Monetary policy:
a delicate balance of employment and stability
Martijn Vruwink* Supervisor: Dr. J.O. Mierau
June 13, 2016
Abstract
Subsequent to the financial crisis, governmental intervention has led to a low interest rate environment. Aggregate investment is worrisome, however, governments need to take the risk-taking channel of monetary policy into account when deciding stance of monetary policy. By performing a fixed effect and generalized method of moments technique for a representative sample set of US banks during the period 1995-2014, this dissertation confirms the presence of a risk-taking channel for the short and long-term interest rate. The effect of the slope of the yield curve remains controversial. The
findings of leverage confirm the asymmetric reaction function regarding risk-taking. To prevent a future crisis, regulators need to monitor loan growth not only on an individual level but also the aggregate level.
2 1. Introduction
The financial crisis has rekindled the debate of the contribution of banks to the financial system. Financials crises tend to exhibit credit growth in the buildup and a sudden decline afterwards according to Dell'Ariccia, Detragiache, and Rajan (2008) and as Brunnermeier (2008) suggests this development destabilized the financial world. Credit seems partially responsible for the collapse of the financial system initiated in 2007. The responsible authorities were powerless and could only watch how events unfolded and are looking for an answer to prevent future catastrophes. The relatively new Basel III provides an additional cushion for a potential future collapse. But what if risk-taking is procyclical? Are these regulations adequate of averting a new crisis?
The consequences of the financial turmoil are still felt worldwide and aggregate investment is low. By launching a Quantitive Easing (QE) program a government aims to revitalize investments. The Federal Reserve started QE in 2008 and stated the objective is to reduce long-term interest rates in order to spur economic activity.1 The
case of Japan acts as a precedent in which a similar policy did not result in the desired steady inflation and high economic growth (Okina and Shiratsuka, 2004). There are signals that the low interest rate encourages reckless investment behavior which is a major factor of the last crisis. Is the possible risk-taking channel a factor which needs to be taken into account for setting new monetary policy?
1.1. Research focus
In essence, banks transform short-term deposits (liabilities) into long-term loans (assets). Mink (2011) states that this maturity transformation enables banks to borrow against lower interest rates than their shareholders. The cost advantage of this maturity transformation is the foundation of the banking industry and is a great source of income for traditional banks. In accordance with the maturity transformation, Allesandri and Nelson (2015) acknowledges the importance of the long-term interest rate as an indicator for its profitability. The short-term interest rate did not prove to be significant due to frictions in non-negligible loan pricing (the stickiness of loan pricing). Therefore, it seems intuitive banks react to changes in interest rate.
3
The experts argue that banks and credit were one of the main culprits of the financial turmoil. Brunnermeier (2008) proves how bank risk-taking is intensified in the period prior to the financial crisis. In addition, Borio (2008) acknowledges that bank risk-taking moves parallel with the performance of the world economy. Borio’s argument is in contrast with Rajan (2006) who argues that in a low interest rate environment banks search for yield, since a low interest rate puts pressure on the income of maturity transformation. A lower income would induce banks to lower lending standards and consequently take additional risks. Additionally, a sudden increase in credit was observed prior the financial turmoil. Levering a bank balance sheet enables banks to issue more loans, and earn more profits at the risk of insolvency which makes the
financial markets more instable. Dell’Ariccia, Laeven, and Marquez (2014) conclude that the level of leverage (leverage and capitalisation are used interchangeable) might have an impact on the size of risk-taking.
To provide clarity regarding the role of interest rates this dissertation
investigates the risk-taking channel and aims to show the role of credit in the area of risk-taking. Despite, the insignificance of the short-term for a bank’s profitability (Allesandri and Nelson 2015), I investigate the effect of the short-term (hypothesis 1), the long-term (hypothesis 2) and the slope of the yield curve (hypothesis 3) on the risk-taking of banks. Furthermore, I investigate the effect of leverage on risk-risk-taking (hypothesis 4). This dissertation exhibits mixed results for the risk-taking channel of monetary policy approximated by abnormal loan growth and Z-score. For readability the results of the Z-score are found in appendix A.2. The results suggest that different levels of leverage trigger an asymmetrical response regarding risk-taking. By employing a time period of 1995-2014 the analysis covers the low and high interest rate environment. To obtain robust results survivorship bias, endogeneity and heterogeneity are taken into account.
4
regulators should incorporate the risk-taking channel in their policy and recommends which signals might indicate a possible credit boom and risk-taking.
