MSc Thesis
The effect of negative interest rate policy on banks’ profitability in
advanced economies.
Gülnur Nasirova 10223665
University of Amsterdam, Amsterdam Business School
MSc Finance: Banking and Regulation
Thesis supervisor: Enrico Perotti
Statement of Originality
This document is written by Gulnur Nasirova who declares to take full responsibility for the contents of this document.
I declare that the text and 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, ECB announced the introduction of monetary policy, namely Negative Interest Rate Policy, and spurred a lot of debate about its potential implications on banking sector. The current research examines the consequences of implementing negative interest rate policy on bank profitability, specifically on net interest margins of the banks in advanced countries. Using difference-in-difference approach, we find evidence that banks’ net interest margins have been affected in a sizable and negative way by introduction of NIRP. Liquidity channel test further confirms the negative nature of the effect of NIRP on bank profitability and significant role of liquidity provisions that determine the banks’ performance. The initiation of other unconventional monetary policies could be associated with alterations in NIM during the period under investigation, and have particular importance for research and policy discourse in the future.
Keywords: Negative Interest Rates, ECB, bank profitability, Unconventional Monetary Policy
Table of Contents
1. Introduction ... 5
2. Understanding Zero Lower Bound ... 6
3. Literature Review ... 7
4. Empirical methodology and Data ... 10
4. 1 Methodology ... 10
4.2 Difference-in-Difference Assumptions ... 13
4.3 Data ... 16
5. Empirical Results ... 19
5.1
Baseline Regression ... 19
5. 2 Liquidity channel test ... 20
6. Conclusions ... 22
7. Bibliography ... 24
1. Introduction
To achieve recovery in economic activity after slowdown caused by global financial crisis many advanced economies decided to implement unconventional monetary policies, such as asset purchase programs, credit easing, forward guidance as well as Negative Interest Rate Policy (NIRP) to support economic growth and to keep inflation close to target. However, already in the course of financial crisis major central banks had cut their policy rates to or slightly above zero, as zero was then regarded the lower bound for policy rates. As policy rates were near zero, the net supply of eligible assets for central bank purchase programs was declining, and commercial banks were hoarding excess reserves rather than lending to households and companies. Shifting the alleged zero lower bound was perceived by some central banks as a needed supplement and promoter of current policy initiatives. These central banks now impose penalty on banks for their surplus reserves. The unprecedented use of NIRP in wide range of countries, sum up for one-fourth of world GDP, has not just extended the boundaries of unconventional monetary policies, but also fueled an already polarized debate on the implications of these policies (Jobst et al, 2016).
Historically, negative interest rates—policy-determined or otherwise—have been an extremely rare phenomenon and widespread emergence of negative interest rates outside of a financial crisis is unprecedented. The introduction of NIRP has led to an intensive debate, where some argue that NIRP have so far served the intended purpose, complementing the broader set of expansionary measures employed by central banks. Others, however, emphasize financial stability risks associated with NIRP and claim that they may have weakened banks’ willingness to lend, contributed to financial market distortions, further inflated asset prices, and delayed the implementation of necessary macroeconomic and structural policies (Jobst and Lin et al, 2016). Regardless of these highly polarized debates, studies on NIRP’s effect on bank profitability have been quite limited.
As such, it is critical to inquire into more detail whether moving towards negative interest rates has changed the balance sheet of the banks. The focus of this paper will be the impact of negative interest rates on the banking industry, more specifically on banks’ profitability. For that reason, this paper aims to shed light on this issue and contribute to existing literature by presenting a comprehensive analysis
behind the implementation of NIR policy. The question under investigation is: To what extent negative interest rates have impact on banks’ profitability in advanced economies? The cross-country sample of panel data will be employed to investigate the effect of NIRP on bank profitability by using Difference-in-Difference approach.
Our paper will have the following structure. In section 2, explanation of zero lower bound is given. Further, of extensive literature review is presented in section 3 is provided, followed by empirical framework and description of the data in section 4. In Section 5 and 6, we demonstrate regression results and present conslusions.
2. Understanding Zero Lower Bound
Before central bank in Denmark (DNB) introduced negative deposit rate in 2014, the notion of below zero interest rates had only theoretical domain. Due to downward-stickiness of deposit rates and reluctance of banks to impose negative yield on the customers, it was assumed that interest rates would not enter below zero territory. Nevertheless, if we consider the various storage and transaction costs associated with keeping lots of cash, then we come to the concept of physical lower boundary of interest rates, depicted by Benoit Coeuré in Figure 1.
