas the financial systems in Luxembourg and the United Kingdom are characterized by an overwhelming majority of shareholder banks relative to the total number of banks (97% and 70%, respectively), the financial systems in countries such as Germany (92%) and Austria (76%) have a dominant presence of stakeholder banks.Besides Italy (69%) and Germany (67%), other countries that exhibit a large number of cooperative banks are Austria (52%) and Switzerland (46%). 3Savings banks are strongly present in the Scandinavian region, especially 1 The validity of employing annual data when studying the risk-taking channel is supported by Delis & Kouretas (2011), who build a quarterly dataset with information collected from Bloomberg and find that their results are not sensitive to the frequency of the data.
2 The survey results used to construct the regulatory indices are available at four points in time (i.e., 2001, 2003, 2007, and 2011) and cover the period from 1999 through 2011. in Norway (86%) and Sweden (83%). 4Taken together, these figures confirm the considerable heterogeneity across Western European countries with respect to the ownership composition of the banking system, calling for a greater understanding of the implications that such heterogeneity might have for the transmission of monetary policy to bank risk.

Bank risk-taking
We measure the risk-taking behavior of banks with two proxies commonly used in the literature, namely the ratio of risky assets to total assets (risky assets) and the ratio of loan loss provisions to total loans (LLPs).Risky assets are calculated as the difference between total assets and the sum of loans and advances to banks, government securities, and cash.Therefore, this ratio includes all assets with non-negligible credit and market risk (Gropp et al., 2011) and captures the overall riskiness of bank portfolios at a given point in time (Delis & Kouretas, 2011).The LLPs variable is defined as the sum of provisions against possible losses on nonperforming loans over net loans (i.e., residential mortgage loans, other mortgage loans, other consumer loans, corporate and commercial loans, and other loans minus reserves for loan losses).This variable reflects the quality of loan portfolios (Bertay et al., 2015) and offers a more direct proxy for credit risk (Iannotta et al., 2007), with a higher ratio denoting a poorer credit quality.5 Data for the risk-taking proxies are collected from Bankscope, and descriptive statistics are summarized in Table 2, where both proxies are expressed in percentage terms. 6Over the sample period, the risky assets ratio has an average value of 80% and a standard deviation of 16%.The mean value of risky assets was 77% in 1999 and 80% in 2004, suggesting a 4% increase in the average risk-taking behavior of banks until the mid-2000s.In turn, the LLPs ratio is characterized by a mean of 0.68% and a standard deviation of 1.02%.While the credit risk appetite of banks seemed at its highest in 2002 (0.89%), it reached its lowest level in 2011 (0.28%).
Table 3 presents summary statistics for the dependent variable by dividing the sample according to ownership type.
At first glance, we notice several important differences among banks with alternative ownership structures.Whereas the average value of risky assets for stakeholder banks is higher than for their shareholder peers, largely reflecting their greater focus on the traditional lending business, the lower standard deviation for cooperative and savings banks implies less volatility in their risk-taking behavior relative to commercial banks.On average, stakeholder banks also show a lower and less volatile loan loss provision ratio compared with shareholder banks.

Monetary policy
Since the onset of the global financial crisis, some researchers and market observers alike have blamed the relatively low interest rate environment in the first half of the 2000s for the softening of lending standards by banks and the subsequent materialization of risks in the economy.More recently, a related discussion has ensued on whether the current environment of exceptionally low interest rates is sowing the seeds for the next financial crisis (Dell'Ariccia Note: The table summarizes descriptive statistics for the main regression variables.Risky assets is the ratio of risky assets to total assets; LLPs is the ratio of loan loss provisions to total loans; Overnight rate is the annual average of the daily overnight interbank rate; Size is the natural logarithm of real total assets; Capitalization is the ratio of equity to total assets; Deposits is the ratio of deposits to total liabilities; OBS items is the ratio of OBS items to total assets; Profitability is the ratio of profit before tax to total assets; Efficiency is the ratio of cost to total income; Income diversity is a measure of income diversification; Concentration is the Herfindahl-Hirschman Index of market concentration; Activity restrictions is an index of the extent to which banks can engage in a number of activities; Capital stringency is an index of the regulatory oversight of bank capital; Supervisory power is an index of the power of the supervisory authority to influence the behavior on the part of banks; Deposit insurance is an index of each country's explicit deposit insurance regime; Private monitoring is an index of the degree to which regulatory and supervisory policies affect the private monitoring of banks; Institutions is a composite measure of country-level governance; GDP growth is the annual growth rate of real GDP; Inflation is the annual change in the CPI; Volatility is the annual average of the daily historical volatility of the country's stock market index; House prices is the annual change in the residential property price index (divided by the GDP deflator).Risky assets is the ratio of risky assets to total assets; LLPs is the ratio of loan loss provisions to total loans; Size is the natural logarithm of real total assets; Capitalization is the ratio of equity to total assets; Deposits is the ratio of deposits to total liabilities; OBS items is the ratio of OBS items to total assets; Profitability is the ratio of profit before tax to total assets; Efficiency is the ratio of cost to total income; Income diversity is a measure of income diversification.et al., 2017).For this reason, the main measure of monetary policy used in this paper is the short-term interest rate (overnight rate), computed as the annual average of the daily overnight interbank rate.
After the collapse of Lehman Brothers in September 2008, many central banks sought to maintain financial and economic stability by implementing an unprecedented set of nonstandard monetary policy measures.As a result, central bank balance sheets expanded sharply in many advanced economies, largely reflecting the increase in the amount of liquidity supplied to the banking sector (Gambacorta et al., 2014). 7To disentangle the effects of nonstandard monetary policy measures on bank riskiness from those due to variations in short-term rates, the estimations for the crisis period also include the ratio of central bank assets to nominal GDP (central bank assets) as a proxy for unconventional monetary policy.8