The remainder of this paper is organized as follows. Section 2 contains the theoretical framework, including existing literature on the subject. Section 3 contains the data collection and the employed empirical methodology. Section 4 describes the findings and section 5 concludes.
1. Literature review
Although, the risk-taking channel of monetary policy is heavily researched. The bridge between macroeconomic and banking models is not extensively researched, an exception is the work of Dell ‘Ariccia, Laeven, and Marquez (2014). To bridge the two models I briefly elaborate on the effect of monetary policy on the real economy and the impact of credit.
2.1. The money view and the financial accelerator
The first link regarding real output changes as a consequence of monetary actions is established by Friedman and Schwartz (1963). The implication of their work is that monetary policy should control the money supply since it affects the US economy and in particular the behavior of the business cycle. Their findings were confirmed by, inter alios, Bernanke and Blinder (1992).
5
The consequence is that the money view considers banks not to affect the real economy. Fama (1980) argues supporters of the interest rate channel see no need to control for deposit creation or security purchasing activities of banks to achieve an equilibrium with respect to prices and real activity. However, Fama (1980) and
Bernanke and Blinder (1995), inter alios, argue the money view contains several cavities which requires an addition to the existing money view
The credit channel is the answer to the cavities and is considered to amplify and propagate the conventional interest rate effects of the money view (see, e.g., Bernanke and Blinder, 1995; De Graeve, 2008). The channel advocates the presence of
asymmetrical information, which violates the essence of Modigliani and Miller, resulting in an external finance premium which is the wedge between internal, i.e. retained
earnings, and external financing. The size of the premium is considered to be the result of bank lending and the balance sheet conduit and moves parallel with the open market interest rates (Bernanke and Blinder, 1995).
The lending conduit argues a tightened monetary policy results in decreasing deposits since the opportunity costs have increased and therefore influences the
liabilities side of a bank. In the absence of reserves and the inability to liquidate assets the supply of loans will decline. The balance sheet channel is inversely related to the borrower’s riskiness, derived by its balance sheet composition. A higher interest rate results in reduced cash flows, a reduced net worth, drop in loans, and decline in aggregate demand.
2.2. The channels of bank risk-taking
6
investment since internal risk limits might not adjust accordingly. The severity of this balance sheet channel depends on the multiplier effect, which in turn is affected by the degree of rising asset prices as stated by Kiyotaki and Moore (1997). The asset prices are prone to changes due a more extensive period of low interest rate and the magnitude of the interest rate change (Borio and Zhu, 2012). As a result of the widely spread VaR methodology a change in interest rate could have an immense impact on the individual and therefore aggregate risk-taking.
A second channel through which interest rate influences bank risk is by the search for yield. The search for yield channel was first identified by Rajan (2006), who argues that a low interest rate induces relatively riskier behavior for pension funds, insurance companies, and banks, if expansionary monetary policy lowers the yield on short-term safe assets relatively to its long-term liabilities. Dell’Ariccia, Laeven, and Marquez (2014) argue a prolonged time of low interest rate could force financial
institutions to alter their portfolio to prevent default on future long-term commitments. To lower the probability of future default, financial institutions swap safe, low yield assets for riskier, higher yield assets.
Furthermore, Dell’Ariccia and Marquez (2006) find that when banks face adverse selection and the bank’s cost of capital has decreased due to expansionary monetary policy, this could lead to a credit bubble or lower lending standards. This is due to the decreased incentive of banks to filter bad borrowers as their cost of funding have
decreased. Concluding, a low interest rate environment could induce banks to switch to risky assets to diminish the probability of future default and would decrease cost of capital which diminishes bank incentive to screen out bad borrowers.
The third channel operates through the perception of stakeholders. Borio and Zhu (2012) and Gambacorta (2009) argue that moral hazard issues arise if communication of policymakers affect the perception of stakeholders regarding risk. Borio and Zhu (2012) argue this Greenspan’s put has an asymmetric reaction function for banks. The
7
The vast majority of literature hints towards a negative relationship between risky behavior and the interest rate environment. However, De Nicoló, Dell’Ariccia, and Laeven (2010) suggest the link between monetary policy and risk-taking is more
ambiguous. A lower interest rate increases the value of a bank’s assets. Consequently the equity stake of a bank increases and potentially mitigates the moral hazard problem between a bank and its depositors, hence risk-taking is decreased. Nevertheless, the analysis is theoretical and not yet proven in empirical literature.
