Bank Risk-taking in Negative Interest Rate
Environments: Evidence from the Euro Area
Name: Imane Ait-ha Student number: 11419954
Program: EBE Specialization: Finance Supervisor: Dr. Mihnea Constantinescu
Statement of Originality
This document is written by Imane Ait-ha who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract:
In June 2014 the ECB introduced negative policy rates which instigated a debate about its possible consequences on the monetary transmission mechanism and the banking sector more specifically. This paper uses a dataset of 3752 banks operating in 19 countries in the Euro Area between 2012-2016 to estimate the impact of the introduction of negative interest rates on banks’ lending behavior and risk-taking. Using a difference-in-differences framework, we find that banks with high-deposit ratio increase lending in response to negative interest rates in comparison with banks with low-deposit ratio. Moreover, high-deposit banks also engage in more risk-taking in negative rates environments.
Keywords:
TABLE OF CONTENTS
1. Introduction………..5
2. Literature Review……….6
2.1 Monetary Transmission Mechanism………..6
2.2 Risk-Taking Channel……….7
2.3 Bank Risk-Taking under Low-Interest Rates………...9
2.4 Unconventional Monetary policy………10
2.5 Negative Interest Rates in the EU………12
2.6 Negative Interest Rates and Bank Risk-Taking………...12
2.7 Hypothesis Development……….14
3. Data Description & Methodology………..…15
3.1 Baseline Model & Choice of Variables………...15
3.2 Data……….18
3.3 Summary statistics………...19
4. Results ………...20
4.1 Impact of NIRP on lending volume………..…...20
4.2 Impact of NIRP on risk-taking……….22
5. Discussion………..24 5.1 Interpretation of results………....24 5.2 Policy implications………...25 5.3 Limitations………...26 6. Conclusion……….26 7. References………..27 8. Appendix………....31
1. INTRODUCTION
The 2000s were characterized by an extended period of low levels of interest rates and loosely regulated banking systems combined with significant growth of credit. These factors heavily influenced the unfolding of the 2008 crisis (Angeloni et al, 2015). Since the 2007-2008 Global Financial Crisis, various central banks have implemented a variety of unconventional monetary policies such as large scale asset purchases or what is known as quantitative easing, forward guidance, and negative interest rate policy NIRP to address low inflation and stimulate economic growth. Since June 2014, the European central bank ECB progressively implemented the NIRP (Arteta et al, 2018).
The intention behind the implementation of this policy is according to Cœuré (2016), member of the executive board of the ECB; “to ensure that sufficient stimulus is provided to the economy to return inflation to the ECB’s objective” . This came after persistent below-target inflation expectations and disappointing growth rates. That period was also characterized by excess liquidity in the banking system. Therefore, i mposing negative rates on banks’ excess deposits held at the central bank was in fact a form of taxing these banks for holding cash (Arteta et al, 2018). This unusual, unconventional monetary policy brought up concerns about potential consequences on banks' profitability and risk-taking behaviors amongst the recent growing literature. First, the introduction of negative rates could impede the transmission of monetary policy by reducing banks' interest incomes and thus banks’ profitability and its equity’s worth (Borio & Gambacorta, 2017). Second, as a reaction to these consequences on profitability, banks could pursue more profitable securities in search of higher yield or increase risk-taking and adjust their lending behavior accordingly (Bubeck et al, 2020).
In this paper, the extent to which bank lending behavior changes under negative interest rates will be tested. A cross-sectional sample of panel data will be utilized to investigate the effect of NIRP on bank-risk taking and lending behavior by using the Difference-in-Differences framework. In the second section an intensive literature review explaining the monetary transmission mechanism, Unconventional monetary policy, and reviewing evidence of risk-taking behavior under low-interest environments and under NIRP. This section will be concluded by formulating a hypothesis. In the third section, the empirical specification and
description of data will follow. In section 4 results will be presented. Finally, section 5 & 6 will include the interpretation of our results along with some concluding remarks.
2. LITERATURE REVIEW
2.1 Monetary Transmission Mechanism
The monetary transmission mechanism outlines the impact of monetary policy decisions such as changes in money supply or short-term interest rates on general economic conditions and its influence on variables such as aggregate output and employment. According to Ireland (2002), there exist two key assumptions for monetary policy changes to affect variables outside the limits of the central bank’s balance sheet. Firstly, agents are not capable of abolishing monetary policy generated changes. That is agents do not issue private securities which can be perfect substitutes for elements of the monetary base. Second, we assume the presence of some friction in the economy that prohibits the immediate adjustment of nominal prices.
The monetary transmission mechanism takes effect through various channels. Kuttner & Mosser (2002) classify these channels as the interest rate channel, the wealth channel, the narrow credit channel, the broad credit channel, and the exchange rate channel. The channels will be concisely defined. The most elementary process is the interest rate channel. Given the stickiness of short-term prices as described in the above-mentioned assumptions, tightening the monetary policy leads to an increase in the real interest rate. Which in turn results in a decrease in investment spending and a reduction in aggregate demand (Mishkin, 1995). Secondly, the narrow credit channel depicts the relationship between the provision of credit by depository institutions and the availability of reserves. Essentially, since banks depend on demand deposits, in case of a contractionary monetary policy, by reducing bank reserves, this will in turn decrease loan supply and therefore reduce aggregate demand (Kuttner & Mosser, 2002).
The third is the wealth channel and the broad credit channel, both closely related. The connection with monetary policy lies in that an increase in interest rates, results in a fall of value for long-lived assets such as stocks, bonds, and real estate. According to the permanent income theory, in addition to changes in current income, changes in permanent income impact the
individual’s consumption pattern. Permanent income is determined by a person’s long-lived assets (Friedman, 1957). Ando and Modigliani (1963) developed the life-cycle model of consumption based on Friedman’s (1957) hypothesis, where wealth is the main driver of consumption. Based on this model, reducing households’ wealth results in a decrease in consumption. Fourth, the exchange rate channel which is often ignored in closed-economy models. The transmission comes initially from the interest rate channel through the uncovered interest rate parity. For instance, an increase in the domestic interest rate in comparison to foreign rates leads to an appreciation of the domestic currency, resulting in a decrease in net exports and aggregate spending. Finally, themonetarist channel,unlike the previously discussed channels, this channel focuses on direct quantitative changes in terms of outstanding assets instead of interest rates. Given the assumption of the nonexistence of perfect substitutes for assets issued by the central bank, changes in the number of assets bought by monetary policy impact relative price changes (Kuttner & Mosser, 2002).