Control variables
We control for a number of bank-level, industry-specific, and macroeconomic factors that might affect banks' appetite for risk.One of the main empirical challenges with studying the transmission of monetary policy to bank risk is to distinguish the risk-taking channel from the partially overlapping bank-lending channel.According to Bernanke and Blinder (1988), an expansionary monetary policy tends to increase credit supply by making it less costly for financial intermediaries to fund loans, leading to an equilibrium with higher aggregate credit and larger bank balance sheets.
Therefore, one concern is that our risky assets measure could go up not because the riskiness of bank loan portfolios increases (risk-taking channel) but because bank balance sheets expand (bank-lending channel).9 In our empirical estimations, we hold these additional supply-side effects constant by including four bank-level characteristics that are commonly used by the literature to capture the bank-lending channel.First, we account for a possible "too-big-to-fail" phenomenon by including bank size (size), defined as the natural logarithm of real total assets.
We also test whether better capitalized banks have stronger or weaker incentives to take on risk by considering the ratio of equity to total assets (capitalization).Finally, because recent empirical evidence suggests that bank-specific characteristics such as deposits and securitization could influence banks' funding needs and explain bank risk (Altunbas et al., 2014;Azam et al., 2022;Guo & Zhang, 2020), we control for the deposits-to-total-liabilities ratio (deposits) and the ratio of OBS items to total assets (OBS items).
Alongside these four bank characteristics, at the micro level, we add the ratio of profit before tax to total assets as a measure of bank profitability (profitability) and the cost-to-income ratio as a proxy for bank efficiency (efficiency), with higher values indicating less efficient banking operations.As empirical findings show that diversification away from traditional lines of business influences bank risk-taking (e.g., Hesse & Cihák, 2007), we also include a variable to control for differences in banks' income (income diversity).10Building on Laeven and Levine (2007), this variable is calculated as follows: The set of industry-related controls comprises the Herfindahl-Hirschman Index as a proxy for market concentration (concentration), computed as the sum of squared market shares of all banks in the country. 11To capture the regulatory environment, we construct five indices using data from the Bank Regulation and Supervision Survey (BRSS) conducted by the World Bank and described in Barth et al. (2001Barth et al. ( , 2004Barth et al. ( , 2006Barth et al. ( , 2012)).12Activity restrictions measures the extent to which banks can engage in a range of activities (e.g., securities underwriting, brokering, and dealing), with higher numbers indicating more regulatory impediments on nonlending activities.Capital stringency proxies for the regulatory oversight of bank capital, with higher values denoting more stringent guidelines on the nature and sources of regulatory capital.Supervisory power reflects the right of the supervisory agency to take actions such as forcing banks to change their organizational structures, suspending directors' decisions to distribute dividends, and declaring insolvency, with a higher index implying greater supervisory power.Deposit insurance captures each country's explicit deposit insurance regime, with greater values pointing to higher protection of depositors in case of bank default.Private monitoring shows the degree to which regulatory and supervisory policies affect the private monitoring of banks; in this case, higher numbers reflect greater incentives for market discipline.
At the country-level, we account for an array of institutional and macroeconomic variables that are likely to influence bank risk-taking.Because there is evidence that greater institutional development contributes to financial stability (Beck et al., 2006;Hesse & Cihák, 2007), the econometric estimations include a composite measure of country-level governance (institutions) constructed using the Worldwide Governance Indicators (WGI) by Kaufmann et al.
(2010). 13ecause the demand for loans is mostly dependent on macroeconomic factors, the empirical literature has tended to discriminate between loan supply and demand by controlling for the state of the macroeconomic environment (e.g., Altunbas et al., 2014;Ferri et al., 2014).We follow this literature and hold demand-side effects constant by conditioning the model on four macroeconomic variables. 14We control for the growth rate of real GDP (GDP growth) and the annual change in the CPI (inflation).Furthermore, we capture developments in stock markets by computing a measure of share price volatility (volatility), calculated as the annual average of the daily historical volatility of a country's stock market index. 15Finally, the empirical setup aims to distinguish the risk-taking channel from the standard "financial accelerator" à la Bernanke et al. (1999).According to Bernanke and Gertler (1995), monetary policy may affect the external finance premium in credit markets via its impact on borrowers' balance sheets (i.e., balance sheet channel).A lower interest rate environment tends to boost borrowers' creditworthiness by reducing their debt payments and increasing their net worth, which in turn may lead to a higher demand for credit in order to support investment and spending.Our empirical setup controls for these demand-side effects by including the annual change in the residential property price index (divided by the GDP deflator) as a measure of the value of borrowers' collateral (house prices). 16able S1 in the supporting material reports the correlation coefficients for the explanatory variables, suggesting that multicollinearity is unlikely to affect the parameter estimates.17