The theoretical analysis above shows the potential impact of a changing equity base, and therefore its leverage and consequently its riskiness. Dell’Ariccia, Laeven, and Marquez (2014) argue the amount of leverage of a bank influences the size of risk-taking as a consequence of change in the monetary policy. De Nicoló, Dell’Ariccia, and Laeven (2010) argue leverage indicates the degree of skin in the game and this will influence the effect of risk-taking in the short-term, since leverage is not easily adjustable. Low
leveraged banks (high capitalisation) exhibit a negative relationship while high leveraged banks (low capitalisation) show a positive relationship, which implies that highly leveraged banks will decrease risk-taking in expansionary monetary policy. To empirically test the effect of leverage a fixed effects regression for differently capitalized banks is performed.
2.3. Previous findings concerning the risk-taking channel
The literature regarding the risk-taking channel of monetary policy is hinting towards a negative relationship. Empirically, Delis and Kouretas (2011) find a
significant negative relationship between risky assets over total assets and the short and long-term interest rate. Altunbas, Gambacorta, and Marques-Ibàñez (2014) investigate the link between low interest rates and bank risk-taking by examining the expected default frequency of 600 listed banks operating in the United States and Europe over the period of 2001-2008. Altunbas, Gambacorta, and Marques-Ibàñez (2014) include the slope of the yield curve in the regression analysis. Nevertheless, the slope’s coefficient transpired to be insignificant. By employing listed banks, the representativeness for the total banking industry has decreased significantly.
8
central bank. They find robust evidence that low short-term interest rates soften lending standards for firms and households, which is amplified by the period of the low interest rate. Such a robust result has not been found for the long-term interest rate. Dell‘Ariccia, Laeven, and Marquez (2014) combine responses to the Federal Reserve’s Terms of
Business Lending Survey and individual bank loan level data and confirm the existence of a risk channel of monetary policy in their paper. The most direct links of a risk-taking channel is proven by the work of Jiménez, Ongena, and Peydró (2014) and Ioannidou, Ongena, and Peydró (2015) which both employ detailed credit register of respectively the Spanish and Bolivian regulators. The first concludes that monetary policy affects the composition and supply of credit. Interestingly, they do not find a relationship between the ex-ante risk taking and the long-term interest rate. The latter discovers that lower policy rates induce banks to grand riskier loans to less creditworthy counterparty’s. In addition they conclude that banks do not price in the extra risk they take.
The existence of a risk-taking channel appears to be present in various countries for the short-term interest rate. However, the literature is divided regarding the effect of the long-term interest rate and the slope of the yield curve. Importantly, Bruno and Shin (2015) among others argue that the short term interest rate is more sensitive to changes in monetary policy and therefore is of more importance for regulators.
This review of literature describes the dynamics of the money and credit view regarding the supply of loans on a macroeconomic level, the operative risk-taking
channels and empirical literature regarding bank risk-taking. The ensemble of academic literature gives insight into the destabilizing factor of credit in the form of a financial accelerator regarding the macro economy and for individual banks allowing them to supply additional loans. The operative risk-taking channels suggest a negative
relationship between risk-taking and the interest rate environment. It can be inferred from the review of literature that expansionary monetary policy will reduce the cost of capital to boost aggregate investment while simultaneously pressurizing the maturity transformation and thereby inducing banks to exhibit relatively riskier behavior.
3. Research methods
9
advantages and disadvantages. The richness of data for the selected banks is advantageous. However, by limiting the sample to listed companies the
representativeness for the total US Banking industry declines drastically. To obtain robust, and general results for the total US banking industry, I strive for a bare
minimum of selection criteria to draw general conclusions about the risk-taking channel of monetary policy in the US. The selected period for this analysis concerns 20 years of observations (1995-2014). Missing data before 1995 and the unavailability of data for 2015 yielded to this period. Appendix A.3. shows the distribution of observations for the sample period.
The data set consists mostly of bank holding companies (23.56%), commercial (68.20%) and saving banks (7.42%). According to Delis and Kouretas (2011) investment banks are not influenced by the maturity transformation. However, they constitute for less than 1% of the sample and therefore I assume that this is of no consequence for the result. Furthermore, bank holding companies do not necessarily engage in traditional banking activities according to Foos, Norden, and Weber (2009). Which would make them less exposed to the maturity transformation. However, excluding bank holding companies would make the sample less representable for the total US banking industry. The effect of interest rate per specialisation is found in appendix A.4.