2.2 Risk-Taking Channel
The financial system has recently been through various changes that stimulated a shift of focus in studying the transmission mechanism from the controls being that of monetary nature to prudential controls. Karagiannis et al (2010) argue the efficiency of the transmission has been perturbed which can be shown in the expansion of the gap between policy rates and money market rates. The latter indicates a probable inefficiency in the interest rate transmission channel. After the financial crisis of 2007-2008, a new monetary transmission channel was identified, namely therisk-taking channel. The risk-taking channel is the relation between monetary policy and the way economic actors perceive and price risk (Borio & Zhu, 2012). This channel transmits in three possible ways.
One is the effect of interest rates on valuations, incomes, and cash flows. An expansionary monetary policy, for instance, lower interest rates, increases risk tolerance. This framework is closely related to the financial accelerator by Bernanke et al (1999), which describes the intensification of shocks that are caused by changes in the credit market. Borio and Zhu, (2012) argue that lower interest rates increase borrowers’ wealth and in turn, their demand
for investment increases as well. An important distinction from the wealth channel, they state, is that this increased net worth lowers the expected probability of default by borrowers, resulting in an increase in investment. This added increase in investment caused by lower interest rates raises asset prices even further and in turn increasing the investors’ net worth even more. This is where the multiplier effect plays a role in this process and that is the financial accelerator. The relationship between risk-taking and the financial accelerator stems from the fact that risk tolerance is affected when profits, incomes, asset values, and the values of collateral increase. Borio and Zhu, (2012) therefore state that there is a positive relationship between risk tolerance and an increased level of wealth and that as a result of lower interest rates risk-taking behavior is stimulated.
The second mechanism takes effect through the link between market rates and target rates of return, namely the search for yield effect (Rajan, 2005). Given fixed rates of return targets, after a decrease in interest rates, it becomes more challenging to meet liabilities that were set at long-term fixed rates, the case for pension funds and insurance companies for instance. This indicates that the effect of this channel is more powerful when the gap between the market and target rates is particularly significant (Borio & Zhu, 2012).
The third effect according to Borio and Zhu (2012) comes through the communication and reaction function of the central bank. Moreover, increased transparency about future policy decisions, thus reducing uncertainty, leads to the reduction of risk premia.
How does bank risk-taking translate the mechanisms behind the recently identified missing link, namely, the risk-taking channel? Banks have an important function in the transmission of monetary policy. Based on the effects described earlier, it is expected that risk tolerance increases during periods of accommodative monetary policy, in other words, lower interest rates. Bonfirm & Soares (2018) obtain evidence of the existence of this channel by assessing loan growth to risky borrowers under high versus lower interest rates. Empirical evidence shows that banks do indeed take more risk when interest rates are lower. Especially when policy rates are expected to stay low for a significant period. Results also showed that risk-taking is more prevalent in banks considered large in size. The latter is also true for less
capitalized banks. Moreover, Bonfirm & Soares (2018) provided significant evidence that more liquid banks may also take on more risks.
2.3 Bank Risk-Taking under Low-Interest Rates
Following the 9/11 attacks on the World Trade Center, interest rates attained low historical levels. Namely, from 6.24% in 2000 to 1.13% in 2013 in the US. The latter also took effect in the Euro area where interest rates decreased from 4.38 in 2001 to 2.09% in 2004. From then interest rates remained exceptionally low (Delis & Kouretas, 2010). Therefore, the 2000s were characterized by an extended period of low levels of interest rates and loosely regulated banking systems combined with significant growth of credit. These factors heavily influenced the unfolding of the 2008 crisis (Angeloni et al, 2015). Before the crisis and the identification of the risk-taking channel, the ECB did not view monetary policy as a potential influential mechanism on the banking sector and its risk-taking activities.
An empirical study conducted by Delis and Kouretas, 2010 on a panel of euro banks shows strong evidence that low interest-rate environments increase risky bank assets and transform their portfolios towards a riskier standpoint. This relationship is more significant for banks engaging in nonconventional banking activities such as off-balance items and less significant for highly capitalized banks. In another study by Vervliet and Blikker (2017), where they investigated the US banking sector for risk-taking under the unusually low-interest rates. No evidence was found in terms of the search for yield as banks were able to sustain profit through non-interest income such as fees and hence did not have to participate in risky investment through trading. However, risky behavior was observed through lowering credit loss provisioning resulting in a shrinking buffer against unexpected defaults in a low-interest-rate environment.
Dell’Ariccia et al (2014) provided a model indicating the effect of lowering interest rates while linking leverage with the level of monitoring depending on the flexibility of its capital structure and its level of leverage. It concluded that a decrease in rates resulted in higher leverage and lower monitoring for banks capable of adjusting their capital structures. On the other hand, for banks with a fixed capital structure, if the bank is well-capitalized, the bank will reduce
monitoring. Otherwise, if the bank is already highly levered, it might increase monitoring and reduce risk. Another interesting aspect to look at would be how these findings differ between long-term and short-term interest rates. Investigating both the Euro-area and the US, Maddaloni and Peydro (2011) found sound indications that low short-term rates soften lending standards rather than low long-term interest rates. Low short-term rates here indicate that they are a result of monetary policy. Moreover, they infer that low for too long policy rates resulted in ever-increasing risk on bank assets and consequently led to the financial crisis. More specifically the effect was more prominent for mortgages. Another striking conclusion is that short-term rates or policy rates are more economically influential in the Euro area than in the US in terms of altering lending standards.
