Master Thesis
The impact of a low interest rate environment on
bank profitability and bank risk-taking: evidence
from the U.S. and the euro area
Author: Mart Menger
m.i.menger@student.rug.nl S2358565
University of Groningen
Faculty: Faculty of Economics and Business Program: MSc. Finance & MSc. Economics
Thesis Supervisor: S. Pool Second Assessor: dr. L. Dam
January 9, 2020
Abstract: In this thesis the effect of a low interest rate environment on bank profitability and risk-taking is examined. The research of two dynamic models is executed with the use of the system GMM estimator and a dataset consisting of more than 9,000 banks from the USA and the euro area. The results show that interest rates and the yield curve slope positively affect return on assets. The effect of interest margins on profitability is sufficient to overcome to the opposite effect of provisioning and non-interest income. The results provide weak evidence of a negative effect of the yield curve slope on bank risk-taking. Finally, this thesis shows that profitability erodes over time, because every additional year of negative short-term interest rates further reduces bank profit.
1
Introduction
With the start of the financial crisis, concerns have emerged about the soundness of banks and the financial sector in general. The low interest rate conditions that are present for a prolonged period create a challenging environment for banks. Interest rates have fallen worldwide since the 2000s due to fear for economic slowdown. After the crisis started, the interest rates decreased even more to mitigate the negative effects of the financial crisis. Central banks in the USA, Europe and Japan have executed unconventional monetary policy to stimulate the economy and to meet the inflation target levels. This monetary policy included credit and quantitative easing (QE) and the reduction of the short-term policy rate by the ECB to negative levels.
The expectation is that a lower interest rate environment in the market puts pressure on the profit model of banks. In the short-run, banks will be able to lower the deposit rates to compensate for the loss of lower lending rates. However, if deposit rates approach zero, banks face the zero lower bound because it is assumed that deposit holders will not accept a negative interest rate on their deposit. In this situation, the deposit rates of commercial banks stabilize, while the lending rates continue to decrease. This compresses the profit margins of banks as a result of the reduced net interest rate margin and possibly weakens the capital position and solvency. A bank’s main activity, that of loans and deposits, is greatly affected by interest rates. However, in addition, banks also gain profit with non-interest income. Earlier research has shown that banks increase their non-interest income in times of low interest rates, mainly driven by capital gains, fees and commissions. (Borio, Gambacorta, and Hofmann, 2017; Altavilla, Boucinha, and Peydr´o, 2018; Claessens, Coleman, and Donnelly, 2018; Lopez, Rose, and Spiegel, 2018) Furthermore, unconventional expansionary monetary policy reduces interest rates which increases bank lending. At the same time it reduces the amount of non-performing loans and loan loss provisions and therefore boosts the profit of banks. In the end, it remains ambiguous what the net effect of low interest rates will be on a bank’s profitability.
return a risk premium. In the short-term, the low interest rates will result in lower credit risks due to lower interest payments on variable interest loans and therefore higher cred-itworthiness of customers. However, the low interest rates could lead to relaxed lending standards in terms of granting loans to more risky borrowers. These risky borrowers have a higher probability of default and therefore the relaxed lending standards lowers the loan portfolio quality. In the long run, the rise of interest rates will cause higher credit risk and eventually higher credit losses.
In this thesis the effect of low interest rates on bank profitability and credit risk-taking in the USA and the euro area is investigated. Dynamic panel data consisting of bank level data and macroeconomic data is used in this research. With the help of the system generalized methods of moments (GMM) estimator, two different models are examined. The focus is on the effect of interest rates and not on expansionary monetary policy, because expansionary monetary policy does not necessarily imply low interest rates. In addition, an analysis on the effect of an extension of the low interest rates period and specifically the negative interest rate environment on bank profitability and bank risk-taking is performed. Longer periods of low interest rates and negative rates could potentially erode bank profit and adjust the risk behavior of banks over time.
The research results show that a decrease in interest rate and the yield curve slope diminishes interest margins and return on assets. Moreover, it is found that every extra year of negative interest rates will deteriorate return on assets. This is evidence that banks are unable to overcome the decreased interest income with income from other banking activities. In the risk-taking analysis, it is found that a decrease in the yield curve slope leads to a higher percentage of risk weighted assets. However, this result is not robust to a larger time frame, different estimation methods and different dependent variables and control variables. Furthermore, the results show that every extra year the short-term interest rate remains below 50 basis points, risk weighted asset intensity at banks increase. Although the results are not robust and the GMM instruments are not valid, this thesis finds weak evidence for the risk taking channel of monetary policy.
adequate credit to the market. Therefore, bank profitability contributes to bank soundness and in the end to financial stability. The results of this study are interesting for central bank policy makers, in order to understand the impact of expanding monetary policy and as a consequence the impact of low interest rates on the profitability and risk behavior of commercial banks. Furthermore, the results of this thesis are helpful for prudential super-visors, because they should recognize the conditions in which commercial banks increase their risk, in order to adapt the rules of the game to these implications.
2
Literature review
2.1
Profitability
This section will discuss the literature on the effect of interest rates on bank prof-itability. Studies such as Claessens, Coleman, and Donnelly (2018), Borio, Gambacorta, and Hofmann (2017) and Cruz-Garc´ıa, de Guevara, and Maudos (2019) found a positive effect of interest rates on profitability and net interest margins. Claessens, Coleman, and Donnelly (2018) concluded that a longer period of low interest rates decreases profitability even more. Furthermore, the paper states that the effect of low interest rates is relatively weaker on return on assets compared to net interest margin of banks. Possible explana-tions for this outcome are valuation gains, cost reduction and raising non-interest income. Borio, Gambacorta, and Hofmann (2017) find that the profitability of a commercial bank diminishes during times of unusually low interest rates and a flat term structure. Cruz-Garc´ıa, de Guevara, and Maudos (2019) find that the net interest margin diminishes with lower interest rates, however they present a smaller effect with a flattening of the yield curve. Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009) find, with evidence from banks in China that higher real interest rates increase the profitability of banks.
the diminished interest margins with either higher volumes of loans, higher asset prices gains, decreased funding costs, greater operational efficiency that lead to cost savings or consolidation. However, there will be a limit to these mitigation actions.
Only Trujillo-Ponce (2013) are able to find a negative effect of interest rates on bank profitability. The negative effect is found on the relation of the MRO1 rate on both return on assets and return on equity with data of Spanish banks.
It is concluded that in the short-run, interest rates and the yield curve slope have a positive effect on interest margins, however this effect is offset by non-interest income in the overall profitability of banks. On the other hand, in the long-run bank profit is eroded by low interest rates. Finally, it is important to note that the generalized methods of moments (GMM) is a common method to study profitability of banks. Many of the above mentioned papers on bank profitability use GMM, such as: Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009), Trujillo-Ponce (2013), Dietrich and Wanzenried (2011), Athanasoglou, Brissimis, and Delis (2008), Alessandri and Nelson (2015), Borio, Gambacorta, and Hofmann (2017), Cruz-Garc´ıa, de Guevara, and Maudos (2019) and Altavilla, Boucinha, and Peydr´o (2018).