Throughout this dissertation two data sets coexist to tackle the survivorship bias. The active sample potentially underestimates the risk-taking channel since the
survivorship bias could be present in this sample. Banks which have defaulted in the sample period could have been triggered by the interest rate environment and excluding those leads to a decrease in power of the results, since I am foremost interested in the quality of credit which potentially leads to possible defaults. For the results of a fixed effects regression concerning defaulted banks see appendix A.5. Therefore, to prevent underestimation of the results every main regression is regressed on the full and active sample whenever possible.
3.1. Bank risk-taking
10
loans on an individual basis. Ioannidou, Ongena, and Peydró (2015) and Jiménez, Ongena, and Peydró (2014) are considered to directly link monetary policy to the risk attitude of individual banks. However, the majority of literature, and this paper included, is incapable of obtaining individual loan decisions. Due to the inability to collect flawless data the results will be slightly distorted but will still touch upon the dynamics of the risk-taking channel.
Foos, Norden, and Weber (2009) explore Abnormal Loan Growth (ALG) of a bank as a proxy for risk–taking. Abnormal loan growth indicates the possible lowering of lending standards. Köhler (2012) argues that banks who exhibit higher loan growth than competitors could attract customers which have not received a loan by other banks because they asked for out of proportion loan rates or are under collateralised relative to their creditworthiness. In addition, the relationship between abnormal loan growth and lagged loan losses is widely recognized. (see, e.g., Sinkey and Greenawalt, 1991; Hees, Grimes, and Holmes, 2009) while the found cross-sectional heterogeneity cannot be fully explained by macroeconomic factors and thus it is reasonable to state lending standards were lowered.
I assume abnormal loan growth is the most adequate proxy for credit risk. However, one minor drawback is the assumption that abnormal loan growth could be fueled by mechanical factors for instance mergers and acquisitions. By truncating the observations of abnormal loan growth and other variables at a 1% level, I follow the procedure of Foos, Norden, and Weber (2009) to control for this external loan growth.
Abnormal loan growth (ALG) is calculated through the difference of the annual loan growth of a bank and the median loan growth of that specific country and year. The sample consists of only one country which results in the following formula to calculate the abnormal loan growth:
𝐴𝐿𝐺𝑖,𝑡= 𝐴𝑛𝑛𝑢𝑎𝑙 𝑙𝑜𝑎𝑛 𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡− 𝑀𝑒𝑑𝑖𝑎𝑛 𝑙𝑜𝑎𝑛 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 (1)
Where ALGi,t denotes the abnormal loan growth of bank i at time t.
11
ratio divided by the standard deviation of the return on assets over the period 1995-2014.
Z − scorei,t=
ROAi,t+ CARi,t
SDROAi
(2) ROA is the return on assets and CAR the ratio of total equity over total assets of
12
Graph 1 Development of the average abnormal loan growth and Z-score over the sample period
This graph depicts the development of the two main variables: abnormal loan growth and the Z-score over the sample period. The left y-axis denotes the average value of the abnormal loan growth. The y-axis on the right depicts the scaled average Z-score for the sample and the x-axes denotes the year of the observations
13
3.2. Interest rates and slope
As opposed to the measures of risk-taking the literature isharmonized regarding the choice of measures to represent the short and long-term interest rate. Bernanke and Blinder (1988) argue the federal funds rate is a competent measure of monetary policy or any interest rate derived from it. In addition, Ioannidou, Ongena, and Peydró (2015) and Jiménez, Ongena, and Peydró (2014) who study respectively the Bolivian and Spanish banking sector argue interest rate is endogenous to the macroeconomic conditions. Therefore, interest rate is considered endogenous in this dissertation.