2.4 Unconventional Monetary policy
The global financial crisis of 2007-2008 affected most financial centers and economies around the globe. The immediate reaction from central banks was to considerably reduce interest rates. As policy rates got closer to zero, central banks issued a selection of unconventional monetary policies to help the financial system recover once it meets the zero lower bound. In general, the mechanism involved liquidity assistance for financial institutions, especially banks, by conducting widespread purchases of public assets. The European Central Bank primarily concentrated on liquidity support to revive ceased interbank transactions as well as the purchase of government bonds to tackle the outbreak of the national debt crisis in Greece, Ireland, and Portugal in 2010. Mid-2014 the ECB conducted a large-scale asset purchase program. UMP has been beneficial to avoid additional distress and helped the recovery of financial markets. Unconventional Monetary Policy is a mechanism used primarily to lower and flatten the yield curve i.e to lower interest rate level and tighten the gap between long- and short-term rates (Dell’Ariccia, Rabanal, and Sandri 2018). UMP takes shape in three forms:
Forward guidance which revolves around supplying information to economic agents about projected policy rates. However, forward guidance might not succeed in adjusting expectations in case agents already predicted those expectations or they simply do not trust the monetary policy’s commitment (Campbell, Evans, Fisher, and Justiniano 2012). The second
form is Quantitative easing which is a widespread acquisition of assets by the central bank funded by reserves. In theory, there is an unrestricted movement of agents across asset classes so that quantitative easing has no effect on bond yields as arbitrage will offsets these effects. In real life, however, the purchase of government bonds causes an increase in bond prices and reduces its yield. The reduction of bond yields can also be a result of signaling that the central bank is devoted to an expansionary policy stance. In this way, QE can also be a form of forward guidance. QE overall aims at banks to reduce their rate, increase lending, and promote prices across markets (Dell’Ariccia, Rabanal, and Sandri 2018).
The third form of UMP is negative interest rates.The ECB implements negative interest rates by making banks pay interest rates on the reserves that they hold at the central bank. This will result in individual banks wanting to decrease their excess reserves at the ECB by increasing credit and buying other financial assets. This approach aims to decrease lending rates, increase lending, and inflate prices throughout financial markets. In theory, it was believed that policy interest rates are not capable of declining below zero, stemming from the idea that agents would rather store cash than deposit money in accounts that charge interest rates. On the other hand, dealing with cash necessitates considerable transaction costs and risks. Given the reliance of banks on interest income, there are concerns about negative interest rates affecting bank profitability but so far banks seem hesitant to charge retail depositors negative interest rates and instead charge their customers through other fees (Borio & Gambacorta, 2017).
Goals and potential benefits arising from unconventional monetary policy have been discussed but there are arising concerns of various UMP side effects. Firstly, as previously expressed, all the aforementioned methods put pressure on bank profitability (Borio, Gambacorta, and Hofmann 2015). Second, the reduction of returns on safe and governmental securities leads banks and other financial institutions to engage in riskier investment opportunities (Rajan 2005). Third, banks have a lower incentive to write off impaired loans (Caballero, Hoshi, and Kashyap 2008). Finally, being that the central banks are more engaged in multiple objectives and policy mechanisms; for instance the use of forward guidance, the monetary body becomes more politically involved. (Taylor, 2016)
2.5 Negative Interest Rates in the EU
Following the financial crisis of 2008, the ECB’s priority was to tackle the effects of the credit crash and the dramatic fall in interbank lending. It was not until mid-2013 when the ECB started implementing the unconventional monetary policies due to weak growth and below-target inflation. On June 5th, 2014, the ECB established a negative interest rate of -0.1% then -0.4% in 2016 (Hutchinson & Smets, 2017).
Results from the Bank Lending Survey of April 2016 in the euro area following NIRP indicate that more than 80% if banks faced lower profitability or at least expect the latter. This should not come as a surprise since the negative deposit facility rate reduced net interest income. The Euro area banks also recorded reduced equity prices which can be attributed to worries about future profitability under NIRP. However, this decreased profitability is compensated, at least for large euro area banks, owing to lower impairments and higher non-interest income (Arteta et al, 2018). It was also recorded that at the end of 2016 in some European countries banks began passing on negative deposit rates but excluded households. The latter was implemented for corporate depositors, in Germany, for instance, this rate was on average -0.03% as of April 2019. Although stickier than with positive rates, this suggests that a complete interest-rate passthrough might occur eventually even to household deposits (Eisenschmidt & Smets, 2018).
2.6 Negative Interest Rates and Implications on Bank Risk-Taking
In this section, the way NIRP affects bank risk-taking will be explained using evidence from various papers trying to disentangle this effect from 2014 onwards. NIRP being an unconventional policy, most literature still did not aggregately settle its implications on the interest-rate passthrough and the rest of the monetary transmission channels.
Firstly, we start by explaining the effect on banks’ profitability as it is closely related to the possible observations in terms of bank lending and risk-taking. One of the most evident assumptions is the lack of interest rate pass-through of negative interest rates to lending rates given that banks are reluctant to charge their corporate and household depositors interest rates which might lead to a reduction in the banks’ reserves (Arteta et al, 2018). This reduction would
come as a result of the possibility of depositors switching to cash as an alternative. Moreover, the notion of depositors switching to cash still seems unlikely given the inconvenience of holding cash and the hustle of storage costs and other transaction costs (Eisenschmidt & Smets, 2018). In general, banks consider negative rates a financial burden which reduces their equity value. This effect would be usually reversed in the case where rates are positive following an interest rate cut. In other words, in past cases where interest rates remained positive, a decrease in policy rates would actually increase the franchise value (Brunnermeier & Koby, 2017). As mentioned earlier, banks have been urged by the ECB to change their business models to adjust profitability under this negative interest rate environment (Coeure, 2016). The most common way banks have been using, is to switch to more fee-based products. A study of non-interest income after NIRP shows that the increase is less significant for high-deposit banks in comparison with low-deposit banks (Altavilla et al, 2019). Furthermore, the hindered interest rate pass-through might magnify the credit channel since holding excess liquidity is even more costly if you have to pay for it, this is especially for banks with high deposit funding (Horwath et al, 2018).