Econometric model
The equation to be estimated has the following functional form: where N is the number of banks, k is the country, and T is the final year.
The risk-taking behavior, y i,k,t , for bank i headquartered in country k at time t is proxied by either risky assets or LLPs. 18ince evidence is found of a relatively high persistence of risk over time (e.g., Delis & Kouretas, 2011), we include the lagged dependent variable among the regressors.
The variable x k,t is the main measure of monetary policy, namely the overnight interbank rate.In line with the risktaking channel literature (Borio & Zhu, 2012), we would expect the coefficient  to be negative.The differential effects of interest rates on banks' appetite for risk owing to ownership structure are captured by interacting the monetary policy variable with an ownership dummy, z i,k , which equals 1 for stakeholder banks as a group (or for cooperative and savings banks separately) and 0 otherwise.
A potential problem with the inclusion of interaction effects in a multiple regression equation is the multicollinearity between the multiplicative term and its constituents.To address this problem, we mean-center the interest rate variable prior to forming the product term.This approach involves transforming the variable x k,t to deviations from its mean and calculating the product term using these deviations.Besides alleviating multicollinearity issues (Iacobucci et al., 2016), mean-centering facilitates interpretation of the constituent term coefficients, in that they will reflect the conditional effect of one term when the other term is at its raw mean (Burks et al., 2019).Consistent with theoretical predictions, we anticipate the parameter  to be positive.As central banks loosen monetary conditions, cooperative and savings banks are expected to take on less risk compared with their commercial peers.
14 Besides separating changes in bank risk that are due to demand rather than supply factors, the inclusion of these demand-side controls allows us to isolate the monetary policy component of interest rate changes (Gambacorta & Mistrulli, 2004).
The bank-specific controls are contained in the vector W i,k,t , while U k,t represents the set of industry-and macrolevel variables.To account for unobserved country-specific factors and time-varying common shocks (e.g., shifts in credit demand through time) that may influence bank risk-taking, all econometric specifications include country,  k , as well as time,  t , fixed effects.
The estimation of Equation ( 2) presents a number of empirical challenges.A major identification limitation when examining the monetary policy-bank risk nexus is that monetary conditions might be endogenous to the risk observed in the banking sector (Jiménez et al., 2014).Within this context, an endogeneity problem can arise if monetary policy actions are also determined by financial stability objectives.This observation might be particularly true since the onset of the financial crisis, as central banks' concerns regarding the financial situation of the banking sector led to a rapid expansion on the set of conventional and unconventional monetary policy measures (Altunbas et al., 2014).
Although one expects endogeneity not to be a major concern in the sample countries prior to the financial crisis, given that their central banks had primarily an inflation-targeting objective, this potential issue can be mitigated further by the use of an appropriate estimation method.From an econometric standpoint, endogeneity implies that the interest rate variable, x k,t , may be correlated with the error term,  i,k,t , thereby biasing our estimates.In addition, the inclusion of the lagged dependent variable on the right-hand side may induce autocorrelation in the residuals and render the ordinary least squares (OLS) estimator biased and inconsistent even if the idiosyncratic errors are not serially correlated (Baltagi, 2013).
To alleviate these concerns and obtain consistent and unbiased estimates of the interplay between monetary policy, ownership structure, and bank risk-taking, we estimate Equation (2) using the system generalized method of moments (GMM) estimator.By building a "stacked" system of equations in both levels and differences, the system GMM allows for unobserved heterogeneity, simultaneity, and the dynamic relationship between dependent and independent variables (Wintoki et al., 2012).Therefore, it ensures the efficiency and consistency of the estimated parameters, provided that there is no second-order serial correlation and that the instrument set is valid.Details about the dynamic panel data estimator employed in this paper are reported in Appendix B.