2.2
Risk
If indeed lower interest rates decrease a bank’s profitability, it will be interesting to investigate if and how banks respond to such a situation. A potential solution to the lower profitability of banks could be the increase of risky investments in the search for a higher yield. Several studies such as Delis and Kouretas (2011), Altunbas, Gambacorta, and Marques-Ibanez (2010), Ioannidou, Ongena, and Peydr´o-Alcalde (2008) and Jim´enez, Ongena, Peydr´o et al. (2014) show that a low interest rate environment indeed increases risk-taking of banks. Delis and Kouretas (2011) find strong evidence of this relation in the period before the crisis, with the exception of French banks, where the effect is less prevalent. Furthermore, they showed that the effect is smaller for banks with higher equity capital and larger for banks with high off-balance sheet activities. Altunbas, Gambacorta, and Marques-Ibanez (2010) examined the effect of unusual low interest rates and yield curve slope on bank risk-taking over a longer period and found significant evidence of a negative relationship. Ioannidou, Ongena, and Peydr´o-Alcalde (2008) use evidence from Bolivia to present the same effect of low interest rates on bank risk-taking. Furthermore, they find that a low federal funds rate before loan origination raises the probability of
bank loan defaults. Jim´enez, Ongena, Peydr´o et al. (2014) take the overnight rate as explanatory variable and examines the Spanish banking market. According to this paper, bank characteristics and market characteristics do play a role. For example, small banks, highly liquid banks and commercial banks take on more risk with low interest rates.
On the other side, there are papers that find (partly) contradicting results. De Nicol`o, Dell’Ariccia, Laeven et al. (2010) argue that the relationship between inter-est rates and bank risk-taking is more complex and nuanced than other papers argue. The paper uses US lending surveys and US call reports for two similar investigations. With both data sources, they found support for the view that banks increase risk in the search for yield and increase risk in the channel of leverage and asset prices. However, in the short run they found evidence for an opposite effect when banks are low capitalized. Therefore, the impact of low interest rates on bank risk-taking will differ across markets, time, regu-lation, position on the business cycle and other factors. According to Dell’Ariccia, Laeven, and Marquez (2014), the relation between interest rates and bank risk-taking is likewise more complex than was thought beforehand, because in this paper evidence is brought forward that the capital structure of a bank is an important factor. In the paper, risk is measured as the inverse of bank monitoring of loans. A decrease in the real interest rate will lead to a higher amount of leverage and higher risk if the capital structure can be altered. When the capital structure is fixed and a bank is highly capitalized, the bank will increase risk as well. However, if the degree of competition is low, highly leveraged banks will decrease their risk. Therefore, it is concluded here again that the effect of interest rates on risk-taking varies across markets, different types of competitions and across time. Next, research on the relation between monetary policy and risk-taking will be discussed. Paligorova and Santos (2017) show that banks give discount to high risk bor-rowers in times of low market interest rates compared to periods with higher interest rates. The effect is not due to relaxing lending standards or a higher risk tolerance, but solely a result of macroeconomic factors. With the help of a survey among loan officers, this paper found that the previous result is prevalent in banks with a higher risk appetite. Therefore, the outcome is in line with the bank risk-taking channel of monetary policy and indicates that banks take on more risk in times of low market interest rates.
in relation to monetary policy changes. The results indicate that monetary policy changes have less effect on both highly capitalized banks and banks with market power. The second paper is written by Andries, Cocri¸s, and Ple¸sc˘au (2015), who executed a similar research with an entity fixed-effects model. An additional finding from this paper is that in the crisis period (2008-2011), the effect of monetary policy on bank risk-taking became more negative.
in the interest rate raises the value of assets and this affects the probability of default, loss given defaults and volatilities. As a result, risk perception goes down, risk budgets are decreased and therefore banks are likely to take on more risky investments.
The market interest rate conditions have been low for over ten years now. A few papers have examined if the relation between low interest rates and bank risk develops with a longer period of low interest rates. Maddaloni and Peydr´o (2011), a working paper of the ECB, did research on the so-called: ‘Too low for too long policy rates’. They argue that in the period leading up to the crisis the too low for too long rates have led to relaxing lending standards and increased risk on banks assets, which contributed to the financial crisis. An important finding here is that it is only the short-term interest rates and not the long-term interest rates that led to the relaxing of lending standards.
The financial stability review of the ECB (2007) examined credit risk in the period after banks increased their risky assets due to low interest rates. This paper calculated hazard rates of individual loans under different policy rates with an extensive database of the Bolivian banking market. According to the research of the ECB, banks soften their lending standards and lower the loan spread with lower interest rates. In the short run this leads to lower credit risk of bank loans as the volume of outstanding bank loans is bigger than the volume of the new loans. Banks will therefore be safer in the short run. However, after interest rates have been low for a long period, in the medium run interest rates will return to normal or even higher values. In this situation, these ‘new’ loans create a higher credit risk over their lifetime, which possibly cause problems for banks and in the end can have a tremendous effect on financial stability.
2.3
Hypotheses
With the insights from previous literature, this thesis comes up with the following research questions and its related hypotheses.
1. (a) Are market interest rates influencing the profitability of banks?
Hypotheses: Both a decrease in the market interest rates and a decrease in the yield curve slope decrease interest margins. However, this is offset by an increases of non-interest income and therefore return on assets is not affected. (b) What is the effect of an extension of the low interest rate period on bank
prof-itability?
Hypotheses: The longer the market interest rate remains below a certain thresh-old, the lower net interest margin and eventually return on assets will be due to the erosion of interest margins.
2. (a) Are market interest rates influencing the credit risk-taking of banks?
Hypotheses: Both a decrease in the market interest rates and a decrease in the yield curve slope,increase the risk attitude of banks in the search for yield and therefore banks increase their risk assets and decrease loan loss reserves. (b) What is the effect of an extension of the low interest rate period on the credit
risk-taking of banks?
3
Data, models and methodology
3.1
Dataset
US and euro area bank data is collected from Orbis BankFocus (Bureau van Dijk) for the period between 2003 and 2018. However, data from banks in the euro area are only available from 2013 onwards. Therefore, in the main examination an unbalanced panel dataset of the years 2013 to 2018 is used. This period was selected because the absence of euro area banks in the years before 2013 is not random, because they are not reported, and GMM works the best with a small amount of years. Table 1 shows the characteristics of the different data samples.
The dataset is filtered for commercial, savings, cooperative and Islamic banks. Investment banks are excluded from the sample, because they do not hold loans and de-posits. To mitigate double counting with consolidated statements, only banks with the consolidation codes C1, C2, and U1 are selected as described by Duprey and L´e (2016)2.