How do these implications materialize when it comes to lending behavior and bank risk-taking? When studying the effect of lower interest rates, it generally boosts credit supply and increases risk-taking. Or at least that was the case in low but positive interest rate environments. Brunnermeier and Koby (2017) find that below a certain level of the policy rate, lowering interest rates can have a contractionary effect being linked to the aforementioned reduced profitability. The lower bound of policy rate where reducing rates becomes contractionary was called “expansionary lower bound” by Cavallino and Sandri (2017). Heider et al (2019) conducted a study on banks in the euro area before and after the introduction of NIRP comparing high vs low deposit banks. They found that high-deposit banks lend less but take more risks when compared to low-deposit banks. However, findings differ among paper, being that this is a fairly new phenomenon. Demiralp et al (2018) argue that high-deposit banks reacted by increasing the issuance of loans and find that negative rates, in line with lowering rates in positive environments, have an expansionary function. Bubeck et al (2020) conducted a difference-in-differences study to examine bank risk-taking in terms of holdings of risky security as a reaction to NPR, distinguishing between high and low deposit banks. Evidence from this
model shows significant proof of reach for yield behavior. Moreover, banks with high-deposits hold more securities with higher yields following the introduction of NIRP relative to their low-deposit counterparts. This is in line with the assumptions of increasing risk-taking after implementing NRPs for banks with higher customer deposits. Nucera et al (2017) find that risk-taking consequences depend on banks’ business models. They find that large banks with diversified sources of income take on less risk under NRP in comparison with smaller banks. Amzallag et al (2019) analyze the behavior of Italian mortgage lenders before and after NIRP, they find that banks with higher overnight deposits, increase rates on fixed-rate mortgages but found no evidence on adjustable-rate mortgages. Tan (2019) finds evidence aligned with the rest of the literature, banks mostly affected by NRP increase their lending in comparison with those less affected and that these effects lessen as negative rates linger.
2.7 Hypothesis development
Assessing the effects of lower interest rates in positive interest rates environments generally boosts credit supply and increases risk-taking. The effect in negative interest rates environments, since 2014, is still somewhat ambiguous. Various papers, using different frameworks often provided different outcomes.
Amongst almost all reviewed literature, it is apparent that there is a common distinction between low and high deposit banks given that negative interest rates transmission affected wholesale deposits primarily. However, due to a disruption in the passthrough to retail deposits in negative rates environments, there has been a debate about the impact and consequences on banks’ lending behavior and risk-taking. (Arteta et al, 2018) The presence of a lower bound on retail and customer deposit rates, namely the reluctance of banks to pass on negative interest rates to their depositors, suggests that the reliance on deposit funding should provide a distinction between banks that would be more impacted by NRP than others. Moreover, the theory suggests that banks with a higher deposit-to-assets ratio should be more impacted by the introduction of negative rates (Heider et al 2019). Banks with high deposit ratios experience a higher drop in equity value given a more significant anticipated lowered profitability as a result of higher reliance on deposits as a fraction of their assets (Brunnermeier & Koby, 2018).
Combining these factors, it is expected that high deposit banks indulge in higher risk-taking. In line with the findings described in the previous chapter, a hypothesis is formulated:
Hypothesis: Negative Interest rate policy increases risk-taking and leads to a reduction of credit issued by banks with high-deposit funding
3. DATA DESCRIPTION & METHODOLOGY
3.1 Baseline Model & Choice of Variables
We exploit the distinction between high and low deposit banks to create a treatment and control group in order to test the above-mentioned hypothesis. This empirical specification tests two outcomes of interest. Firstly, the lending volume and how it may be affected by a negative interest rate policy between low and high deposit banks. Secondly, the risk-taking level of banks and how it is influenced by negative rates between low and high deposit banks. In order to test these effects, a Difference-in-Differences method is conducted loosely based on Heider et al (2019) and Bubeck et al (2020). The convenience behind using Difference-in-Differences methodology lies in the fact that it enables causal conclusions and is fairly robust against omitted variables. The baseline specification is as follows:
= (1)
yijk α + β * Deposit ratio i × After(2014) j+ uijk
● γ ijk: The dependent variable is associated with lending behavior of bank at time in i j country in log form to estimate the percentage changek
● After(2014) jis the treatment dummy variable equaling 1 after the negative rate policy was introduced.
● eposit ratio D i is the treatment group variable which is the deposit ratio as a percentage of total assets. It is a dummy equaling 1 if in 2014 the deposit ratio was above median.
The justifiability of the difference-in-differences depends on the validity of underlying assumptions. According to Abadie (2006) & Bonhomme and Sauder (2011), the first assumption
is that the reference group should be a sound counterfactual against the treatment group. Namely, that banks with a high deposit are more affected by NIPR than banks with a low deposit ratio. The latter is proven over and over in the underlying empirical literature reviewed in the previous section. The second assumption is that of time-invariant heterogeneity, parallel trends assumption must be fulfilled. The graphical representation in figure 1 and figure 2 in Appendix A1 shows that on average the parallel trends assumption holds for the dependent variables, total assets, and loan loss provisions respectively. In other words, it assumes the average variation in the output variable at baseline for the treated group in the absence of treatment (before the introduction of NIRP) to be comparable to the average variation in the dependent variable for the control group. Finally, the assumption that the ECB’s announcement in 2014 of establishing NRP was unexpected and that banks did not already incorporate expectations of negative rates before the treatment period.
To build on the first equation (1), control variables and fixed effects are added to attain the final model (2):
= (2)
yijk α + β * Deposit ratio i × After(2014) j+ γx t−1 ij + c kj+ δ i + η j + uijk
represents lagged bank-specific control variables to avoid endogeneity issues. Firstly, the
xt−1 ij
size of the bank proxied by the natural logarithm of total assets. Moreover, liquidity and capital ratio are also used. The inclusion of balance sheet characteristics is backed by previous literature covering the transmission of policy rates, for example (Jiménez et al., 2012). ckjincorporates country-specific variables. Following Delis and Kouretas (2010), we include GDP growth to control for the business cycle and CPI as a measure of inflation and macroeconomic conditions.
and are bank and year fixed effects, according to Tan (2019), this is important to isolate
δ i η j
the impact of negative rates. Finally, bank-level clustered standard errors are used when running the regression to account for serial correlation. (Bertrand et al, 2004)
The deposit ratio which is customer deposits as a percentage of total assets was used to split the sample into treatment and control groups. The treatment group consists of banks with a high-deposit ratio and the control group consists of banks with a low-deposit ratio. The latter was done using the median deposit ratio in 2014 across our sample (71.59 %) . Table 2 shows
descriptive statistics of the two groups in terms of loans and the size of banks. Overall low-deposit banks are considered larger banks given a higher mean of assets and loans.