As banks tend to cut deposit rates at a slow pace, they will run into economic lower bound before reaching physical lower bound. Economic lower bound, also called “Reversal Rate” introduced by Brunnermeier, is the point, when benefits of diminishing policy rates do not compensate the costs anymore, and already start having adverse effect on the volumes of credit supply, for instance. Thus, breaching economic lower rate can cause unpredicted losses to the entire economy, among other excessive risk-taking and disequilibrium in financial markets.
Figure 1. Breaching Zero Nominal Rate
Source: Benoit Coeuré (2016): Presentation at Yale Financial Crisis Forum.
3. Literature Review
Empirical evidence on the impact of the negative interest rates on bank profitability is somewhat limited, and even fewer papers focus specifically on the effect of NIRP on profitability.
One of the earliest papers on this topic is a cross-country study using bank specific data on bank margins written by Demirgüç-Kunt and Huizinga (1999). They analyzed how a range of macroeconomic and bank characteristics influence banks’ net interest income and profitability.They conclude that higher interest rates are correlated with higher NIMs and profits, particularly in developing countries, partly because interest rates on deposits there are more likely to be regulated and below market- rates. A recent paper Borio, C., Gambacorta, L. & Hofmann,B. (2015) is most relevant to our study for building econometric model. By investigating the period 1995-2012, they discover non-linear relationships between bank NIMs and yield curve slope. They
determine that the effect on NIMs are more intense as interest rates get lower, where there is an unusually flat term structure. All of this implies that, over time, low interest rates might erode bank profitability.
Studies for several individual countries point to the greater negative effects induced by low interest rates on net interest margins. Paper by Genay and Podjasek (2014) suggest through a narrower spread and for a longer period of time U.S. banks have been negatively influenced by diminishing interest rates. They also mention, however, that the direct effects of low rates are small relative to the cost advantages, also through higher asset quality. Likewise, while not explicitly investigating the effects of interest rates on banks, a research of 98 EU banks (ECB, 2015) infer that macroeconomic factors, and not interest rates, have played the crucial role for bank health, as the GFC Analysis for Germany propose that in “normal” interest rate environments, the long-run effect of a 100 basis points change on NIMs is very limited at around 7 basis points (Busch and Memmel, 2015). The following survey was not conducted in a specifically low interest rate environment; however, the Bundesbank Financial Stability Review of September 2015, investigating 1,500 banks, did not discover that long-term low interest rates depress German banks’ performance.
Research on other countries on the impact of negative interest rates on profitability is scarcer. Inspection using Japanese bank data shows that low-for-long interest rates are also responsible for diminishing net interest margins of Japanese banks (Deutsche Bank, 2013). Over time, however, portfolio shifts towards investment in securities, a greater reliance on non-interest earnings, and lower costs allowed Japanese banks’ profitability to be kept for the most part positive. In addition, recently after almost two decades with low-for-long interest rates, banks in Japan began expanding internationally, supposedly boosting profitability.
The literature has discovered that the direct impact of shifts in interest rates on margins and profitability can vary by bank. Study of U.S. banks indicates that in short-run small banks that are more reliant on traditional intermediation of retail deposits are more heavily affected by changes in interest rates (Genay and Podjasek, 2014). The differences in profitability of small and large banks caused by changes in interest rates depends partly on composition of banks’ asset and liabilities, in the competition for funds and lending opportunities, and in general business models. At the same time, as documented by Landier, Sraer and Thesmar (2013), US banks are exposed to interest
rate risks since their assets are more sensitive than their liabilities are, a risk which is typically not fully offset by banks’ use of derivatives. As presented by Drechsler, Savov, and Schnabl (2014), interest rates on deposit is less sensitive to changes in the Fed funds rate if there is less deposit competition.
English et al. (2012) show that while equity prices of U.S. banks normally drop after unanticipated rise in interest rates or a steepening of the yield curve, a substantial maturity gap lessens this effect, implying that due to their maturity transformation function, banks lose relative to a lower interest rate or a shallower yield curve. However, to offset reduced margins it is possible to broaden non-interest income given that the bank is relatively large and diversified, and thus have more potential to expand lending abroad. At the same time, consistent wit other evidence, Calomiris and Nissim (2014) show that banks have experienced significant changes in the valuation of their various growth opportunities since the global financial crisis and as rates have declined. For instance, a permanent reduction in interest rates brings down the gross value of core deposits, and since branches still have non-interest expenses, sustaining deposit relationships could turn into negative present value business.