Banks that contain missing values in a year for a certain variable are excluded from the dataset for that particular year. Furthermore, the data is winsorized on the lowest and highest 0.01 percentile, as the dataset consists of several outliers. The dataset possibly contains a so-called “survivorship bias”, for the reason that not all banks exist over the
full time period due to failures, mergers and acquisitions. However, a balanced panel will contain only the healthiest banks and is therefore no solution. It is difficult to control for survivorship bias and it is probably negligible because of the large sample size. How-ever, estimates from a balanced and unbalanced sample will be compared to check for the presence of this bias. Delis and Kouretas (2011) and Gropp and Heider (2010) report a survivorship bias magnitude of approximately 10% with comparable datasets. As done by amongst others Borio, Gambacorta, and Hofmann (2017) and Bikker and Vervliet (2018), all level variables are divided by total assets, to make variables comparable and stationary. To test for stationarity, the Fisher unit root test with the Augmented Dickey Fuller (ADF) test is used without drift term or time trend. The Fisher test is used since this test handles unbalanced panel data with gaps. The test shows that all variables are stationary except the market concentration and importance of banks variable. The results can be found in table A.3 of the appendix. The data on interest rates and other macroeconomic variables will be collected from Eurostat and yearly averages are used for all variables.
3.2
Variables
3.2.1 Dependent variables
Two different profitability variables and two different risk-taking variables are used for robustness purposes. The following variables will be dependent variables in the bank profitability model:
(a) Net interest margin (NIM) is the difference between interest income and interest ex-penses as a ratio to interest-earning assets. This variable will reflect the profitability of balance sheet items, which is expected to be sensitive to interest rates changes. (b) Return on Assets (ROA) is calculated as net income divided by total assets. This
variable will reflect the total profitability of a bank and is used by other relevant papers. A higher value of this variable indicates a more profitable bank.
(a) Risk Weighted Assets (RWA) is calculated by dividing risk weighted assets over total assets. This variable includes all bank assets except safe assets such as cash, government securities and balances due from other banks. All bank assets that are sensitive to changes in credit quality through re-pricing opportunities and changes in market conditions are included. This variable is used by Delis and Kouretas (2011) and Bikker and Vervliet (2018) as well. A higher value of this variable indicates a riskier position of the bank.
(b) Loan loss reserves (LLR) is a measure of credit risk and expressed by the reserve for losses as a percentage of total loans. It measures how much the total portfolio has been provided for but not charged off. Given a consistent charge off-policy, the higher the ratio, the poorer the quality of the loan portfolio will be. More risky loans will result in a higher probability of default and higher expected loan losses which requires a higher provision level. Therefore, this variable serves as a proxy for credit risk-taking.
In contradiction to Bikker and Vervliet (2018) and Delis and Kouretas (2011), this thesis includes loan loss reserves instead of non-performing loans, because it is expected that this variable has a larger effect on profitability and therefore risk-taking as well. However, at the end of this thesis the variables non-performing loans and total capital ratio are used in a robustness analysis.
3.2.2 Interest rate variables
of the relative stability of Germany and the USA, and therefore they will mainly be influ-enced by monetary policy. The long-term interest rate includes a term premium, which is suppressed in both the USA and the euro area the last years due to asset purchases. In the robustness analysis, this thesis will control for the suppressed rates. Another interest rate variable that will be used is the difference between the short-term and long-term rate, which is the yield curve slope. It is well known that banks borrow short and lend long and therefore earn money on the difference between the long-term rate and the short-term rate.
Figure 1 and figure 2 show the evolution of the interest rates in the USA and euro area for the period 2013 to 2018. In the United States the federal fund rate started to rise in the year 2015 and as a consequence the yield curve slope decreased. In the euro area the EONIA turned negative from 2015 onwards, while the interest rate on the 10-year German bond decreased during the examined period. Unfortunately, there is limited variability of interest rates over the examined period. Especially the short-term rate in the euro area is barely changed. Therefore, it is questionable whether the changes in interest rate are sufficient to cause significant effects in the profitability and risk-taking of banks.
3.2.3 Bank control variables
on bank size and profitability is inconclusive. Papers such as Demirg¨u¸c-Kunt and Huizinga (1999) argue that large banks are more profitable due to economies of scale, because they have a higher degree of loan and product diversification. On the other hand, papers such as Tregenna (2009) suggest diseconomies of scale due to agency costs, large overhead costs and other costs that are related with managing large banks. In the risk-taking analysis, it is assumed that smaller banks take on relatively more credit risk, because smaller banks have less net worth, lower diversification and are less able to access liquidity (see e.g. Jim´enez, Ongena, Peydr´o et al. (2014)).
Liquidity is approximated as liquid assets over deposits and short-term fund-ing. Goddard, Molyneux, and Wilson (2004), Trujillo-Ponce (2013), Alshatti (2015) and Molyneux and Thornton (1992) suggest a negative relation between bank liquidity and profitability. On the other hand, Bourke (1989) suggests a positive relation. Bordeleau and Graham (2010) argue that profitability improves with more liquid assets until some point where it will deteriorate bank profitability. In the risk-taking model, the effect of liquidity on risk-taking is ambiguous. Khan, Scheule, and Wu (2014) argue that banks with higher liquidity levels are providing more credit at lower rates and therefore possess more risky loans on their balance sheet. Alternatively, Jim´enez, Ongena, Peydr´o et al. (2014) argue that banks with more liquidity lend out more safe loans.
Concentration is measured by the Herfindahl-Hirschman index and is calculated by the sum of the squared market shares of each bank, where market is defined by one country and market share as total assets in a bank over total assets in the market. The following equation is used:
HHIt= N X i−1 (PAi jAj )2 (1)
Listed is a dummy variable which takes the value 1 if the bank is publicly listed. This variable is unimportant when explaining bank profitability, according to Athana-soglou, Brissimis, and Delis (2008), Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009), Bourke (1989) and Molyneux and Thornton (1992). However, Dietrich and Wanzenried (2011) find evidence that banks listed on the stock exchange are less profitable compared to banks that are not listed. In the risk-taking analysis, publicly owned banks will take on more risk in the search for return to satisfy shareholders, according to Jim´enez, Ongena, Peydr´o et al. (2014).
Capitalization is a variable that indicates the ratio of capital in a bank by divid-ing total equity capital over total assets. It is expected that due to regulations, banks with higher capital levels possess higher levels of risky assets and therefore generate higher profit (Iannotta, Nocera, and Sironi (2007)). Berger (1995b) argues that sufficiently capitalized banks face lower bankruptcy costs which reduces their cost of funding and increases prof-itability. Furthermore, the paper argues that the increase of capital, is a signal from the management of good prospects. Goddard, Molyneux, and Wilson (2004) argue that high capital ratios could be the reason of overcautious banks that ignore potentially profitable opportunities and highly capitalized banks tend to generate lower profit. In the research of risk-taking, a higher amount of capital is associated with more prudent bank behavior and therefore less credit risk, according to Delis and Kouretas (2011).
Diversification is calculated by dividing non-interest income over total income. Non-interest income consists of fees and commissions and trading income. Higher amounts of non-interest income are associated with lower bank profitability, according to Demirg¨ u¸c-Kunt and Huizinga (1999), Stiroh (2004), and Kok, M´or´e, Pancaro et al. (2015). Opposite findings are from Dietrich and Wanzenried (2011), Valverde and Fern´andez (2007) and Elsas, Hackethal, and Holzh¨auser (2010), who argue that non-interest income yield higher margins, the so-called ‘diversification premium’. Gambacorta, Scatigna, and Yang (2014) argue that these benefits are present until a certain level of diversification. The effect of non-interest income on bank profitability remains ambiguous. In the risk-taking analysis, Bikker and Vervliet (2018) find that more diversified banks will have riskier assets.