Full period
Table 4 presents the results of the main empirical estimations when bank risk-taking is proxied by risky assets. 19The results with bank risk-taking measured by LLPs are provided in Table 5 and discussed in the next subsection. 20The Arellano-Bond test for autocorrelation in the errors rejects the presence of second-order serial correlation, AR(2), while the Hansen test of overidentifying restrictions confirms the validity of the instrument set.
The coefficient on overnight rate for the full period is negative and strongly significant, suggesting that lower interest rates alter the composition of commercial banks' portfolios toward riskier positions.This evidence is consistent with a risk-taking channel that operates via the risk decisions of shareholder banks.The interaction term between monetary policy and the stakeholder bank dummy takes a positive and highly significant coefficient, indicating that the effects of monetary conditions on the riskiness of financial intermediaries are lower for stakeholder banks (−0.299 + 0.038 = −0.261).These results offer preliminary support for our initial hypothesis in that banks with alternative ownership structures appear to respond differently to changes in interest rates.

Note:
The table presents the results of the main empirical estimations with bank risk-taking proxied by risky assets.Risky assets is the ratio of risky assets to total assets; Overnight rate is the annual average of the daily overnight interbank rate; Stakeholder is a dummy that equals 1 for either cooperative or savings banks and 0 otherwise; Central bank assets is the ratio of central bank assets to nominal GDP; Size is the natural logarithm of real total assets; Capitalization is the ratio of equity to total assets; Deposits is the ratio of deposits to total liabilities; OBS items is the ratio of OBS items to total assets; Profitability is the ratio of profit before tax to total assets; Efficiency is the ratio of cost to total income; Income diversity is a measure of income diversification; Concentration is the Herfindahl-Hirschman Index of market concentration; Activity restrictions is an index of the extent to which banks can engage in a number of activities; Capital stringency is an index of the regulatory oversight of bank capital; Supervisory power is an index of the power of the supervisory authority to influence the behavior on the part of banks; Deposit insurance is an index of each country's explicit deposit insurance regime; Private monitoring is an index of the degree to which regulatory and supervisory policies affect the private monitoring of banks; Institutions is a composite measure of country-level governance; GDP growth is the annual growth rate of real GDP; Inflation is the annual change in the CPI; Volatility is the annual average of the daily historical volatility of the country's stock market index; House prices is the annual change in the residential property price index (divided by the GDP deflator).All econometric specifications include country as well as time fixed effects.Robust standard errors (clustered at the bank level) are reported in parentheses.*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
With respect to the bank-specific variables, a lower deposits-to-total-liabilities ratio seems to reduce banks' capacity to take on risk, while less profitable but more efficient financial intermediaries tend to have a greater share of assets with non-negligible credit and market risk.By contrast, the results provide no evidence of a statistically significant relationship between other bank-level variables such as capitalization or income diversity and risky assets.In line with the concentration-stability view (Beck et al., 2006), financial intermediaries in more concentrated markets tend to have lower incentives to take on risk.Support is found for a moral hazard problem induced by explicit deposit insurance while other features of the regulatory environment do not appear to exert a material impact on the overall riskiness of bank portfolios.At the country level, the only variable with a statistically significant coefficient over the 1999-2011 period is GDP growth.Consistent with other empirical findings (Lee and Hsieh, 2013), financial intermediaries operating in countries with higher GDP growth hold less risky portfolios.