Dell'Ariccia, G., and Marquez, R. (2013) have summarized emerging literature on why and how banks’ decisions concerning the overall risk of their portfolios, and their capital structures, may be influenced by changes in the interest rate environment and by monetary policy that affects it. They also present a simple model that is used to analyse the likely effect of the low interest–rate environment in the run-up to the crisis that may have created incentives for banks to take on excessive leverage and lower their lending standard, thus weakening bank portfolios.
Below is given the literature that is of particular interest for us as it has specifically investigated the effect of interest rate changes on banks’ balance sheet when interest rates are low or negative. Papers examining the effects of NIRP (Arteta et al., 2016; Jobst and Lin., 2016; Bech and Malkhozov., 2015) indicate that NIMs are compressed due to decline of new loans’ lending rates, while deposit rates stay sticky-downward, but in their studies they do not employ bank-level data to comprehensively check if this is indeed the case. Demiralp et al. (2016) confirm that bank performance did not suffer by introduction of negative rates in spite of expectations, yet intermediation margins have fallen slightly. They argue that as solvency of debtors improve (due to lower lending rates), the higher value of asset portfolio will neutralize
possible losses due to NIRP.
Demiralp. S., Eisenschmidt, J. and Vlassopoulos, T. (2015) examine whether there is an evidence of portfolio adjustment triggered by the implementation of negative interest rates through several channels. They found out that indeed there is a significant portfolio adjustment when banks tend to hold more non-domestic bonds and rely less on wholesale funding. Furthermore, Blot and Hubert (2016) analyze consequences of shifting to negative rates on the banking system of the euro area and conclude that the NIRP brings down the net interest margin and hence the profitability of the maturity transformation activity carried out by the banks. Thus, while in the short term, lower interest rates might raise the profitability of banks, in the medium term, their margins will start diminishing and banks in the euro area will become more responsive to monetary policy normalization.
According to Daniel Gros (2016) the effect of negative interest rates on bank profits is a priori ambiguous. It depends on the business structure of the bank, the degree of competition and the general economic conditions. The overall conclusion is that negative interest rates and low long-term rates count more as a concurrent factor rather than the central underlying cause for the low profitability of euro area banks. Other economic trends, such as he savings surplus of the euro and ongoing tightening of regulation are also likely to contribute to current situation in banking profitability.
The review of literature shows that our study is about a contemporary topic of great interest at the time of still ongoing historically low interest rates applied by major central banks around the globe, and aforementioned studies will be useful in both building the theoretical framework, as well as constructing econometric model to answer our research question.
4. Empirical methodology and Data
4. 1 Methodology
To capture the impact of NIRP on bank performance, the difference-in-differences technique is employed, which has been frequently encountered in the literature for assessment of policy effects, and over the last years to evaluate banking sector and
financial industry challenges (Calderon & Schaeck). The convenience associated with using this methodology is the possibility to employ a panel data setting to cross check a cluster of treated countries versus control cluster of untreated countries.. Below is given the baseline regression that will be used to answer our research question:
𝑌!"# = 𝛼 + 𝛽!𝑃𝑜𝑠𝑡𝑃𝑒𝑟𝑖𝑜𝑑!"+ 𝛽!𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡!"+ +𝛽! 𝑃𝑜𝑠𝑡𝑃𝑒𝑟𝑖𝑜𝑑!"∗ 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡!" + 𝛿!+ 𝜃!+ 𝛽!𝑋!+ 𝜀!"#
where
• 𝒀𝒊𝒕𝒋 is the NIM of bank i at time t in country j
• 𝑷𝒐𝒔𝒕𝑷𝒆𝒓𝒊𝒐𝒅𝒋𝒕 is a binary variable that equals 1 the year that NIRP was adopted in country 𝑗 at time 𝑡 and takes value 0 prior the time NIRP was introduced
• 𝑻𝒓𝒆𝒂𝒕𝒎𝒆𝒏𝒕𝒊𝒋is a binary variable that equals 1 if in bank I in country 𝑗 NIRP has been introduced and 0 otherwise
• 𝜷𝟑 stands for the average variation in NIM among states that underwent policy change and those that did not.