(1994) argue that especially the ability in controlling costs, the so-called X-efficiency, is important. In the risk-taking analysis, Delis and Kouretas (2011) argue that technically more efficient banks will be more capable of managing risk and therefore contain more risk. The dependent variables Risk Weighted Assets (RWA) and Loan loss reserves (LLR) are used in the profitability model as independent variables. Higher credit risk exposure via loans is related to lower profit margins which is widely validated in the literature. Higher credit risk will lead to an increased amount of provisioning in the short-term, which will be deducted from net profit. In the medium term profit will be deteriorated as well due to actual loan losses. (Athanasoglou, Brissimis, and Delis, 2008; Brissimis and Delis, 2010; Kok, M´or´e, Pancaro et al., 2015)
The dependent variables Net Interest Margin (NIM) and Return on Assets (ROA) will be used in the risk-taking model as independent variables. Profit can be used in the next period to generate new loans according to Delis and Kouretas (2011) and Andries, Cocri¸s, and Ple¸sc˘au (2015). Delis and Kouretas (2011) find an insignificant effect of profit on risk assets and a positive effect on non-performing loans, indicating that profitable firms use profit from previous periods to fund qualitatively more risky projects instead of increasing risk assets.
3.2.4 Macro control variables
Inflation is the consumer price index (CPI) inflation in a country and reflects aspects of the business cycle as well. Demirg¨u¸c-Kunt and Huizinga (1999), Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009) and Trujillo-Ponce (2013) find a positive relation between inflation and bank profitability. Perry (1992) comes up with two theories. First, if banks have more nominal assets than nominal liabilities, rising prices would decrease the value of nominal assets more than the value of nominal liabilities, resulting in a loss. On the other hand, it depends whether inflation expectations are fully anticipated in the interest rates, in order to increase revenues faster than costs and increase profitability. In the risk-taking analysis, higher inflation before loan origination is expected to increase risk, while higher inflation during the lifetime of a loan reduces the real value of debt and therefore reduces default risk according to Jim´enez, Ongena, Peydr´o et al. (2014).
Importance of banks indicates whether the banks operates in a bank-based or market based economy. The variable is defined as the amount of domestic credit provided by the banking sector as a share of GDP. There is no literature on the effect of bank importance on bank profitability. In the risk-taking analysis it is expected that banks in a bank-based economy should take on more risk to satisfy the inelastic demand for credit according to Delis and Kouretas (2011).
3.2.5 Descriptive statistics
Table 2 presents the descriptive statistics, table A.1 gives the source and de-scription of every model variable, table A.2 presents the Fisher unit root test, table A.3 presents the characteristics of the dataset per country, table A.4 presents the descriptive statistics of the variables in the euro area and the USA separately and table A.5 provides a correlation matrix.
from the USA have the longest average number of observations, namely 5.27 years.
3.3
Models
The following model will be used to analyze profitability of banks:
πit = c + απi,t−1+ β1IRjt+ β2Xit+ β3Zit+ it (2)
Where πit is the profitability variable for bank i in year t. According to Berger,
Bonime, Covitz et al. (2000), bank profitability is persistent over time and therefore a dynamic model is used in this thesis. In equation 2, πi,t−1 is the lagged profitability
0, there is a high speed of adjustment present. If α is between 0 and 1, it implies that profit persists but eventually will return to its normal level. If α is close to 1, persistence of bank profitability is strong and the speed of adjustment is low. The lagged profitability variable will affect the interpretation of the coefficients, for the reason that if α is close to 1, the coefficient of other independent variables will explain the change in profitability3. This model is based on the idea of Bikker and Vervliet (2018) and Athanasoglou, Brissimis, and Delis (2008). The interest-rate variable is IRit for country j in year t and the three different
interest rate variables (ST rate, LT rate and Yield curve) will be used separately in three different estimations for collinearity reasons. Xit is a set of bank specific control variables
and Zit is a set of macro control variables. The error term is given by it= ηi+ uit where
ηi is the unobserved bank-specific effect, which is time-invariant, and uit the idiosyncratic
error. The model will be estimated with two different profitability dependent variables (NIM and ROA) as πit. The coefficient β1 is the key coefficient and indicates the relation
between interest rates and bank profitability. A positive β1 should be interpreted as lower
interest rates decrease bank profitability and a negative β1 should be interpreted as an
opposite relation. The size of β1 is the amount of the profitability ratio (NIM and ROA)
changes with one percentage point increase of the used interest rate. If β1 is statistically
not different from zero, this would imply that no relation between interest rates and bank profitability can be found. The following model will be used to analyze risk-taking of banks:
rit = c + αri,t−1+ β1IRjt+ β2Xit+ β3Zit+ it (3)
Where rit is the risk variable for bank i in year t. This is a dynamic model that
builds on the work of Delis and Kouretas (2011) and Bikker and Vervliet (2018). Delis and Kouretas (2011) argue that risk exposure is carried over to the next period in the form of trading assets or loans portfolios and therefore risk is persistent. The degree of persistence is captured by α and ri,t−1 is the lagged risk variable. The other variables and the error
term are similar to those in the profitability model. The model will be estimated with two different dependent risk-taking variables (RWA and LLR) as rit. The coefficient β1 is the
key coefficient and indicates the relation between interest rates and bank risk-taking. A negative β1 should be interpreted as lower interest rates increase credit risk-taking with
banks and a positive β1 should be interpreted as an opposite relation. The size of β1 is
3If α is 1, the equation will be the following: π
the amount of the risk-taking ratio (RWA and LLR) changes with one percentage point increase of the used interest rate. If β1 is statistically not different from zero, this would
imply that there is no relation between interest rates and bank risk-taking.
3.4
Methodology
The models in this thesis have a dynamic structure as a result of persistence in bank profitability and bank risk-taking. Previous research has shown the dynamic nature of these variables, e.g. Athanasoglou, Brissimis, and Delis (2008), Delis and Kouretas (2011), Dietrich and Wanzenried (2011), Trujillo-Ponce (2013) and Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009). Berger, Bonime, Covitz et al. (2000) argue that bank profitability is persistent due to market power following from impediments in product market competition and information opacity. Delis and Kouretas (2011) give several reasons for persistence of bank-risk, which are: intensive competition which enhances risk-taking, relationship-banking with risky borrowers, banks require time to smooth the effects of macroeconomic shocks and finally regulations such as deposit guarantees and capital requirements enhance moral hazard, leading to inefficient and risky investments over a substantial amount of time. Furthermore, they argue that the effect of stock variables on flow variables is preferably approximated by a dynamic model, because a static model will be biased.
that remain constant over time and do affect the profitability and risk-taking of banks. Finally, the dynamic nature of the models complicates the estimation approach. In dy-namic environments, the pooled OLS estimator, the first-difference estimator or the within estimator have proven to be insufficient, because of the dynamic panel bias according to Nickell (1981).4 The lagged dependent variable is in large samples of the within estimate biased downwards and the pooled OLS estimate is biased upwards as shown by standard results in Bond (2002).