The endogenous short-term interest rate is approximated by the 3-month US London Interbank Offered Rate (LIBOR) which is charged for selected banks in the United States if a bank decides to borrow short-term funds from other banks. A robustness analysis using the federal funds rate as a proxy for the short-term interest rate is found in appendix A.6. The long-term interest rate is approximated by the 10-year US government bond. By using these interest rates I follow the procedure of Delis and Kouretas (2011). To calculate the slope I follow the procedure of Altunbas,
Gambacorta, and Marques-Ibàñez (2010) and Maddaloni and Peydró (2011) who state the slope is the difference of the annual average of the short and the long-term interest rate. Appendix A.7. denotes the developments of the different interest rates over the sample period
3.3. Control variables
By incorporating control variables on a bank and country level the omitted variable bias is controlled for. Regarding the bank control variables I follow the
procedure of Delis and Kouretas (2011), and Demirguc-Kunt, Detragiache, and Tressel. (2008), who include size, efficiency, capitalisation, and profitability into their
specification. According to Konishi and Yasuda (2004) size decreases risk-taking. Larger banks, in terms of total assets, are more capable of diversifying risks. Moreover, size is considered to be predetermined as banks are aware of their relative size at the moment of decision making (Delis and Kouretas, 2011). Efficiency, approximated by the cost-to-income ratio, positively affects bank risk-taking, since efficient banks are more capable in managing risks. In addition, Delis and Kouretas (2011) argue efficiency is endogenous since, higher risks could also explain a higher efficiency, as a higher interest rate
14
as has been argued by Boyd and Nicoló (2005). Furthermore, capitalisation is treated as endogenous by Delis and Kouretas (2011). Dell’Ariccia, Laeven, and Marquez (2014) state leverage is mediating the effect of bank risk-taking. The relationship between profitability and risk-taking is ambiguous according to Delis and Kouretas (2011), a higher profitability could indicate higher risk, however too much risk could reduce profitability. To ensure robustness I treat profitability as endogenous.
Regarding the control variables on a country level, Laeven and Levine (2009) argue the regulatory environment is of influence for the risk-taking. However, the
sample of this dissertation comprises of only one country, therefore the need for an array of macroeconomic variables has declined. Furthermore, Čihák et al. (2012) argue
regulations has not changed significantly post crisis and therefore I assume the growth of the gross domestic product and the inflation is capable of capturing the
15 Table 1 descriptive statistics
This table depicts the summary statistics used in the subsequent analysis. The full sample contains balance sheet info of 11,836banks for the period 1995-2014. Loan growth describes the abnormal loan growth of bank i, which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. The Z-score represents the solvency risk associated for a bank, the solvency is the capital to asset ratio of bank i minus the return on assets for the full sample divided by the standard deviation of the return on assets for the whole sample. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Inflation depicts the annual rise in consumer prices as measured by the CPI.
Variables Obs. Mean Std. Dev. Min. Max.
Table 2 Correlation matrix
This table depicts the correlations of the variables employed in this dissertation. The full sample contains balance sheet info of 11,836banks for the period 1995-2014. Loan growth describes the abnormal loan growth of bank i, which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. The Z-score represents the solvency risk associated for a bank, the solvency is the capital to asset ratio of bank i minus the return on assets for the full sample divided by the standard deviation of the return on assets for the whole sample. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Inflation depicts the annual rise in consumer prices as measured by the CPI.
Loan growth Z-score Short-term rate Long-term
rate Slope Size
Lagged
profit. Capital. Efficiency GDP Inflation
17
3.4. Econometric specification
As stated in the hypothesis I intend to regress the short-term, long-term interest rate and the slope of the yield curve on abnormal loan growth and the Z-score. The following model captures the basic model in this dissertation:
ri,t = α + β1irt+ β2bi,t+ β3mi,t+ εi,t (3)
The risk-taking variable r, of bank i at time t is a function of the interest rate variable ir at time t and an array of bank (b), and macroeconomic (m) control variables of bank i at time t as described in the control variables section. The desire to include control
variables on the bank and country level arises from the discussion of several papers (see, e.g., Delis and Kouretas, 2011; Foos, Norden, and Weber 2009) who argue exclusion leads to the omitted variable bias, which generate potentially distorted test results.
To estimate the relationship I employ two techniques: a fixed effects regression and a dynamic two step system Generalized Method of Moments (GMM) panel estimator. The employed fixed effects regression corrects for autocorrelation and contains robust standard errors on a bank level. The latter is proposed by Blundell and Bond (1998) and deals with potential dynamic dependent variables, endogeneity of variables in the model and is robust for autocorrelation, heteroskedasticity, and the Nickell bias.
3.4.1. Fixed effects
The econometric model, as mentioned in the previous section, aims to capture all heterogeneity in the data set. However, unobserved heterogeneity could still be present and influence results of the regression (see, e.g. Gambacorta and Mistrulli 2014). Although the sample is limited to one country there are multiple potential sources of heterogeneity, for instance unobserved bank heterogeneity. Delis and Kouretas (2011) argue off-balance sheet items influence the risk-taking statistic. Although, the sample consists of one country this sample consists of multiple states which are autonomous to a certain extent. Delis and Kouretas (2011) include regulatory variables to prevent omitted variable bias. Therefore, there is a desire to control for state heterogeneity. By employing a fixed effects regression I assume state fixed heterogeneity is controlled for.