The main dependent variables for our results are the natural logarithm of total loans and the natural logarithm of loan loss provisions. The log of total loans is used as a proxy for the volume of lending. In addition, bank risk-taking can also be represented using credit risk, measured by the log of loan loss provisions. This variable measures the quality of the assets held by the bank, indicating the allowance for uncollected loans associated with the expectations of future losses. Banks have to maintain loan loss provisions if there is a possibility of loan impairments. LLPs can also reflect the aggressiveness of banks’ lending decisions. Boungou (2019) suggests that higher LLPs would indicate an increase in the bank's risk-taking. Despite Non-performing loans being the most commonly used proxy for risk, the frequent unavailability of NPL data has forced us to use LLPs. Loan loss provisions as a proxy of risk were used by García-Alcober et al (2019), Boungou (2019), and Khan et al (2017).
Banks variable N mean sd p5 p95
Low-deposit ratio Total Loans 7150 7387.05 18961.25 54.21 40489.43 Total Assets 7150 12711.14 33713.02 101.25 70408.48
High-deposit ratio Total Loans 8034 1546.68 6694.14 25.75 4989.72 Total Assets 8034 2737.93 12638.27 53.47 8669.64
Total Total Loans 15184 4296.85 14194.86 32.65 20365.00
Total Assets 15184 7434.22 25386.03 65.97 34757.77
3.2 Data
For this study, bank-level data was mainly gathered from Bureau van Dijk’s BankFocus database. The sample consists of 3,752 commercial, savings, and cooperative banks from 19 countries in the Euro Area. Table 1 with the bank distribution for each country in the sample is linked in Appendix A2. The sample ranges from 2012 to 2016, covering the before and after NIRP implementation in 2014. The estimation window is kept relatively short around negative interest rates’ implementation to reduce the effect of overlapping factors (Lechner, 2011). Country-level or macro data was sourced from the World Bank database.
For the data cleaning process, observations with assets’ growth rate below the 5th and above the 95th percentile were removed, as well as observations with 5 or more missing variables. To avoid duplication of variables and double counting, branches of banks were excluded from the sample. Furthermore, all of the dependent and bank-specific control variables were winsorized at the 1st and 99th percentile to minimize the effect of outliers. Finally, we limit the sample to banks that actually took deposits between 2012 and 2016.
In Appendix A3, we test our variable for multicollinearity. According to Allen (1997), the problem of multicollinearity occurs when there is a high correlation between two or more independent variables. A high correlation is usually considered when it is 0.7 or higher. Based on table 4 in appendix A3, correlations between independent and control variables are all relatively low, with the highest being -0.185 between GDP and liquidity ratio.
3.3 Summary statistics
Variable Units N mean Median St. Dev. p5 p95
Dependent variables
Total loans 15184 4296.85 396.79 14194.86 32.65 20365.00 Loan Loss Provisions 15184 28.57 0.85 126.60 -3.34 100.86
Log (Loans)
in million
euro 15184 6.23 5.98 1.93 3.49 9.92 Log (Loan Loss Provisions) 11479 0.90 0.83 2.40 -2.92 5.11 Deposit variables
Deposit Ratio % 15184 66.54 72.12 18.77 28.20 87.44 Deposit Ratio in 2014 % 14298 65.86 71.59 18.92 28.12 87.14 Bank-level control variables
Total assets in million euro 15184 7434.22 698.85 25386.03 65.97 34757.77 Liquidity ratio % 15184 24.80 20.01 18.40 3.78 60.81 Capital ratio % 15184 9.78 9.07 3.83 4.82 16.84 Country-level control variables Inflation (CPI) (2010 = 100) 14093 107.25 107.20 1.77 104.71 110.69 GDP Growth % 14093 1.00 1.01 1.53 -1.84 2.23
4. RESULTS 4.1 Impact of NIRP on lending volume
Equation (3) shows the effect of a negative interest rate on the volume of lending for high deposit ratio banks:
= (3)
og(T otal loans)
L ijk α + β * Deposit ratio i × After(2014) j+ γx ij t−1+ c kj+ δ i + η j + uijk
is the lending volume measure for bank i in country k at time j. Table 5 og(T otal loans)
L ijk
presents the results of equation (3). For each column, we add more control variables. All regressions were run as a fixed effect panel regression with bank-level clustered standard errors to account for serial correlation. Column (1) estimates equation (3) without control variables. Column (2) includes only bank-level control variables while in column (3) country-level control variables are added. We notice R squared increases as we add more variables. It reaches 0.707 in the 3rd column. This is a sign that the new terms improve our model.
The treatment effect is quite stable and on average positive. It is also highly significant across the 3 columns with a p-value of less than 1%. Fixing for bank, country, and time effects, the first column shows a significant coefficient on our difference in differences variable (0.0607). In the 2nd column, it is still significant but decreases to (0.0239), probably due to the additional control variables explaining the increase in lending more. In the 3rd column, high-deposit banks increased lending with 1.15% ((e 0.0114− 1 × 1) 00) in comparison with low-deposit banks after the introduction of negative interest rates. This result is with the inclusion of bank and country-level control variables.
Size is strongly correlated with loans but shows a decrease in its coefficient after including country-specific controls. Moreover, an increase in the bank's assets is accompanied by increased lending. The capital ratio is insignificant for both specifications, this can be explained by the ambiguity of its effect on lending in general. According to Kashyap and Stein (2000), the effect of monetary policy on lending behavior is more robust when banks are less liquid. This is in line with the findings, where the liquidity ratio’s coefficient is negative with high significance.