The system Generalized methods of moments (GMM) estimator introduced by Arellano and Bover (1995) and Blundell and Bond (1998) is suitable for dynamic panel models with the issues of dynamic panel data, unobserved heterogeneity and endogeneity. The GMM estimator makes use of lagged values of the variables in levels and in differences as instrument for the dependent variables and other endogenous variables. The GMM estimator controls for the problems of endogeneity, unobserved heterogeneity and persis-tence of the dependent variables. The GMM estimator is widely used in related empirical studies in the field of bank profitability and bank risk-taking as discussed in the litera-ture review. The GMM estimator is amongst others suitable for panel data with a small amount of periods (T) and a large amount of panel members (N), which is the case in this thesis.5 GMM tackles the fixed effects problem by taking the first difference of regressors
as instruments and with this approach the time-invariant effects are eliminated.
In the case of an unbalanced panel dataset, this first difference approach amplifies the gaps in the dataset. If an observation is missing, both the difference of this observation with the previous and the difference with the next observation will be missing in the GMM estimation. A solution to this problem is the forward orthogonal deviations (FOD) trans-formation introduced by Arellano and Bover (1995). This transtrans-formation subtracts the average of all available future transformations from the current value and drops the last observation for each individual instead of dropping the first observation as is done by the first-difference transformation, The system GMM estimator is proposed as more efficient compared to the difference GMM estimator. Compared to the difference GMM estimator
4The dynamic panel bias implies that the lagged dependent variable and the error term are correlated which causes inconsistent estimates.
it adds a set of equations in levels to the system with suitable lagged first difference as instruments (Roodman, 2009a; Bond, Hoeffler, and Temple, 2001). A technique commonly used with the GMM estimation is the Windmeijer finite sample correction (Windmeijer, 2005), for the reason that the poorly estimated weighting matrix produces downward bi-ased outcomes. Furthermore, Windmeijer (2005) found that the two-step GMM estimator produces less biases, lower standard errors and improves the estimation of the coefficients compared to the one-step GMM estimator. For this reason, this thesis uses the finite-sample correction in the two-step system GMM estimator in both the profitability and the risk-taking model. The amount of instruments has a substantial effect on the valid-ity of instruments.6 There is no clear consensus on how many instruments is too many, because even with a few instruments there will be bias present. In order to decrease the instrument count and improve the Hansen test, the collapse technique is implemented in all estimations to reduce the amount of instruments to avoid overfitting and reducing the bias. This technique collapse the instrument matrix into a single column which discloses less information however represents the same information.
The choice whether a variable is endogenous or exogenous is an important input in the estimation. If a variable is determined inside the model by other model variables, it is an endogenous variable and on the other hand if a variable is predetermined, it is an exogenous variable. Endogenous variables are treated as ‘GMM-style’ instruments while exogenous variables are treated as standard instruments (Holtz-Eakin, Newey, and Rosen, 1988; Arellano and Bond, 1991). Almost all bank-specific variables are considered endogenously and the macro and interest rate variables will be treated as exogenous, comparable to related empirical literature. Interest rates are not expected to be influenced by bank-level risk or profitability, because there is no evidence that the ECB or Federal Reserve sets policy rates in correspondence to bank profitability or bank risk-taking. The variable bank size is assumed to be endogenous for the reason that profitability, efficiency and risk-taking will impact bank size in terms of assets. The concentration variable is endogenous in the model of Garc´ıa-Herrero, Gavil´a, and Santab´arbara (2009), while it is exogenous in the model of Trujillo-Ponce (2013). The variable concentration will be taken as exogenous, because it is expected that bank profitability or bank risk-taking is not able to influence concentration in a market. Finally, it is assumed that being listed on the stock exchange is predetermined and therefore the variable listed is treated as exogenous.
4
Empirical results
4.1
Bank profitability
Figure 3 presents the evolution of the average net interest margin and shows a large difference between the net interest margin in the USA and the euro area. The net interest margin is approximately twice as high in the USA over the period 2013 to 2018 and the euro area net interest margin is decreasing over the period. On the other side, figure 4 presenting the evolution of average return on assets for banks shows an upward trend in the USA in the same period and a steady level in the euro area. The graphs give a first indication that non-interest banking activities compensate for decreasing interest margins or additionally increase profitability levels. The gap between the USA and the euro area is amplified in figure 4 as average return on assets is roughly three times as high in the USA compared to the euro area. There is hardly any variation over time of the average net interest margins and return on assets. This could lead to identification problems if interest rates are unable to explain variation in profitability variables when there is barely variation in the examined period. However, the descriptive statistics show that both variables have a larger standard deviation between banks.
variable (short-term rate, long-term rate and yield curve slope). The Wald test indicates a fine goodness of fit for the model. The Arellano-Bond (AR) test on autocorrelation of the first and second order is presented. Failure to reject the null hypothesis in this test indicates no autocorrelation. The AR test shows the presence of first-order auto correlation with dependent variable NIM. However, according to Arellano and Bond (1991), inconsistency would be implied if second-order auto correlation would be present, which is the case with the regressions on dependent variable ROA. A solution to this problem would be to add higher order lags of the dependent variable to the estimation, however this does not significantly improve the AR(2) test results.
The highly significant coefficients of the lagged profitability variables show that profitability is persistent as expected. It can be observed that net interest margin is more persistent than return on assets. This is because it takes time to adjust interest rates on outstanding loans and deposits, while profitability of nontraditional banking activities has a higher speed of adjustment. The EONIA and the federal funds rate have a significant effect on return on assets and surprisingly not on the net interest margin. However, the 10-year government bond has a large and significant effect on both net interest margin and return on assets. The effect of the interest rates is larger on return on assets. There is a positive relation between interest rates and provisioning, as is observed in the next section on bank risk-taking. This is explained by Altavilla, Boucinha, and Peydr´o (2018) as follows: An increases (decrease) of interest rates increases (decreases) the probability of default and loss given default7 on loans and therefore increase (decrease) the cost of
loan loss provisioning which decreases (increases) profit levels. Apparently this negative effect is not large enough to overcome the positive effect of interest rates on net interest margins and potentially other banking activities, resulting in a positive net effect on return on assets. Therefore, it is concluded that interest rates affect bank profitability positively, confirming studies such as Claessens, Coleman, and Donnelly (2018), Borio, Gambacorta, and Hofmann (2017) and Cruz-Garc´ıa, de Guevara, and Maudos (2019).
The effect of long-term interest rates on profitability is relatively larger compared to short-term interest rates. The yield curve slope, defined as the difference between the short-term rate and the long-term rate has a significant effect on net interest margin and return on assets as well. This is in agreement with Alessandri and Nelson (2015), who found a positive effect of the yield curve slope on net interest margins and profit levels. The yield curve slope increases return on assets via the net interest margin for the reason that banks earn interest income on the difference between long-term assets and short-term liabilities.