Precrisis period
As we noted in Section 3, the last few years covered by the estimations saw the adoption by many central banks of unprecedented actions aimed at restoring financial stability.For this reason, we provide an insight into the functioning of the risk-taking channel during times of financial distress by distinguishing between two periods, namely the years before the outbreak of the crisis (i.e., 1999-2007) and the period after the bankruptcy of Lehman Brothers (i.e., 2008Brothers (i.e., -2011)). 21he results for the former period largely resemble those for the whole sample period.Lower interest rates are associated with an increase in risky assets by shareholder banks, with the coefficient on overnight rate being statistically significant at the 1% level.The impact of looser monetary policy on financial intermediaries' risk-taking is reduced for stakeholder banks as denoted by the positive and highly significant coefficient on the multiplicative term.
This evidence is consistent with recent empirical findings (Ferri et al., 2014), according to which the loan supply of stakeholder banks prior to the start of the crisis was less influenced by changing monetary conditions compared with shareholder banks.
Again, we find that banks with lower profitability but greater efficiency have riskier balance sheets whereas the coefficient on deposits becomes insignificant over the precrisis period.As indicated by other studies (Hesse & Cihák, 2007), the results for the years prior to the crisis also show that higher income diversity tends to be associated with increased bank risk.At the industry level, greater stringency in terms of capital regulations, power of the supervisory authority, and market discipline seem to be effective in limiting the risk-taking incentives of banks while further support is provided for a risk-shifting effect associated with deposit insurance.Compared with the full period, market concentration is no longer important in explaining differences in banks' appetite for risk.Moreover, the results for the 1999-2007 period confirm that improved macroeconomic conditions, as captured by faster GDP growth, reduce the overall riskiness of bank portfolios.In accordance with theoretical predictions (Paligorova & Santos, 2017), there is evidence that lower stock market volatility offers financial intermediaries incentives to take on additional risk.

Crisis period
Turning to the crisis period, we note several striking results.The coefficient on the overnight rate is insignificant, implying that a risk-taking channel is no longer operative for shareholder banks.This finding is not surprising, as the havoc wrought by the financial turmoil resulted in a marked increase in risk aversion and a widespread seizure of liquidity in financial markets (Acharya et al., 2009).In contrast, we find that standard monetary policy is still effective at influencing the composition of stakeholder banks' portfolios, as captured by the negative and highly significant coefficient on the interaction term between overnight rate and stakeholder.In showing that stakeholder banks' risk appetite varies less with interest rates than does that of shareholder banks, this evidence is consistent with the literature that points to a less cyclical behavior by stakeholder banks (e.g., Meriläinen, 2016Meriläinen, , 2019)).
To account for the effects of unconventional monetary policy on the functioning of the risk-taking channel, we add the ratio of central bank assets to GDP (central bank assets) and its interaction with the stakeholder bank dummy.The coefficient on central bank assets is positive but statistically insignificant, suggesting that nonstandard monetary policy plays a limited role in counteracting the shift by shareholder banks toward riskless assets.The interaction term reveals that the expansion of central bank assets is associated with a change in portfolio composition only for stakeholder banks.Similar evidence of a positive relationship between unconventional monetary policy measures and risky assets is put forward by Lambert and Ueda (2014) who consider US banks' balance sheet data after the start of the crisis and find that the ratio of risk-weighted assets to total assets tends to increase as central bank assets expand.Lambert and Ueda (2014) show that an expansion in central banks' balance sheets may also delay loss provisioning on existing loans, as discussed in more detail in Section 4.2.
Concerning the bank-level controls, we observe that larger financial intermediaries exhibit a greater exposure to asset risk.Contrary to the precrisis period, the results for the crisis years indicate that more profitable but less efficient banks are more likely to engage in risk-taking.The coefficient on income diversity turns negative in the years following the bankruptcy of Lehman Brothers, implying that more diversified banks responded to the reduced opportunities for income diversification in the aftermath of the crisis by decreasing the riskiness of their loan portfolios.Consistent with this observation, we find that financial intermediaries operating in banking systems characterized by greater regulatory restrictions on nonlending activities hold more assets with non-negligible credit and market risk.The results for the crisis period also provide evidence of a positive relationship between concentration and bank risk-taking, possibly reflecting a moral hazard problem caused by "too-big-to-fail" policies since the onset of the crisis (Mishkin, 1999).
Turning to the macroeconomic environment, the coefficients on GDP growth and volatility lose their significance when examined over the crisis period.In addition, we establish that the deterioration of a country's institutional environment that was observed in a number of economies after the outbreak of the crisis, captured by a drop in the institutions variable from 1.38 in 2006 to 1.31 in 2011, resulted in an increase in risky assets.The same conclusion cannot be drawn for the precrisis period, during which there is no evidence of an economically and statistically significant relationship between country-level institutions and the riskiness of the banking sector. 22The results also reveal that the fall in house prices in the aftermath of the crisis led to a decrease in the share of risky assets in bank portfolios, suggesting that borrowers had to reduce their demand for credit as a consequence of their lower net worth.