Log(Total loans)
(1) (2) (3)
Deposit ratio × After(2014) 0.0607*** 0.0239*** 0.0114*** (0.00879) (0.00473) (0.00333) size 0.930*** 0.901*** (0.0377) (0.0342) Capital ratio 0.00391 0.00357 (0.00394) (0.00305) Liquidity ratio -0.0133*** -0.0117*** (0.00138) (0.000824) GDP growth 0.000779 (0.00147) inflation (CPI) 0.0109*** (0.00126) constant 6.138*** 0.174 -0.832*** (0.00523) (0.225) (0.175) N 15184 15184 14087 R-squared 0.143 0.673 0.707
bank FE yes yes yes
Year FE yes yes yes
Country FE yes yes yes
* p<0.1, ** p<0.05, *** p<0.01
Table 5 Impact of NIRP on banks’ lending volume
Notes: The sample comprises annual data of 3,752 European banks operating in 19 countries from the Euro Area over the period 2012–2016.
4.2 Impact of NIRP on risk-taking
= (4)
og(LLP )
L ijk α + β * Deposit ratio i × After(2014) j+ γx ij t−1+ c kj+ δ i + η j + uijk
is the risk-taking measure for bank i in country k at time j, which is loan loss og(LLP )
L ijk
provisions. Table 6 presents the results of equation (4). For each column, we add more control variables. All regressions were run as a fixed effect panel regression with bank-level clustered standard errors to account for serial correlation. Column (1) estimates equation (4) without control variables. Column (2) includes only bank-level control variables while in column (3) country-level control variables are added.
The treatment effect is on average positive. It is not that significant in the first 2 columns however in the 3rd column and after the inclusion of the rest of the controls, the coefficient for our difference-in-differences variable is significant with a p-value less than 1%. Fixing for bank, country, and time effects, the first column shows a significant coefficient but only with a p-value of less than 10% (0.0890). In the 2nd column, the coefficient seems to have decreased but it does so insignificantly. In the 3rd column, high-deposit banks significantly increased their loss loan provisions by 13.64% ( e ( 0.128− 1 × 1) 00) relative to low-deposit banks after the introduction of negative interest rates. This result is with the inclusion of bank and country-level control variables.
Size is strongly correlated with loan loss provisions, this is expected as larger banks often keep more provisions. Moreover, it shows a decrease in its effect after the inclusion of country-specific controls. The capital ratio is in this case significant for both specifications, and it shows that the less capitalized banks have higher loan loss provisions. This is in line with Bonfirm & Soares (2018) they found that banks with lower capital ratio indulge in more risk-taking. The liquidity ratio’s coefficient is again negative with high significance. This again can be explained by the fact that less liquid banks take more risks.
Log(LLPs)
(1) (2) (3)
Deposit ratio × After(2014) 0.0890* 0.0525 0.128*** (0.0463) (0.0455) (0.0492) size 0.712*** 0.628*** (0.147) (0.144) Capital ratio -0.0426*** -0.0428*** (0.013) (0.0129) Liquidity ratio -0.00874** -0.00714* (0.00392) (0.00384) GDP growth -0.00214 (0.0158) inflation (CPI) -0.0529** (0.0218) constant 0.974*** -3.313*** 2.694 (0.0253) (1.049) (2.354) N 11479 11479 10594 R-squared 0.0276 0.0491 0.0423
bank fixed effects yes yes yes
Year fixed effects yes yes yes
country fixed effects yes yes yes
* p<0.1, ** p<0.05, *** p<0.01
Table 6 Impact of NIRP on banks’ risk-taking
Notes: The sample comprises annual data of 3,752 European banks operating in 19 countries from the Euro Area over the period 2012–2016.
5. DISCUSSION
5.1 Interpretation of results
The first finding from our difference-in-differences specification is that banks with a high deposit to asset ratio increased their lending volume after the introduction of negative rates in comparison with low deposit banks. This finding rejects the first part of the formulated hypothesis. However, it is in line with Demiralp et al (2018)’s results, namely that high-deposit banks reacted to NIRP by increasing the issuance of loans. We find that indeed, lowering interest rates in negative interest rates environments still has the same expansionary function as lowering rates in a positive interest rate environment. This relationship between negative rates and increased issuance of credit has been confirmed by other recent papers such as (Schelling and Towbin, 2018; Lopez et al, 2018; and Tan, 2019)
On the other hand, Heider et al (2019), found that an increase in negative rates resulted in a lower lending volume. One reasoning could be that Heider et al (2019) did not use total loans as their dependent variable but syndicated loans instead, which only consist of a small percentage of bank loans. Their sample also only consisted of 23 euro banks, whereas our sample is much larger with 3,752 banks across the Eurozone.
As opposed to what some papers predicted, according to this finding, negative rates do not yet have a contractionary effect. It is apparent that rates did not yet reach the “expansionary lower bound” or the reversal rate described by Brunnermeier and Koby (2017) and Cavallino and Sandri (2017)
The second result focuses more specifically on banks’ risk-taking. Our findings indicate that banks with a high deposit to asset ratio increased their loan loss provisions after the introduction of negative rates in comparison with low deposit banks. According to Boungou (2019), LLPs reflect the aggressiveness of banks’ lending decisions and that a higher LLP would indicate an increase in the bank's risk-taking. Our result, therefore, implies banks with a high deposit ratio increase their risk-taking following the implementation of NIRP compared to low deposit banks.
This finding is in line with the second part of the formulated hypothesis. Looking at the recent findings, this result is in accordance with most literature. Heider et al (2019) and Bubeck et al (2020) for instance conducted a study on banks in the euro area before and after the introduction of NIRP comparing high vs low deposit banks. They found that high-deposit banks take more risks when compared to low-deposit banks. Furthermore, from observing our control variables, we found that in line with Bonfirm and Soares (2018) less capitalized banks have higher loan loss provisions. Therefore, banks with lower capital ratios indulge in more risk-taking.