A fixed effects regression is preferred to a random effects model since the
18
In addition the Hausman test exhibits a preference towards a fixed effects regression. The risk-taking literature acknowledges the presence of autocorrelation in the field of risk-taking. The Wooldridge (2010) tests confirms autocorrelation. To control for autocorrelation the first difference approach is adopted. By first differencing the dependent variable the autocorrelation is compensated. The result of first differencing the dependent variable is shown in appendix A.8. The fixed effect regression has the functional form of:
∆ri,t = α + β1irt+ β2bi,t+ β3mi,t+ γi+ ui,t (4)
where Δ denotes the first difference of the dependent variable r, and γi denotes the
individual fixed effects intercept.
The sample exhibits heteroskedasticity according to the Breusch-Pagan test for homoskedasticity. However, the consequences of homoskedasticity are mild according to Gujarati (2009), to control for potential heteroskedasticity the fixed effects regression adopts robust standard errors clustered on a bank level2.
3.4.2. Generalized method of moments
A fixed effects regressions could be an inappropriate estimation technique for risk-taking. Delis and Kouretas (2011) dispute why risk-taking is possibly dynamic in nature and thus persists over multiple years. Risk-taking could persist due to increased competition. However, Boyd and Nicoló (2005) argue banks become more risky as their markets become more concentrated. Secondly, new loans are not necessarily given to new clients, existing clients also engage in new loans. The issuance of new loans to existing (risky) clients could persist over years. In addition, risk-taking is affiliated with the business cycle and its macroeconomic shocks which generally last longer than one year. Lastly, Delis and Kouretas (2011) state that that moral hazard issues are persistent due to existing regulations for instance depositor insurance.
Moreover, a system GMM is a popular estimation technique within the risk-taking literature (see, e.g., Altunbas, Gambacorta, and Marquéz-Ibàñez, 2010; Foos, Norden, and Weber, 2009). Mainly because a dynamic GMM controls for dynamic
2 The inclusion of time fixed effects does not alter the results and therefore are not included in the
19
dependent variables, endogenous predictors, unobserved heterogeneity, and the Nickell (1981) bias. Thus, in addition of the fixed effects estimator a generalized method of moments (GMM) is used for analyzing the nexus of risk-taking and monetary policy.
The proposed GMM model by Blundell and Bond (1998) is used in this dissertation.
ri,t= α + δ(ri,t−1) β1irt+ β2bi,t+ β3mi,t+ εi,t (5)
Ri,t-1 denotes the lagged dependent variable and δ represents the speed of
convergence of the dependent variable to its normal average level. Delis and Kouretas (2011) conclude a value between 0 and 1 would suggest the persistence of risk. It is argued in the article by Arellano and Bond (1991) that first order autocorrelation is of no concern for the estimators. The paper argues there is no autocorrelation if the test for first order autocorrelation is negative and there is no evidence of second order
autocorrelation. The results suggest the absence of second order autocorrelation. By performing a Difference-in-Hansen test, Z-score, abnormal loan growth, the interest rate and profitability transpired to be endogenous Therefore, the endogenous variables are instrumented by their second lag and up in the GMM model, which is common practice in the literature (see, e.g., Foos, Norden, and Weber, 2009).
4. Findings
20
4.1. Results
Table 3 depicts the results of the fixed effects regression concerning ALG. The short-term interest rate transpired to be insignificant. However, the long-term and slope are negatively correlated with the dependent variable. Interestingly, the coefficient of the long-term interest rate is considerably greater than the short-term interest rate. The slope of the yield curve is significant at the 10% for the full sample and at the 5% for the active sample. As is shown in the table, size is positively
correlated for all the regressions, similarly the coefficients of profitability,
capitalisation, and efficiency are positive. Interestingly, the coefficient for size and profitability is relatively large and is observed for roughly all regressions in table 3. Implying a change has a relatively large impact on risk-taking.