Full period
The results of Table 5, where bank risk-taking is proxied by LLPs, confirm and qualify the results based on risky assets.
Over the full period, lower interest rates increase credit risk-taking by both shareholder and stakeholder banks, with this effect being more pronounced for shareholder banks.The results for the bank-level controls show that better capitalization is associated with higher credit quality whereas greater diversification across income sources translates into higher credit risk-taking.In turn, the regulatory environment has a bearing on banks' credit risk, yet at varying degrees.While the stringency of capital regulations is negatively linked to our measure of credit risk, greater power of the supervisory authority and stronger market discipline seem to encourage more loan loss provisioning by financial intermediaries.Concurrently, credit risk-taking is curtailed by faster rates of inflation, more volatile stock markets, and higher values of borrowers' collateral.

Precrisis period
The moderating role of stakeholder banks with respect to the risk-taking channel is corroborated by the results for the precrisis period during which the credit risk decisions of stakeholder banks are found to be less affected by changes in interest rates.Less concentrated markets are likely to increase credit risk-taking by financial intermediaries while a 22 A potential explanation for the result on the institutions-bank risk nexus is that our sample covers 17 Western European countries that are among the most financially and industrially developed economies in the world.These countries tend to be characterized by a strong institutional environment, which remained by-and-large stable in the years before the crisis (the mean value of the institutions variable over the 1999-2007 period is 1.38).

Note:
The table presents the results of the main empirical estimations with bank risk-taking proxied by LLPs.LLPs is the ratio of loan loss provisions to total loans; Overnight rate is the annual average of the daily overnight interbank rate; Stakeholder is a dummy that equals 1 for either cooperative or savings banks and 0 otherwise; Central bank assets is the ratio of central bank assets to nominal GDP; Size is the natural logarithm of real total assets; Capitalization is the ratio of equity to total assets; Deposits is the ratio of deposits to total liabilities; OBS items is the ratio of OBS items to total assets; Profitability is the ratio of profit before tax to total assets; Efficiency is the ratio of cost to total income; Income diversity is a measure of income diversification; Concentration is the Herfindahl-Hirschman Index of market concentration; Activity restrictions is an index of the extent to which banks can engage in a number of activities; Capital stringency is an index of the regulatory oversight of bank capital; Supervisory power is an index of the power of the supervisory authority to influence the behavior on the part of banks; Deposit insurance is an index of each country's explicit deposit insurance regime; Private monitoring is an index of the degree to which regulatory and supervisory policies affect the private monitoring of banks; Institutions is a composite measure of country-level governance; GDP growth is the annual growth rate of real GDP; Inflation is the annual change in the CPI; Volatility is the annual average of the daily historical volatility of the country's stock market index; House prices is the annual change in the residential property price index (divided by the GDP deflator).All econometric specifications include country as well as time fixed effects.Robust standard errors (clustered at the bank level) are reported in parentheses.*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
stricter regulatory environment in terms of scope of banking operations, power of the supervisory agency, and private monitoring tends to reduce credit risk.The results for the years before the crisis also confirm that a boost in house prices raises the value of borrowers' collateral and improves credit quality.