• 𝜹𝒕 and 𝜽𝒕 are year and bank fixed effects, that control potential bias in 𝜷𝟑 calculation • 𝑿𝒊 is an array of country- and bank-specific traits to control possible inter-country and inter-bank heterogeneity that might have impact on NIM. Publications studying NIM determinants are used to choose the appropriate set of control variables.
Net interest margin is calculated as the net interest income adjusted relative to the average interest-generating assets. Different strategies can be used in order to achieve projected range of targeted NIM, such as expanding interest rate spread between loans and deposits. The choice of appropriate set of control variables in this paper is substantially affected by experimental and theoretical input from extensive collection of publications on factors influencing bank profitability.
To estimate default risk, we use loan loss reserve divided by gross loans, the proxy that reflects the comparative deterioration of the key bank asset element, i.e. portfolio of assets. The riskiness level of asset portfolio at the moment of reporting is displayed by loan loss reserve, which is substantially less apt to ‘window-dressing’ of balance sheet.
Furthermore, liquidity is measured as the ratio of liquid assets to short-term funding and deposits, which is intended to secure that banks possess sufficient cash and working capital available to overcome temporary disruptions and industry-wide turmoil.
Size of the bank is computed with its total assets, which supposedly is positively correlated with NIM via the realization of economies of scale, yet it is also arguable statement, since the larger and the more productive is the bank, the smaller interest margins it can afford to request from customers. With this metric we will be able to adjust for cost differentials between banks associated with their higher capacity to diversify (Staikouras & Wood).
According to Ashraf et al., evaluation of risk exposure of bank is complicated and several proxies have been used in the literature to estimate it. In our study, we use RWA density, which reflects investments in assets with different levels of risk exposure. For instance, government bonds have lowest share in risk-weighted assets computations, while commercial loans are seen as having the highest risk weight
To gain better understanding of the competition level in the banking industry and distribution of market power, we use the Herfindhal-Hirschman Index (HHI). The HHI will get larger as the number of banks in the market declines, and also when the size discrepancy between banks becomes larger.
The country-specific variables consist of the measure of macroeconomic performance. GDP growth rate describes economic health of the country, and studies indicate that the general macroeconomic conditions in which banks run indeed affect their performance. We recognize a dual role of macroeconomic conditions on bank profits: in one respect, a steady GDP growth gives beneficial impact on bank profitability caused by higher loan demand. But, there might be an adverse impact on profitability as well, if asset supply decreases due to the growth in expenditure.
In Table 1, we provide short description of variable that will be used in our analysis and regressions later in the paper.
Table 1: Variable description
Proxy
Expected relation
Output Variable
NIM Net interest income/Average interest-generating assets
Control variables
Bank specific elements
Bank size Log of Total Assets +/-
Credit risk Loan loss reserves/Gross loans -
RWA density RWA/Total assets +/-
Liquidity Liq. assets/Short-term funding and deposits +
Industry specific factor
HHI Herfindahl-Hirschman Index +/-
Macro-economic factor
GDP growth rate GDP per capita growth (annual %) +
4. 2 Difference-in-Difference Assumptions
The appropriateness of the difference-in-differences evaluation method relies critically on the justifiability of core underlying assumptions that must be satisfied. They have been extensively discussed in the papers of Abadie (2006), Bonhomme and Sauder (2011), Donald and Lang (2007) and etc.
Underlying assumptions:
1. Implied counterfactual: Foremost, the reference group should present a sound counterfactual against treatment group.
2. Placement exogeneity: Decision to allocate banks to the treated group should be unrelated to bank performance and not determined by it, that is to say, the change in the monetary policy influence performance of the banks and not other way around. As mentioned earlier, the goal behind introduction of negative interest
rate policy was to raise penalty for keeping surplus reserves at CB, and as a result encourage banks to expand their credit supply. Thus, it is more likely to be a spillover effect, in case if banks’ profitability has been altered by policy intervention.
3. Time-invariant heterogeneity: According to Mora and Reggio, DID evaluation method is considered well-grounded and can be applied in the research analysis only if the third restrictive condition of a “Parallel Paths” assumption is satisfied. This core assumption assumes that the variation between treated and control groups, if untreated, is time-invariant. In other words, it requires the average variation in the outcome at baseline (banks’ NIM) for the treated in the absence of treatment (before the introduction of NIRP) to be similar to recorded average variation in banks’ NIM in the control group.