Table 4 denotes the results of the Generalized Method of Moments estimation for abnormal loan growth. The lagged variable of loan growth in the dynamic GMM transpired to be significant for all the regressions and its coefficient is fairly constant but relatively low in comparison with the Z-score. The results of the system GMM regarding interest rates largely coincides with the fixed effects regression. The coefficient of the short-term interest rate the relationship is negative and significant at the 1%, while the fixed effects regression found no significant result. The long-term interest rate is negative and significant for both samples. However, the coefficient is smaller than the short-term interest rate. Implying it has less impact on risk-taking than the short-term. The slope of the yield curve is large, positive, significant and in contrast with the fixed effects regression. Implying a decrease of the yield curve by 1% the risk-taking of a bank would decrease with 0.567%, which is fairly large.
21
4.2. Economic impact
Delis and Kouretas (2011) argue risk-taking is highly persistent, although the lagged dependent variable in the GMM is significant the coefficient of 0.141 is no evidence for high persistence. Abnormal loan growth relies solely on new loans and therefore the risk-taking could quickly return to equilibrium. To return to
equilibrium a bank needs to decrease the issuance of new loans to the median of that particular year. Delis and Kouretas (2011) however, use risky assets over total assets in which old loans are taken into account. Prematurely ending existing loans is complicated and therefore the ratio is more rigid due to this inability. Furthermore Dell’Ariccia, Laeven, and Marques (2014) argue risk-taking is procyclical, to
investigate the procycliality different periods in the sample are regressed to investigate changes in sign for the relevant variables see appendix A.9.
The results of the two techniques confirm the presence of a risk-taking channel as described in the literature (see, e.g. Borio and Zhu, 2012; Delis and Kouretas, 2011). The short-term interest rate is insignificant for fixed effects
regression. However, the GMM proves the existence of a negative relationship for an endogenously treated short-term interest rate. The impact is that banks increase their supply of loans by 0.357% above the median on average if the short-term interest declines by 1%. Monetary policy is therefore capable of influencing the risk-taking of individual banks.
22
Mink (2011) expects risk-taking to increase when the slope is steep, since the relative cost advantage of borrowing is greater. However, Gambacorta (2009) argues the impact of the slope is ambiguous. A steeper slope could incentivize banks to increase risk-taking or could decrease additional risk-taking since profits have already increased. The fixed effects found a weakly significant negative sign while the GMM finds a positive sign. Possibly the endogenous treatment of interest rates alters the sign of the slope and the true effect of the slope remains unknown. A fixed effects regression during periods which exhibit a steep slope can be found in appendix A.10. The robustness analysis shows different effects for two periods which are
characterized by relatively steep slopes.
Size is controversial in my results and in the risk-taking literature. De Nicoló (2000) states that there is a positive and significant relationship between failure probabilities and bank size in the United States. Banks consider themselves too big to fail and therefore exhibit additional risk-taking. Delis and Kouretas (2011), inter alios, claim larger banks are more capable of diversifying their risks and therefore a negative sign is expected.
Moreover, the impact of profitability is controversial, for a discussion on the subject see Delis and Kouretas (2011). A positive and relative large coefficient suggests that profitability is utilized for issuing new risky loans. No convincing evidence is presented in the results for the influence of profitability on risk-taking.
23 Table 3 Abnormal loan growth fixed effects regression
This table depicts the results of the fixed effects regression regarding abnormal loan growth. The full and active sample contain balance sheet info of 11,836 and 6,358 banks respectively, for the period 1995-2014. Loan growth describes the abnormal loan growth of bank i, which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3) (4) (5) (6)
Variables Loan growth Loan growth Loan growth Loan growth Loan growth Loan growth