Crisis period
Consistent with the results for risky assets, the estimations for the crisis period indicate that standard monetary policy has an impact only on the risk behavior of stakeholder banks.However, Table 5 reports an opposite sign on the interaction term between overnight rate and stakeholder, implying that stakeholder banks responded to the extremely low interest rate environment following the collapse of Lehman Brothers by increasing the share of risk-related assets in their portfolios (Table 4) and delaying their loan loss provisioning (the same pattern is reproduced in Tables S2-S9 in the supplementary material).This result suggests that the accommodative monetary policy in the aftermath of the crisis might have boosted borrowers' creditworthiness and reduced the need for stakeholder banks to add to their loan loss reserves.23By contrast, it is possible that the rapid expansion of shareholder banks' balance sheets prior to the crisis, exemplified by a 6% increase in the average share of risky assets held in their portfolios between 1999 and 2007 against a 1% increase for stakeholder banks, left them with a plethora of poor quality assets that were little if at all affected by the decrease in interest rates in the aftermath of the crisis. 24Alternatively, the delay in loan loss provisioning could be taken as evidence of income-smoothing on the part of stakeholder banks in an effort to restore their profitability, which was weakened by the prolonged period of low interest rates.Although profit maximization is not their ultimate goal, profitability is central to stakeholder banks' business model in that retained earnings constitute their primary source of capitalization, as opposed to shareholder banks, which generally face fewer obstacles in raising capital (Llewellyn, 2017).By delaying provisions for loan losses, stakeholder banks might have thus sought to preserve their ability to build up capital at a time of financial distress. 25imilarly, nonstandard monetary policy is found to influence credit risk-taking only for stakeholder banks, with an increase in central bank assets leading to a decline in their loan loss provisioning.The opposite sign on the interaction term between central bank assets and stakeholder compared with Table 4, confirmed by Tables S2-S9 in the supplementary material, is consistent with Lambert and Ueda (2014), who show that an expansion in central banks' balance sheets may delay loss provisioning on existing loans while increasing risky assets.According to Lambert and Ueda (2014), there are two possible explanations for the negative relationship between unconventional monetary policy and loan loss provisions.First, banks' incentives to engage in "loan evergreening" practices, whereby banks rollover existing loans or issue new loans to troubled borrowers to prevent them from becoming nonperforming, may be higher within a context of highly accommodative monetary policy.Alternatively, unconventional monetary policy may support economic activity and improve borrowers' creditworthiness, which in turn may lead banks to decrease their loan loss provisioning.Therefore, it is possible that nonstandard monetary policy reinforced the positive impact of lower interest rates in the aftermath of the crisis on the quality of stakeholder banks' portfolios while such impact was not observed for shareholder banks due to their holding of poor quality assets that were marginally if at all affected by unconventional monetary policy.A third explanation that is plausible within the context of our research refers to stakeholder banks' expectations about the effects of nonstandard monetary policy on their profitability.Insofar as stakeholder banks were concerned that nonstandard monetary policy measures, such as quantitative easing, would have flattened the yield curve and reduced their net interest income, they might have postponed provisions for loan losses to maintain their ability to accumulate capital through retained earnings.
Looking at the control variables for the crisis period, we see evidence that financial intermediaries with a higher share of deposits to total liabilities have a higher credit risk exposure whereas capitalization and OBS items are negatively associated with LLPs.At the industry level, we qualify the results based on risky assets by showing that banks sought to alleviate the potentially negative effects of greater activity restrictions on their profitability, which was already under pressure due to the outbreak of the crisis and the ensuing extremely low interest rate environment, by delaying loan loss provisioning.This evidence points to income-smoothing in the form of provisions for loan losses by Western European banks in the aftermath of the crisis.We also find that this effect was at least in part offset by the decline in house prices, which diminished the value of borrowers' collateral and resulted in an increase in loan loss provisions by banks to address the deterioration in credit quality.

Robustness tests
To evaluate the robustness of the results, we perform the following tests: (1) we split stakeholder banks into cooperative and savings banks; (2) we include loans and advances to banks in the definition of risky assets; (3) we replace the risky assets and LLPs proxies with the Z-score as an alternative measure of bank riskiness; (4) we quantify conventional monetary policy using the central bank's official rate (central bank rate) instead of the overnight interbank rate; (5) we exclude banks that are listed on the stock exchange;26 (6) we drop banks that are ultimately owned by the government or another public authority; and (7) we run the analysis on a smaller sample that excludes German banks as they dominate the sample.The results of these tests are presented in Tables S2-S9 in the supplementary material.The coefficients on the variables of interest are similar to those reported in Tables 4 and 5, leaving our findings qualitatively unchanged.