The validity of “Parallel Path” assumption is confirmed in Figure 1, which depicts that before the policy introduction, NIMs of adopter and non-adopter countries had similar pattern of movement, which drastically changes after policy intervention. Thus, we conclude that the “Parallel Paths” assumption holds.
Figure 1. Parallel Path assumption: Average level of NIMs among treated banks and non-treated banks in the pre-treatment and post-treatment period.
Figure 2. (a) Company profit margins %
In order to shed light to NIMs’ parallel trend development in Figure 1, we can also graph publicly listed firms’ profit margins and macroeconomic performance of the control and treated countries in Figure 2.
It is visible that the average profit margins of firms move in identical direction with average NIM of treated countries after introduction of NIRP; margins drop in 2015 and then rise to slightly lower level again in 2016. However, GDP growth rate evolves in a different pattern. Both treated and untreated countries were growing until 2015, when they reached 3%-3.3%, and then had a drop in 2016, with larger fall in NIRP-adopting countries.
According to ECB Monthly Bulletin, such discrepancy in the movements of GDP versus profits of banks and firms might be explained by the fact that in most advanced economies companies have become more multinational, which may blur the profits-GDP relationship (2007, p. 45). It is important to mention that those companies that are international and operate worldwide will be subject only to local accounting principles, implying that economic growth in developing countries will show up in the earnings of european listed companies, even though GDP growth might not be high in that region.
3. 3 Data
The bank-level data was pooled from new Orbis Bank Focus database, sourced by Bureau van Dijk. Within this database we use a history of 5 year- bank balance and performance data. The sample covers a period from 2012 to 2016, spanning both non-negative interest rate environment as well as period when NIRP was put into action. The sample period is deliberately chosen to be short, as according to Roberts and Whited, the internal validity of our empirical analysis would be jeopardized if study period was chosen wide apart from the inception of policy intervention in 2014, as many other external factors could have impact on the outcome of interest.
Since the sample of bank balance sheet data has the structure of panel data, our research framework benefits from improvements in accuracy of predicting model parameters and control over unobservable bank- and country specific factors that are taken account for when using panel model.
(commercial banks, savings banks, cooperative banks) throughout period of study. It includes 13107 banks for a period from 2012 through 2016, with a sample of 13 control group countries and 20 countries that adopted NIRP. Countries that introduced NIRP consist of Sweden, Hungary, Switzerland and Euro Area. In order to impede outliers from tampering the research, the sample is inspected for missing values and outliers. In addition, we winsorize variables with extreme outliers at level 1% and 99%. Using above-mentioned techniques we attempt to limit amount of deleted data, so that eventual distribution displays sensible shape when assessed from a statistical standpoint.
When both serial correlation and variable correlation within the country are present, then we have high probability of getting standard errors that may considerably underestimate the standard deviation of the coefficients. To solve this problem, we carry out error correction method by applying clustered standard errors to address both autocorrelation and heteroscedasticity (Borio, 2015). Relevant studies point out that endogeneity problem should not be a big concern for out empirical framework. Because, even though broad macroeconomic conditions do affect central banks’ choice of monetary policy, the performance of independent banks within sovereign state is not the key factor when selecting financial policy.
Table 2. Sample Descriptive Statistics for Treatment and Control Group
Treatment Group
Variable Obs. Mean Std. Dev. 5th
percentile 95th percentile NIM 11,403 1.91% 0.94 0.61% 3.26% Bank Size 11,499 13.26 2.01 10.71 17.02 RWA density 6,662 55.31% 20.62 27.16% 80.08% Liquidity 11,344 19.54% 25.36 3.82% 80.22% Credit risk 8,171 3.67% 4.39 7.2% 13.56% GDP growth 13,780 -0.90% 1.55 -1.75% 2.89% HHI 13,275 0.062 0.025 0.036 0.105
Control Group
Variable Obs. Mean Std. Dev. 5th percentile 95th percentile
NIM 49,854 3.59% 1.64 1.68% 5.06% Bank Size 49,925 14.17 1.60 12.61 17.57 RWA density 22,705 65.34% 15.40 38.51% 87.57% Liquidity 61,068 16.25% 2.49 1.67% 35.24% Credit risk 49,138 1.89% 1.22 0.26% 3.26% GDP growth 51,755 2.19% 0.61 1.48% 2.86% HHI 51,755 0.073 0.076 0.055 0.083 Note: Summary statistics of the main bank balance sheet elements are reported that will be used in
the empirical research. Data is collected from Orbis Bank Focus database for the period spanning from 2012 through 2016, and compose 13107 banks in the sample .