5.2 Policy Implications
In general, the positive effect on bank lending and risk-taking by negative rates is consistent with the expansionary monetary policy in a normal or positive interest rate environment and its typical intention to provide sufficient monetary policy accommodation. Evidence from this study and other precedent studies strongly suggest that the reversal rate was not yet breached. This proposes that the European central bank has more room to implement negative rates in an effort to boost economic growth and raise inflation levels without prompting contractionary behavior.
Furthermore, in some European countries banks began passing on negative deposit rates but excluded households. The latter was implemented for corporate depositors, in Germany, for instance, this rate was on average -0.03% as of April 2019. Although stickier than with positive rates, this suggests that a complete interest-rate passthrough might occur eventually even to household deposits (Eisenschmidt & Smets, 2018). This implies that individuals might eventually become more tolerant towards being charged negative rates, this can happen as a result of two factors. Firstly, switching to cash still seems unlikely given the inconvenience of holding cash and the hustle of storage costs and other transaction costs (Eisenschmidt & Smets, 2018). Secondly, banks already started charging more fees as a form of non-interest income to their customers (Borio & Gambacorta, 2017). If these fees keep increasing then it is more likely to assume customers would prefer a low negative rate on their deposits.
5.3 Limitations
In this section, two main limitations of this study are acknowledged. Firstly, our sample period of 2012-2016 confounds with other unconventional tools. Namely, quantitative easing and forward guidance. Although we have used a difference-in-differences framework, it might be that some effects were due to quantitative easing which has been found to promote risk-taking (Heider et al, 2017). Secondly, the use of yearly panel data provided perhaps a less precise estimation of the pre and post negative rate implementation period. The use of monthly data would have been more effective to estimate the effect more precisely before and after June 2014. However, monthly data for all our variables were scarce and would leave the study with a very small sample of banks and a decreased statistical power.
6. CONCLUSION
On June 5th, 2014, the ECB implemented negative interest rates to stimulate economic growth and tackle the below-target inflation levels. This unconventional monetary policy spurred a debate in monetary transmission literature. In this paper, we used a dataset of 3752 banks operating in 19 countries in the Euro Area between 2012-2016 to estimate the impact of the introduction of negative interest rates on banks’ lending behavior and risk-taking using a difference-in-differences framework. We find that negative Interest rate policy increases risk-taking and leads to a growth of credit issued by banks with high-deposit funding.
Furthermore, these results suggest that the European central bank has more room to implement negative rates in an effort to boost economic growth and raise inflation levels without prompting contractionary behavior. However, further research is still needed to disentangle the effects of persistent negative rates on bank-level as well as macro level. Further implications of NIRP on the economy as a whole can help unravel mechanisms behind these effects.
7. REFERENCES
Abadie, A. (2006). Semiparametric Difference-in-Differences Estimators. The Review of Economic Studies, 72, pp. 1-19.
Allen, M. P. (1997). The problem of multicollinearity. Understanding regression analysis, 176-180.
Altavilla, C., Burlon, L., Giannetti, M., & Holton, S. (2019). Is There a Zero Lower Bound? The Effects of Negative Policy Rates on Banks and Firms. ECB Working paper 2289.
Amzallag, A., Calza, A., Georgarakos, D., & Sousa, J. (2019). Monetary Policy Transmission to Mortgages in a Negative Interest Rate Environment. ECB Working Paper, 2243.
Ando, A., & Modigliani, F. (1963). The "Life Cycle" Hypothesis of Saving: Aggregate Implications and Tests. The American Economic Review, 53, 55-84.
Angeloni, I., Faia, E., & Duca, M. L. (2015). Monetary Policy and Risk Taking. Journal of Economic Dynamics and Control.
Arteta, C., Kose, M. A., Stocker, M., & Taskin, T. (2018). Implications of negative interest rate policies: An early assessment. Pacific Economic Review, 23(1), 8-26.
Basten, C. & Mariathasan, M. (2018). How banks respond to negative interest rates: Evidence from the swiss exemption threshold. CESifo Working Paper Series, 6901
Bernanke, B. S., Gertler, M., & Gilchrist, S. (1999). The financial accelerator in a quantitative business cycle framework. Handbook of Macroeconomics, 1341–1393.
Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences in-differences estimates? The Quarterly Journal of Economics, 119 (1), 249–275.
Bikker, J. A., & Vervliet, T. M. (2017). Bank profitability and risk-taking under low interest rates. International Journal of Finance & Economics, 23(1), 3–18.
Bonfim, D., & Soares, C. (2018). The Risk-Taking Channel of Monetary Policy: Exploring All Avenues. Journal of Money, Credit and Banking, 50(7), 1507–1541.
Bonhomme, S., & Sauder, U. (2011). Recovering distributions in difference-in-differences models: A comparison of selective and comprehensive schooling. Review of Economics and Statistics, 93(2), 479-494.
Borio, C., & Zhu, H. (2012). Capital regulation, risk-taking and monetary policy: A missing link in the transmission mechanism? Journal of Financial Stability, 8(4), 236–251.
Borio, C., & Gambacorta, L. (2017). Monetary policy and bank lending in a low interest rate environment: Diminishing effectiveness? Journal of Macroeconomics, 54, 217-231.
Borio, C., Gambacorta, L., & Hofmann, B. (2017). The influence of monetary policy on bank profitability. International Finance, 20(1), 48-63.
Boungou, W. (2019). Negative interest rates, bank profitability and risk-taking. Available at http://dx.doi.org/10.2139/ssrn.3416762
Brunnermeier, M. K., & Koby, Y. (2017). The Reversal Interest Rate: An Effective Lower Bound on Monetary Policy. Unpublished, Princeton University, Princeton, NJ. Available at: https://scholar.princeton.edu/markus/publications/reversal-interest-rate-effective-lower-bound-m onetary-policy
Bubeck, J., Maddaloni, A., & Peydró, J. (2020). Negative monetary policy rates and systemic banks’ risk-taking: evidence from the euro area securities register. ECB working paper, 2398. Caballero, R. J., Hoshi, T., & Kashyap, A. K. (2008). Zombie Lending and Depressed Restructuring in Japan. American Economic Review, 98(5), 1943-1977.