CONCLUSIONS
Recent years have witnessed a revived interest in the far-reaching effects of bank risk-taking on financial stability and economic performance.This paper contributes to a better understanding of how financial intermediaries' appetite for risk is influenced by the monetary conditions prevailing in the economy.Theory suggests that a key determinant of firms' risk-taking is their ownership structure, which ultimately impacts the extent to which multiple stakeholder claims find recognition alongside those by shareholders.By building a large panel of commercial, cooperative, and savings banks across Western Europe, we find robust evidence that heterogeneity of ownership types accounts for a differential impact of monetary policy on financial intermediaries' risk-taking.
While the risk-taking channel is shown to be operative for both shareholder-and stakeholder-oriented banks, the results indicate that the effects of lower interest rates on bank risk are greater for shareholder banks.Comparison of the analyses before and after the onset of the global financial crisis shows that these results are driven by the years prior to the bankruptcy of Lehman Brothers, during which commercial banks are found to adjust the riskiness of their portfolios to changes in interest rates more actively than cooperative and savings banks.The results for the period since the outbreak of the crisis highlight that standard monetary policy is no longer effective in changing the proportion of risk-related assets held by shareholder banks, possibly as a consequence of the sharp increase in risk aversion and the average loss of trust in counterparties brought about by the market turmoil.Conversely, we find that stakeholder banks responded to the accommodative monetary policy in the aftermath of the crisis by continuing to adjust the share of risky assets in their portfolios along with their loan loss provisioning.The observed reaction of stakeholder banks to the unprecedented set of conventional and unconventional monetary policy measures points to fewer procyclical risk-taking policies on the part of cooperative and savings banks relative to their commercial peers.
These results feed into an intense academic and policy debate over the causes of the 2008 crisis.Our evidence concurs with the increasing role of monetary authorities on macroprudential regulation and supervision, as epitomized by the creation under the responsibility of the European Central Bank (ECB) of the European Systemic Risk Board (ESRB) in late 2010.In addition, this study finds that the heterogeneity of ownership types in the European banking sector is important in explaining the effects of monetary policy on bank risk-taking.Therefore, the findings call for a closer monitoring of the ownership composition of the banking sector when setting monetary policy, as bank ownership is shown to influence the functioning of the risk-taking channel.Such monitoring may be particularly important for the euro area as it recovers from the COVID-19 pandemic, in that dissimilarities across countries may account for a differential impact of the common monetary policy on financial and economic outcomes.
The results presented in this paper could be extended in a number of important ways.First, future research might examine how various features of financial intermediaries' ownership structures (e.g., concentrated ownership) influence the functioning of the risk-taking channel.Second, efforts could be directed at examining the impact that conversion of financial cooperatives to joint-stock banks has on their risk appetite and ensuing response to fluctuations in monetary policy.To this end, one might consider a smaller sample of depository institutions than the sample built in this study and construct time-varying proxies for bank ownership.Third, a fruitful line of inquiry could be to extend the results of this research by developing measures capturing the ownership composition of the banking sector that might serve as valuable instruments for monetary authorities and other banking regulators.We are currently working on this major endeavor.

OBS items
Ratio of OBS items to total assets.

Profitability
Ratio of profit before tax to total assets.

Efficiency
Ratio of overheads to total operating income.

Bankscope; authors' calculations
Income diversity Bankscope; authors' calculations Industry-specific controls Concentration Herfindahl-Hirschman Index of market concentration.The index is calculated as the sum of squared market shares of all banks in the country in terms of total assets.

Activity restrictions
Index that captures the extent to which national regulations restrict banks from engaging in: (1) securities activities, (2) insurance activities, (3) real estate activities, and (4) ownership of nonfinancial firms.

Capital stringency
Index that measures the stringency of regulatory capital requirements.

Supervisory power
Index that proxies for the power of the supervisory authority to influence the behavior on the part of banks.

Deposit insurance
Index that describes the explicit deposit insurance regime adopted in the country.

Private monitoring
Index that quantifies the incentives for private investors to monitor and exert effective governance over banks.

Institutions
Simple average of six country-level governance indicators, namely "voice and accountability," "political stability and absence of violence," "government effectiveness," "regulatory quality," "rule of law," and "control of corruption."WGI; Kaufmann et al. (2010); authors' calculations

GDP growth
Annual growth rate of real GDP.

Inflation
Annual change in the CPI.

Volatility
Annual average of the daily historical volatility of the country's stock market index with a 30-day window.

House prices
Annual change in the residential property price index (divided by the GDP deflator).

BIS; ECB; WDI; authors' calculations
Note: BRSS is the Bank Regulation and Supervision Survey by the World Bank (Barth et al., 2001(Barth et al., , 2004(Barth et al., , 2006(Barth et al., , 2012)); IFS are the International Financial Statistics by the IMF; WDI are the World Development Indicators by the World Bank; WGI are the Worldwide Governance Indicators by the World Bank (Kaufmann et al., 2010).

3
Cooperative banks in Italy include the group of Banche Popolari that converted to joint-stock banks following the 2015 reform.TA B L E 1 Distribution of banks by country and ownership structure The table shows the composition of the sample by country and ownership structure.Shareholder banks are commercial banks while stakeholder banks include cooperative and savings banks.TA is the annual average of total assets in billions of US dollars.EA-12 are the founding euro area countries, namely Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain.EU-15 are the economies that joined the European Union before the 2004 accession, that is, the EA-12 countries plus Denmark, Sweden, and the United Kingdom.The sample period goes from 1999 through 2011.Sources: Bankscope; authors' calculations.
The table presents summary statistics for the bank-specific variables by dividing the sample according to ownership type.Shareholder banks are commercial banks while stakeholder banks include cooperative and savings banks.
TA B L E 3 Summary statistics of bank-level variables by ownership structure