The test is performed on the sample for possible multicollinearity and high correlation that might tamper and influence the estimate of the treatment effect, the correlation matrix is constructed for the bank specific and macroeconomic variables in Table 3. We observe that the correlations are quite low, partially because of the nature of the data. The highest correlation exists between RWA density and NIM (0.3534), which it still lower than the threshold of 0.7.
Table 3. Correlation Matrix
Correlation NIM Bank
size Liquidity
RWA
density Credit risk
GDP growth HHI NIM 1 Bank size -0.2201 1 Liquidity 0.0019 -0.0035 1 RWA density 0.3534 -0.0273 0.0184 1 Credit risk 0.0050 0.0762 0.0605 -0.0309 1 GDP growth 0.1998 -0.1139 0.0146 0.1024 -0.1627 1 HHI 0.1241 0.1386 -0.0004 -0.0311 0.0188 0.1021 1
5. Empirical Results
5.1 Baseline Regression
The output of baseline specification is presented in Table 4, which reports the estimates with dependent variable NIMs, where we incrementally add a set of bank specific variables. We focus on the coefficient of NIRP-impact that denotes the average variation in NIM between NIRP-adopting group versus control group.
The coefficient on NIRP-impact in column 1 is negative, sizeable and has significance level of 5%, implying that countries that adopted NIRP witnessed a drop in NIM level of 18.97% relative to those where CBs did not shift their policy. If banks are not capable to reduce loan rates as much as deposit rates, then shifting to NIRP will most likely lead to shrinkage of net interest margin, according to regression outcome. The remaining columns provide the outcome from introducing country- and bank- specific controls, consecutively. The baseline specification output withstands well in the light of adding other control variables; the marginal effect stays statistically significant and negative, varying from -8.62% to -12.1%.
The remaining control variables are for the most part significant at accepted levels, the signs are in line with expectations. Positive correlation between RWA density and NIMs imply that the greater is the risk taken by banks, the higher NIMs are registered. In contrast, liquidity and NIM are negatively correlated. Arguably, to counterbalance for higher risks, banks with smaller amount of liquidity request higher margins from their customers. Bank size also is one of the factors driving banks’ NIMs down.
The default risk was expected to have an adverse impact on the possible losses arising from loans of inferior quality, however, regression results display significant, but positive effect on NIM. Good economic performance also contributes to NIM, however not statistically significant. Higher banking market concentration leads to better performance as well, significant at 1% level.
Table 4. Baseline Regression (1) (2) (3) (4) (5) NIRP-impact -0.1897** -0.0862*** -0.123*** -0.104*** -0.121*** (0.078) (-3.31) (-6.10) (-6.26) (-6.75) Bank size -0.0631 -0.106*** -0.145*** -0.114*** (-1.83) (-6.93) (-9.85) (-6.09) RWA density 0.00931*** 0.0101*** 0.0102*** (4.56) (3.79) (3.77) Liquidity -0.0005 0.0003*** -0.0005 (-1.04) (4.63) (-1.02) Credit risk 0.0371*** 0.0240* 0.0370*** (3.55) (2.32) (3.53) GDP growth rate 0.0030 0.0002 (0.40) (0.02) HHI 6.198*** 1.405 (8.41) (1.04) Fixed year effects Yes Yes Yes Yes Yes Fixed bank effects Yes Yes Yes Yes Yes
Adjusted R2 0.0630 0.1141 0.1429 0.1081 0.1595 # of banks 10,470 8,954 8,643 9,476 8,642 # of observations 47,760 40,592 39,367 43,143 39,326 Note: This table displays results of panel regression of Net Interest Margin (NIM) on GDP growth rate and bank characteristics. NIM is annual net interest income adjusted relative to the average interest-generating assets. Size is logarithm of bank total asset; Proxy for risk taking behavior is the RWA density; Liquidity is bank liquid assets to short-term funding and deposits ratio; Credit risk is loan loss reserve to gross loans ratio. Standard errors are displayed in parentheses. Statistical significance at the 1%, 5%, and 10% level are marked as ***, **, and *, respectively
5. 2 Liquidity channel test
We observe peculiar change in the sign of liquidity coefficient from negative to positive when RWA density is not included in column (4) Table 4. In order to have better understanding of this phenomenon, we perform additional test using 3-way interaction.