Next, the results of the independent variables will be discussed. The coefficients of more or less all variables behave as the theory predicted. Only the variables capitalization, efficiency, risk weighted assets and loan loss reserves are consistently significant. The amount of loan loss reserves as a proxy for credit risk has a negative relation with net interest margins and return on assets. The theory explains that higher credit risk leads in the short run to more provisioning or reserves which is deducted from net profit. The
coefficients of macro variables GDP growth, inflation and bank importance do not show a uniform, convincing or significant results.
4.2
Bank risk-taking
Figure 5 presents the evolution of the risk weighted asset intensity in the period 2013 to 2018. The graph shows a rather stable average RWA level for the euro area, while the average RWA level in the USA increased approximately with 5 percentage points. This indicates that banks in the USA obtained relatively more risky assets in this period. Noteworthy is the difference between banks in the USA and the euro area. Apparently banks in the US possess relatively more risky assets compared to euro area banks. Figure 6 presents the evolution of the average loan loss reserves in the same period. Evidently, there is a downward trend observable in the graph for both the USA and the euro area. Again, USA banks are more risky by having less loan loss reserves to the gross amount of loans, however the graph suggest that loan loss reserves levels converged in the last years.
test of over-identifying restrictions is worrisome here as well. All tests have a p-value of 0.000 and therefore reject the null hypothesis that all over identifying restrictions are jointly valid. Likewise, the coefficients and standard errors should be interpreted with care here as well, because both will be biased. Relationships can be derived out of Table 4, however no causations can be determined and no numerical weight can be placed on the relationships.
The lagged risk-taking variables are highly significant for both dependent vari-ables. The persistence of risk assets is rather low, indicating a slow rate of adjustment for risk weighted asset intensity. This is contradicting with the paper of Delis and Kouretas (2011), since they find higher values for lagged risk assets and this could be the result of invalid instruments. The persistence of loan loss reserves is above 1. This is an inconsistent finding, because it is found that LLR is stationary and in figure 6 it is downward sloping over the estimated period. Furthermore, the first-differences of loan loss reserves are plot-ted over time and it does not show evidence of explosive series.8 The short-term interest
rate coefficient is highly significant and surprisingly positive for RWA. The coefficient of the long-term interest rate is significant as well and substantially higher. These findings are contradicting to previous literature and it is expected that these results are caused by an endogeneity problem, which implies that the interest rate variable correlates with the error term. Although this is what GMM attempts to solve, in practice this is not always the case. In the robustness analysis other control variables will be included in an attempt to solve the endogeneity problem. Moving on to the next variable, if the yield curve slope decreases, provisioning and risk asset intensity increases. This is evidence of increased credit risk-taking in times of converged interest rates. The short-term and long-term rate coefficients in the LLR regression are positive and significant. These results suggest that banks increase their reserves with higher interest rates. As mentioned earlier in this thesis, the reason behind this is that higher interest rates increase the probability of default and loss given default of loans and therefore more reserves are necessary. The slope of the yield curve drops with lower economic growth, raising the needs for higher reserves, which is predicted by the negative and significant coefficient of the yield curve slope variable.
Ozili and Outa (2017) come up with several hypotheses that explain how banks use their loan loss reserves. First, the capital management hypothesis explains that banks use provisions to keep minimum capital requirements. If LLR is included in the compu-tation of minimum regulatory capital ratios, banks have the incentive to adjust the level of LLR to meet the minimum requirements. Second, the signaling hypothesis explains the use of LLR as a signal of information about the loan portfolio or future earnings. Third, several papers have provided evidence for income smoothing of banks with the use of LLR. The motives to do so are to meet objectives, to improve the stability of the bank or to op-portunistically receive a bonus. In times of lower interest rates and therefore lower profit, there is more incentive smoothen income by decreasing LLR. Fourth, the procyclicality hy-pothesis of LLR implies that banks decrease provisioning in times of high economic growth (low interest rates) and increase provisioning in economic downturns (high interest rates), because higher losses are expected in economic downturns.
Next, the coefficient of the independent variables will be discussed. Smaller banks, banks in a concentrated market and publicly listed banks will hold relatively more risk weighted assets according to the results. Banks that have higher net interest margins contain more risk weighted assets compared to banks with lower net interest margins. Delis and Kouretas (2011) argue that profitable firms do not increase their risk assets, however they fund qualitatively more risky projects. Banks with higher return on assets have relatively lower loan loss reserves. Furthermore, evidence is found that higher GDP growth decreases the relative amount of risk weighted assets and loan loss reserves. Finally, with high inflation levels and in countries where banks provide more credit to the economy, banks will possess relatively less risk weighted assets.
4.3
Robustness analysis
4.3.1 Pooled OLS and within estimator
with the use of the pooled OLS and the within estimator.9 Table 5 and table 6 present
the results of the pooled OLS and within estimator of the profitability model and risk-taking model respectively. The table shows that the coefficients of the lagged dependent variable are within the credible range of the pooled OLS and the within estimator. The profitability model estimations of the pooled OLS and the within estimator do not show a lot of differences compared to the GMM estimations. The main differences are the different sizes of the lagged dependent variables and the effect of the short-term rate on profitability is smaller with the alternative estimation methods. There are considerably more differences between the system GMM, pooled OLS and WE in the risk-taking model. Coefficients become significant or insignificant or turn sign in the pooled OLS or within estimator estimations. The main difference is the significant negative coefficient of the short-term rate with risk weighted assets as dependent variable in the Pooled OLS estimation.
4.3.2 Low for long interest rates
4.3.3 Other robustness analysis
This section will introduce a few complementary control variables in an attempt to solve the possible endogeneity problem. First, Delis and Kouretas (2011) argue that bank risk is driven by country specific regulatory conditions and therefore they control for this factor when estimating the relation of interest rates on bank risk. In this robustness analysis the same three indices, namely the capital regulatory index, the supervisory power index and the market discipline index constructed from the World Bank database of Barth and Levine (2001); Barth, Caprio, Levine et al. (2006); Barth, Caprio, and Levine (2008) are used. The exact structure of the indices can be found in appendix table A.8 which is extracted from Delis and Kouretas (2011). Next, expected macroeconomic conditions are included as control variables, inspired by the idea of Altavilla, Boucinha, and Peydr´o (2018). This includes the forecast of GDP growth and the forecast of CPI inflation for the next year and they are obtained from the OECD database. The forecasted GDP growth and the forecast of CPI inflation were not available for the countries Cyprus and Malta and therefore these countries drop out of the sample. The results are presented in the first column of table A.9 and table A.10 in the appendix. The coefficients of the lagged dependent variables are similar to the main results in both models. However, the yield curve slope appears to have a negative and significant effect on return on assets with the additional control variables. Furthermore, the negative relation between the yield curve slope and risk weighted asset is insignificant with the new controls. Therefore, these results show that the main results are not robust for different control variables.