Campbell, J. R., Evans, C. L., Fisher, J. D., & Justiniano, A. (2012). Macroeconomic Effects of Federal Reserve Forward Guidance. SSRN Electronic Journal. doi:10.2139/ssrn.2166310
Cavallino, P., & Sandri, D. (2017). The Expansionary Lower Bound: A Theory of Contractionary Monetary Easing. IMF Working Papers, 236.
Coeure, B. (2016, July 28). Assessing the implications of negative interest rates. Speech presented at Yale Financial Crisis Forum in Yale School of Management, New Haven. Retrieved from https://www.ecb.europa.eu/press/key/date/2016/html/sp160728.en.html
Delis, M., Kouretas, G. (2010) Interest rates and bank risk-taking. Journal of Banking & Finance, 35(4), 840-855.
Dellʼariccia, G., Laeven, L., & Marquez, R. (2014). Real interest rates, leverage, and bank risk-taking. Journal of Economic Theory, 149, 65–99.
Dell’Ariccia, G., Rabanal, P., & Sandri, D. (2018). Unconventional Monetary Policies in the Euro Area, Japan, and the United Kingdom. Journal of Economic Perspectives, 32(4), 147-172.
Demiralp, S., Eisenschmidt, J., & Vlassopoulos, T. (2018). Negative Interest Rates, Excess Liquidity and Bank Business Models: Banks Reaction to Unconventional Monetary Policy in the Euro Area. SSRN Electronic Journal. doi:10.2139/ssrn.2941377
Eisenschmidt, J., & Smets, F. (2018). Negative Interest Rates: Lessons from the Euro Area. Monetary Policy and Financial Stability: Transmission Mechanisms and Policy Implications, 26, 13-42.
Friedman, M. (1957). A theory of the consumption function. Princeton: Princeton University Press, 20-37.
García-Alcober, M., Prior, D., Tortosa-Ausina, E., & Illueca, M. (2019). Risk-taking behavior, earnings quality, and bank performance: A profit frontier approach. BRQ Business Research Quarterly.
Heider, F., Saidi, F., & Schepens, G. (2019). Life below Zero: Bank Lending under Negative Policy Rates. The Review of Financial Studies, 32(10), 3728-3761.
Horwath, R., Kotlebova, J., & Siranova, M. (2018). Interest rate pass-through in the euro area: Financial fragmentation, balance sheet policies and negative rates.Journal of Financial Stability, 36, 12-21.
Hutchinson, J., & Smets, F. (2017). Monetary Policy in Uncertain Times: ECB Monetary Policy Since June 2014. The Manchester School, 85.
Ireland, P. N. (2005). Monetary Transmission Mechanism. Federal Reserve Bank of Boston working paper No. 06-1. Available at http://dx.doi.org/10.2139/ssrn.887524
Jiménez, G., Ongena, S., Peydro, J.-L., & Saurina, J. (2012). Credit supply and monetary policy: Identifying the bank balance-sheet channel with loan applications. American Economic Review, 102 (5), 2301–26.
Karagiannis, S., Panagopoulos, Y., & Vlamis, P. (2010). Interest rate pass-through in Europe and the US: Monetary policy after the financial crisis. Journal of Policy Modeling, 32(3), 323–338.
Kashyap, A. K., & Stein, J. C. (2000). What do a million observations on banks say about the transmission of monetary policy?. American Economic Review, 90(3), 407-428.
Khan, M. S., Scheule, H., & Wu, E. (2017). Funding liquidity and bank risk taking. Journal of Banking & Finance, 82, 203-216.
Kuttner, K.N., Mosser, P.C. (2002). The Monetary Transmission Mechanism: Some Answers and Further Questions. Federal Reserve Bank of New York, Economic Policy Review, pp.15-26.
Lechner, M. (2011). The estimation of causal effects by difference-in-difference methods. Foundations and Trends in Econometrics, 4(3), 165–224.
Maddaloni, A., & Peydró, J.-L. (2011). Bank Risk-taking, Securitization, Supervision, and Low Interest Rates: Evidence from the Euro-area and the U.S. Lending Standards. Review of Financial Studies, 24(6), 2121–2165.
Mishkin, F.S. (1995). Symposium on the Monetary Transmission Mechanism. The Journal of Economic Perspectives 9-4, pp. 3-10.
Nucera, F., Lucas, A., Schaumburg, J., & Schwaab, B. (2017). Do negative interest rates make banks less safe? Economics Letters, 159, 112-115.
Rajan, R. (2005). Has Financial Development Made the World Riskier? Working paper 11728. Available at http://www.nber.org/papers/w11728
Schelling, T. & Towbin, P. (2018). Negative interest rates, deposit funding and bank lending.
Taylor, J. (2016). “Independence and the Scope of the Central Bank’s Mandate.” Sveriges Riksbank Economic Review, 3, 96–103.
Tan, G. (2019). Beyond the zero lower bound: negative policy rates and bank lending. DNB working paper, 649.
8. APPENDIX A1: Parallel trends assumption
Figure 1. Parallel trends for total loans (in million euro)
A2. Banks per country Country Banks Austria 627 Belgium 36 Cyprus 35 Germany 1,281 Estonia 9 Spain 278 Finland 164 France 254 Greece 10 Ireland 23 Italy 767 Lithuania 8 Luxembourg 54 Latvia 23 Malta 13 Netherlands 24 Portugal 109 Slovenia 21 Slovakia 16 Total 3,752
A3. Test for multicollinearity Total loans LLPs Total Assets Capital ratio Liquidity ratio GDP CPI Total loans 1 LLPs 0.7288 1 Total Assets 0.9316 0.631 1 Capital ratio -0.163 -0.1118 -0.184 1 Liquidity ratio -0.0143 -0.0078 0.065 0.0875 1 GDP -0.0437 -0.1568 -0.0354 0.0257 -0.185 1 CPI -0.0907 -0.0999 -0.0833 0.0865 0.1086 0.1934 1