A 3-way interaction means that the interaction between the two factors (Time * Treatment) is different across the levels of the third factor (Liquidity). Thus, the
interaction term consists of 𝑇𝑖𝑚𝑒 ∗ 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ∗ 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦 , where 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦 is continuous variable dichotomized into High and Low categories with 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦|𝐻𝑖𝑔ℎ = 1 and 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦|𝐿𝑜𝑤 = 0.
Positive sign on 3-way interaction term of 0.157 can be interpreted as a relative increase in NIM among countries where central banks implemented NIRP compared to countries in which policy shift did not happen, given that those banks have high liquidity ratios.
Table 5. Regression with additional interaction term
NIM(1) NIM(2) NIM(3) NIM(4) NIM(5)
NIRP-impact -0.148** -0.108*** -0.148*** -0.142*** -0.146*** (-2.95) (-5.09) (-7.00) (-8.38) (-7.71) (NIRP-impact)* Liquidity 0.157* 0.0968 0.115* 0.0969 0.118* (2.51) (1.29) (2.31) (1.80) (2.31) Bank size -0.0756* -0.112*** -0.172*** -0.122*** (-2.17) (-7.51) (-11.75) (-6.58) Liquidity -0.129* -0.182* -0.133* (-2.31) (-2.34) (-2.39) RWA density 0.0082*** 0.0093*** 0.0093*** (3.71) (3.62) (3.60) Credit risk 0.0394*** 0.0271* 0.0394*** (3.78) (2.53) (3.75) GDP growth rate 0.0045 0.0001 (0.60) (0.01) HHI 6.224*** 1.713 (8.53) (1.25) Adjusted R2 0.0190 0.1404 0.2087 0.1362 0.2181 # of banks 12,95 6,393 6,17 12,113 6,17 # of observations 60,996 29,004 28,192 57,222 28,158
The coefficient on 3-way interaction term stays positive across all columns, while individual terms NIRP-impact and Liquidity remain negatively related to NIMs. This implies that liquidity plays crucial role in the way banks are affected by NIRP, as individual terms have the opposite sign when they are regressed on NIM independently from each other.
The sign of remaining variables has not changed after adding 3-way interaction term, hence they affect bank NIMs the same way as described in the first regression Table 4.
6. Conclusions
In 2014, the conventional monetary policy entered unchartered territory with Denmark’s central bank crossing the assumed “Zero Lower Bound” and charged a negative interest rate on deposit. Since then, central banks in Euro area, Sweden and Switzerland also started this experiment with NIR.
Throughout our analyses, we worked with the central question of whether banks’ net interest margins have been affected by implementation of NIRP in the countries where policy intervention was made. It is important to mention that in majority of countries under research, NIRP was introduced together with several other monetary policy tools. Thus, our data reflects the impact of alliance of monetary policies, and it is hard to extract the causal effect of NIRP alone.
The NIRP was aimed at improving real spending and promote growth in credit supply. However, this monetary policy did raise dispute with skeptics identifying possible aspects that might hinder the transmission mechanism to reach higher volumes of loan supply. The most frequently used argument they referred to is that NIRP might compress NIMs and, hence will restrict banks’ capacity to provide loans. Empirical research on the actual impact of policy intervention on banks’ performance is scarce. Thus, in this paper we take a look at how bank margins had behaved right after NIRP introduction, also displaying graphically that “Parallel Path” assumption is in line with unobservable time-invariant heterogeneity. According to both regression results and graphic representation, countries that implemented NIRP encountered decline in net interest margins in year 2015 compared to non-adopter countries, where interest rates had not move below zero. This result stands up in the face of inclusion of various bank- and country-specific variables.
Finally, our findings suggest that variation in the provision of liquidity services create heterogeneity across banks in the way they react to ongoing reduction in interest rates. We discover that the more liquid assets banks keep on their balance sheet, the
better they perform in terms of net interest margins.
Overall reduction in net interest margins might induce banks to invest in more risky assets in order to accumulate higher returns. This in turn might bring down cushioning potential of banks and lead to higher financial sector vulnerability. For that reason, preventive measures should be implemented by policy-makers to ensure financial stability.
7. Bibliography
Abadie, A. (2006). Semiparametric Difference-in-Differences Estimators, The Review of Economic Studies, 72, pp. 1-19.