Next, to control for the suppressed term premiums of the long-term interest rates, average total central bank assets in millions of the particular currency are included as a control variable in the estimations. This variable serves as a proxy of the central bank’s purchase programs and controls for the suppressed term premium of the long-term interest rate. The results can be found in the second column of table A.9 and table A.10 in the appendix. The results show an increased positive relation between the yield curve slope and return on assets. Furthermore, the coefficient of the yield curve slope in the risk-taking model is contradicting to the main results, because it is positive but insignificant.
show substantial differences compared to the previous results and therefore it is concluded that the results are robust to the survivorship bias.
Other papers have used different measures of bank profit and credit risk. To check the robustness of the dependent variables in both models, the same models with two different dependent variables are estimated. In the profitability model, profit before tax over total assets (Profit) and return on equity (ROE) will be used. In the risk-taking model, total capital ratio (TCR) and non-performing loans ratio (NPL) will be used. Total capital ratio is defined as the total capital over risk weighted assets. Non-performing loans ratio is defined as non-performing loans divided by total loans. The results of these regressions can be found in the fourth and fifth column of table A.9 and table A.10 in the appendix. Return on equity is not persistent, however an increase in the yield curve slope increases return on equity. On the other hand, the yield curve slope does not have a relation with profit before tax. This could be the result of higher interest rates that creates valuation gains or losses that show up in profit and not in return on assets. Furthermore, no evidence is found of a relation between the yield curve slope and total capital ratio. The yield curve slope affects non-performing loan ratio negatively, however this result is only slightly significant.
5
Conclusion
In this thesis the effect of the low interest rate environment on the profitability of banks is examined. The hypothesis is that lower interest rates and a suppressed yield curve put pressure on the interest margins of banks. However, provisions decreases and non-interest income increases in this situation resulting in higher profit levels. The net effect on profitability is examined in this thesis. Furthermore, the effect of low interest rates on the credit risk-taking of banks is investigated. The hypothesis here is that in case of lower profit margins, banks would increase their risk appetite in the search for yield. The research of two dynamic models is executed with the use of the system GMM estimator as econometrical method and a dataset consisting out of more than 9,000 banks from the USA and the euro area.
The research results on bank profitability show that a lower long-term interest rate and a smaller yield curve slope decreases interest margins. Banks borrow short and lend long and therefore earn money on the difference between the short-term deposit rates and the long-term loan rates. Therefore, if the difference between both rates is smaller, interest income will be lower. Lower interest rates decrease provisioning and creates valuation gains which increases profit levels. Nevertheless, the decrease in interest margins is strong enough to overcome the profit increasing effect of provisioning, resulting in a negative return on asset as net results. The results are robust for different estimation methods and a longer time frame. This thesis concludes that interest rates and the yield curve slope have a positive relation with return on assets, for the reason that lower interest rates and yield curve slope decrease interest margins and therefore interest income.
In the examination of the risk-taking model, the results show that if the yield curves slope decreases, banks start to increase risk-taking. Therefore, if the difference between the short-term and the long-term interest rate, where banks earn money on, decreases and therefore interest margins diminish, the risk appetite of banks increases in the search for yield. On the other side, an individual decrease of the short-term or long-term interest rate would surprisingly decrease risk-taking. However, the relation of risk-taking with the yield curve slope is not robust to a larger time frame, different estimation methods and different dependent and control variables. Therefore, this thesis cannot conclude that the yield curve slope has a negative relation with risk weighted asset intensity.
rates reduces bank profit. This is because banks can alter deposit rates immediately for most of the deposits, while loans need more time to incorporate the lower rates in the loan portfolio. Furthermore, this thesis finds that every additional year of short-term interest rates below the threshold of 0.5% increases credit risk significantly. The reason for this is that with every additional year of profit erosion, banks have more incentive to increase their credit risk in the search for yield.
For robustness purposes multiple profitability and risk-taking dependent variables and multiple interest rates as explanatory variables are used in the estimations. Other robustness analyses are the estimations of a balanced sample and a longer time frame, which is 16 years instead of 6 years, and the inclusion of several additional control variables that relate to macro forecasts, banking supervision and asset purchase programs of central banks.
There are a few limitations to the GMM estimation method. First of all, the estimations with dependent risk variables potentially suffer from an endogeneity problem. Secondly, GMM results are unstable in case of invalid instruments. None of the estimations appears to have valid instruments which results in biased results and therefore the GMM results should be interpreted carefully. Other limitations of this study are the short time frame of 6 years that is examined and the second-order autocorrelation in the estimation of return on assets. Furthermore, GMM is highly sensitive to specification changes such as the choice whether a variable is treated endogenous or exogenous. Another potential problem is the imbalance of banks from the USA and the euro area in the sample. There are four times as much banks from the USA compared to the euro area.
References
Albertazzi, U., Gambacorta, L., 2009. Bank profitability and the business cycle. Journal of Financial Stability, 5(4):393–409.
Alessandri, P., Nelson, B.D., 2015. Simple banking: profitability and the yield curve. Journal of Money, Credit and Banking, 47(1):143–175.
Alshatti, A.S., 2015. The effect of the liquidity management on profitability in the jorda-nian commercial banks. International Journal of Business and Management, 10(1):62. Altavilla, C., Boucinha, M., Peydr´o, J.L., 2018. Monetary policy and bank profitability in
a low interest rate environment. Economic Policy, 33(96):531–586.
Altunbas, Y., Gambacorta, L., Marques-Ibanez, D., 2010. Does monetary policy affect bank risk-taking?
Andries, A.M., Cocri¸s, V., Ple¸sc˘au, I., 2015. Low interest rates and bank risk-taking: Has the crisis changed anything? evidence from the eurozone. Review of Economic and Business Studies, 8(1):125–148.
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte carlo evidence and an application to employment equations. The review of economic studies, 58(2):277–297.
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. Journal of econometrics, 68(1):29–51.
Athanasoglou, P.P., Brissimis, S.N., Delis, M.D., 2008. Bank-specific, industry-specific and macroeconomic determinants of bank profitability. Journal of international financial Markets, Institutions and Money, 18(2):121–136.
Barth, J.R., Caprio, G.J., Levine, R., 2008. Bank regulations are changing: for better or worse? The World Bank.
Barth, J.R., Caprio, G., Levine, R., et al., 2006. Rethinking bank regulation. Till angels govern.
Berger, A.N., 1995a. The profit-structure relationship in banking–tests of market-power and efficient-structure hypotheses. Journal of Money, Credit and Banking, 27(2):404– 431.
Berger, A.N., 1995b. The relationship between capital and earnings in banking. Journal of money, credit and Banking, 27(2):432–456.
Berger, A.N., Bonime, S.D., Covitz, D.M., Hancock, D., 2000. Why are bank profits so persistent? the roles of product market competition, informational opacity, and region-al/macroeconomic shocks. Journal of Banking & Finance, 24(7):1203–1235.
Berger, A.N., Humphrey, D.B., et al., 1994. Bank scale economies, mergers, concentration, and efficiency: The us experience.
Bikker, J.A., Hu, H., 2002. Cyclical patterns in profits, provisioning and lending of banks and procyclicality of the new basel capital requirements. PSL Quarterly Review, 55(221). Bikker, J.A., Vervliet, T.M., 2018. Bank profitability and risk-taking under low interest
rates. International Journal of Finance & Economics, 23(1):3–18.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of econometrics, 87(1):115–143.
Bolt, W., De Haan, L., Hoeberichts, M., Van Oordt, M.R., Swank, J., 2012. Bank prof-itability during recessions. Journal of Banking & Finance, 36(9):2552–2564.
Bond, S.R., 2002. Dynamic panel data models: a guide to micro data methods and practice. Portuguese economic journal, 1(2):141–162.
Bond, S.R., Hoeffler, A., Temple, J.R., 2001. Gmm estimation of empirical growth models. Bordeleau, ´E., Graham, C., 2010. The impact of liquidity on bank profitability. Technical
report, Bank of Canada working paper.
Borio, C., Gambacorta, L., Hofmann, B., 2017. The influence of monetary policy on bank profitability. International Finance, 20(1):48–63.
Bourke, P., 1989. Concentration and other determinants of bank profitability in europe, north america and australia. Journal of Banking & Finance, 13(1):65–79.
Brissimis, S.N., Delis, M.D., 2010. Bank heterogeneity and monetary policy transmission. Claessens, S., Coleman, N., Donnelly, M., 2018. “low-for-long” interest rates and banks’ interest margins and profitability: Cross-country evidence. Journal of Financial Inter-mediation, 35:1–16.
Cruz-Garc´ıa, P., de Guevara, J.F., Maudos, J., 2019. Determinants of bank’s interest margin in the aftermath of the crisis: the effect of interest rates and the yield curve slope. Empirical Economics, 56(1):341–365.
De Nicol`o, G., Dell’Ariccia, G., Laeven, L., Valencia, F., 2010. Monetary policy and bank risk taking. Available at SSRN 1654582.
Delis, M.D., Kouretas, G.P., 2011. Interest rates and bank risk-taking. Journal of Banking & Finance, 35(4):840–855.
Dell’Ariccia, G., Laeven, L., Marquez, R., 2014. Real interest rates, leverage, and bank risk-taking. Journal of Economic Theory, 149:65–99.
Dell’Ariccia, G., Marquez, R., 2006. Lending booms and lending standards. The Journal of Finance, 61(5):2511–2546.
Demirg¨u¸c-Kunt, A., Huizinga, H., 1999. Determinants of commercial bank interest margins and profitability: some international evidence. The World Bank Economic Review, 13(2):379–408.
Dietrich, A., Wanzenried, G., 2011. Determinants of bank profitability before and dur-ing the crisis: Evidence from switzerland. Journal of International Financial Markets, Institutions and Money, 21(3):307–327.
Duprey, T., L´e, M., 2016. Bankscope dataset: getting started. Available at SSRN 2191449. ECB, 2007. The impact of short-term interest rates on bank credit risk-taking. Financial
Stability review December 2007.
Gambacorta, L., 2009. Monetary policy and the risk-taking channel. BIS Quarterly Review December.
Gambacorta, L., Scatigna, M., Yang, J., 2014. Diversification and bank profitability: a nonlinear approach. Applied Economics Letters, 21(6):438–441.
Garc´ıa-Herrero, A., Gavil´a, S., Santab´arbara, D., 2009. What explains the low profitability of chinese banks? Journal of Banking & Finance, 33(11):2080–2092.
Goddard, J., Molyneux, P., Wilson, J.O., 2004. Dynamics of growth and profitability in banking. Journal of Money, Credit and Banking, pages 1069–1090.
Gropp, R., Heider, F., 2010. The determinants of bank capital structure. Review of finance, 14(4):587–622.
Holtz-Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with panel data. Econometrica: Journal of the Econometric Society, pages 1371–1395. Iannotta, G., Nocera, G., Sironi, A., 2007. Ownership structure, risk and performance in
the european banking industry. Journal of Banking & Finance, 31(7):2127–2149. Ioannidou, V.P., Ongena, S., Peydr´o-Alcalde, J.L., 2008. Monetary policy, risk-taking, and
pricing: Evidence from a quasi-natural experiment. Tilburg University.
Jim´enez, G., Ongena, S., Peydr´o, J.L., Saurina, J., 2014. Hazardous times for monetary policy: What do twenty-three million bank loans say about the effects of monetary policy on credit risk-taking? Econometrica, 82(2):463–505.
Jobst, A., Lin, H., 2016. Negative interest rate policy (NIRP): implications for monetary transmission and bank profitability in the euro area. International Monetary Fund. Khan, M.S., Scheule, H.H., Wu, E., 2014. The impact of bank liquidity on bank risk taking:
Do high capital buffers and big banks help or hinder? In 27th Australasian Finance and Banking Conference.
Lopez, J.A., Rose, A.K., Spiegel, M.M., 2018. Why have negative nominal interest rates had such a small effect on bank performance? cross country evidence. Technical report, National Bureau of Economic Research.
Maddaloni, A., Peydr´o, J.L., 2011. Bank risk-taking, securitization, supervision, and low interest rates: Evidence from the euro-area and the us lending standards. the review of financial studies, 24(6):2121–2165.
Manganelli, S., Wolswijk, G., 2009. What drives spreads in the euro area government bond market? Economic Policy, 24(58):191–240.
Marcucci, J., Quagliariello, M., 2009. Asymmetric effects of the business cycle on bank credit risk. Journal of Banking & Finance, 33(9):1624–1635.
Molyneux, P., Thornton, J., 1992. Determinants of european bank profitability: A note. Journal of banking & Finance, 16(6):1173–1178.
Nickell, S., 1981. Biases in dynamic models with fixed effects. Econometrica: Journal of the Econometric Society, pages 1417–1426.
Ozili, P.K., Outa, E., 2017. Bank loan loss provisions research: A review. Borsa Istanbul Review, 17(3):144–163.
Paligorova, T., Santos, J.A., 2017. Monetary policy and bank risk-taking: Evidence from the corporate loan market. Journal of Financial Intermediation, 30:35–49.
Perry, P., 1992. Do banks gain or lose from inflation? Journal of Retail Banking, 14(2):25– 31.
Rajan, R.G., 2006. Has finance made the world riskier? European Financial Management, 12(4):499–533.
Roodman, D., 2009a. How to do xtabond2: An introduction to difference and system gmm in stata. The stata journal, 9(1):86–136.
Roodman, D., 2009b. A note on the theme of too many instruments. Oxford Bulletin of Economics and statistics, 71(1):135–158.
Tregenna, F., 2009. The fat years: the structure and profitability of the us banking sector in the pre-crisis period. Cambridge Journal of Economics, 33(4):609–632.
Trujillo-Ponce, A., 2013. What determines the profitability of banks? evidence from spain. Accounting & Finance, 53(2):561–586.
Valverde, S.C., Fern´andez, F.R., 2007. The determinants of bank margins in european banking. Journal of Banking & Finance, 31(7):2043–2063.