Ahmad, R., & Matemilola, B. T. (2013). Determinants of Bank Profits and Net Interest Margins. In Emerging Markets and Financial Resilience (pp. 228-248). Palgrave Macmillan, London.
Alessandri, P., & Nelson, B. D. (2015). Simple banking: profitability and the yield curve. Journal of Money, Credit and Banking, 47(1), 143-175.
Andries N., & Billon, S. (2016). Retail bank interest rate pass-through in the euro area: an empirical survey. Economic Systems, 40, pp. 170-194.
Ashraf, B. N., Arshad, S., & Hu, Y. (2016). Capital regulation and bank risk-taking behavior: Evidence from pakistan. International Journal of Financial Studies, 4(3), 16.
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.
Baltagi, B. (2008). Econometric analysis of panel data. John Wiley & Sons.
Bindseil, U. (2016), Evaluating monetary policy operational frameworks, Speech at the Jackson Hole conference on 31 August 2016.
Blot, C., & Hubert, P. (2016). Negative interest rates: incentive or hindrance for the
banking system? (No. info: hdl: 2441/4becfeb5av97abu32kpfiu3srd). Sciences Po.
Bonhomme, S., and U. Sauder (2011). Recovering distributions in difference-in- differences models. The Review of Economics and Statistics, 93(2), pp. 479-494.
Brunnermeier, M. K., & Koby, Y. (2016). The reversal interest rate: An effective lower bound on monetary policy. Unpublished paper, Princeton University.
Claessens, S., Coleman, N., & Donnelly, M. (2017). “Low-For-Long” interest rates and banks’ interest margins and profitability: Cross-country evidence. Journal of
Financial Intermediation.
Coeuré, B. (2016). Assessing the Implications of Negative Interest Rates. Speech at Yale Financial Crisis Forum. 28 July, 2016.
Demirguc-Kunt, A., Laeven, L., & Levine, R. (2003). Regulations, market structure,
institutions, and the cost of financial intermediation (No. w9890). National Bureau of
Economic Research.
Donald, S., and K. Lang (2007): Inference with difference-in-differences and other panel data, The Review of Economics and Statistics, 89(2), pp. 221-233.
ECB (2007). “Economic and Monetary Developments”. Monthly Bulletin, September.
ECB (2016a). Euro area financial institutions, Financial Stability Review, May.
Gambacorta, L., Hofmann, B., & Peersman, G. (2014). The effectiveness of unconventional monetary policy at the zero lower bound: A cross‐country analysis. Journal of Money, Credit and Banking, 46(4), pp. 615-642.
Jobst, A., & Lin, H. (2016). Negative Interest Rate Policy (NIRP): Implications for
Monetary Transmission and Bank Profitability in the Euro Area. International
Monetary Fund.
Kerbl, S., & Sigmund, M. (2016). From low to negative rates: an asymmetric dilemma. Financial Stability Report, (32), 120-137.
Marinković, S., & Radović, O. (2014). Bank net interest margin related to risk, ownership and size: an exploratory study of the Serbian banking industry. Economic
research-Ekonomska istraživanja, 27(1), 134-154.
Mirzaei, A., Moore, T., & Liu, G. (2013). Does market structure matter on banks’ profitability and stability? Emerging vs. advanced economies. Journal of Banking &
Finance, 37(8), 2920-2937.
Molyneux, P., Reghezza, A., & Xie, R. (2018). Bank Profits and Margins in a World
of Negative Rates (No. 18001).
Mora, R., & Reggio, I. (2012). Treatment effect identification using alternative
parallel assumptions. Universidad Carlos III de Madrid. Departamento de Economía.
Pandalai, T., Chen, W., Daly, C., Huang, Q., Lowe, D. & Hermann, F.V. (2017). Negative Interest Rates: A comparative study of implementation and effects across four central banks. Columbia University, School of International and Public Affairs Capstone Project, pp. 1-147.
Petria, N., Capraru, B., & Ihnatov, I. (2015). Determinants of banks’ profitability: evidence from EU 27 banking systems. Procedia Economics and Finance, 20, 518-524.
Roberts, M. R., & Whited, T. M. (2013). Endogeneity in empirical corporate finance1. In Handbook of the Economics of Finance, Vol. 2, pp. 493-572.
Staikouras, C. K., & Wood, G. E. (2004). The determinants of European bank profitability. International business and economics research journal, 3, pp. 57-68.