Short-term rate -0.0372 -0.0324 (0.0493) (0.0558) Long-term rate -0.342*** -0.368*** (0.0909) (0.102) Slope -0.124* -0.175** (0.0651) (0.0722) Size 1.832*** 1.401*** 1.837*** 1.894*** 1.373*** 1.857*** (0.266) (0.320) (0.257) (0.276) (0.342) (0.264) Profitability 0.594* 0.590* 0.592* 0.391 0.388 0.389 (0.352) (0.350) (0.351) (0.254) (0.253) (0.253) Capitalisation 0.120** 0.102** 0.120** 0.195*** 0.173*** 0.194*** (0.0507) (0.0512) (0.0503) (0.0543) (0.0555) (0.0539) Efficiency 0.0854*** 0.0819*** 0.0850*** 0.0899*** 0.0857*** 0.0890*** (0.0186) (0.0187) (0.0186) (0.0145) (0.0147) (0.0145) Economic growth 0.347*** 0.314*** 0.396*** 0.142** 0.0998* 0.204*** (0.0576) (0.0525) (0.0563) (0.0587) (0.0530) (0.0562) Inflation -0.0175 0.143 -0.277** 0.143 0.344*** -0.190 (0.146) (0.119) (0.133) (0.165) (0.129) (0.150) Constant -32.32*** -25.67*** -31.80*** -33.15*** -25.34*** -31.85*** (4.083) (4.964) (4.085) (3.765) (4.884) (3.719) Observations 121,169 121,169 121,169 83,207 83,207 83,207 R-squared 0.004 0.004 0.004 0.004 0.005 0.004 Number of banks 11,836 11,836 11,836 6,358 6,358 6,358
24
Table 4 Abnormal loan growth dynamic system GMM estimation This table depicts the results of the dynamic system GMM regarding the abnormal loan growth. The full sample contains balance sheet info of 11,851 banks for the period 1995-2014. Loan growth describes the abnormal loan of bank i, which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Lagged loan growth is loan growth at t-1. Short-term rate is the 3-month LIBOR (USD). The long-Short-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Lagged profitability represents the pretax operating income divided by total assets at t-1. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3)
Variables Loan growth Loan growth Loan growth
Lagged loan growth 0.141*** 0.142*** 0.142***
(0.0106) (0.0109) (0.0109) Short-term rate -0.375*** (0.0654) Long-term rate -0.283** (0.119) Slope 0.576*** (0.102) Size -1.775*** -1.894*** -1.863*** (0.308) (0.329) (0.313) Lagged profitability 0.101 0.162 0.223 (0.254) (0.261) (0.267) Capitalisation 3.718*** 3.337*** 3.553*** (0.256) (0.232) (0.228) Efficiency -0.0670*** -0.0694*** -0.0690*** (0.00866) (0.00896) (0.00883) Economic growth -0.574*** -0.421*** -0.541*** (0.0617) (0.0474) (0.0567) Inflation 1.951*** 1.296*** 1.905*** (0.201) (0.139) (0.180) Constant -23.63*** -14.53** -18.89*** (6.352) (6.120) (5.698) Observations 121,335 121,335 121,335 Number of banks 11,851 11,851 11,851 Wald score 4520.22 4481.32 4490.93 AR (1) 0.000 0.000 0.000 AR (2) 0.138 0.111 0.124 J-test 0.000 0.000 0.000 Sargan 0.000 0.000 0.000
25
4.3. The impact of leverage
Dell’Ariccia, Laeven, and Marquez (2014) argue that the effect of leverage on risk-taking depends on the degree of leverage. A major hurdle is that interest rates directly affect risk-taking and through the level of capitalisation. The authors suggest the effect of leverage on risk-taking depends on the ability of changing said leverage. Dell’Ariccia, Laeven, and Marquez (2014) argue the easing of monetary policy
increases risk-taking and leverage. However, if leverage is assumed to be fixed, well capitalized banks increase risk-taking while low capitalized banks decrease it. Low capitalized banks are vulnerable to changes in assets prices and high costs are involved in defaulting. In addition, the effect depends on the degree of competition. To provide some clarity I try to assess the influence of leverage on risk-taking for different levels of capitalisation.
As is shown in table 5 different levels of capitalisation exhibit quite uniform reactions regarding taking. An increase in capitalisation leads to decreased risk-taking for different levels capitalisation to a certain extent, higher levels of
26
Table 5 Abnormal loan growth dynamic for different levels of capitalisation
This table depicts the results of the fixed effects regression regarding the abnormal loan growth and different levels of capitalisation. Loan growth describes the abnormal loan of bank i, which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Range depicts the scale of the capitalisation variable, which is equity over total assets. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3) (4) (5) (6)
VARIABLES Loan growth Loan growth Loan growth Loan growth Loan growth Loan growth
Range (%) 0-5 5-8 8-11 11-14 14-17 <17
Capitalisation -15.45* -0.887** -0.641*** -0.999*** -0.513 -0.294
(8.060) (0.372) (0.136) (0.254) (0.364) (0.205)
Short-term interest rate 0.729 -0.301** -0.185** -0.0270 -0.163 0.0705
(0.863) (0.146) (0.0795) (0.160) (0.259) (0.342) Size 1.388 -0.596 3.179*** 2.778** -1.205 1.253 (3.215) (0.884) (0.460) (1.357) (2.519) (2.755) Efficiency -0.0144 0.191*** 0.235*** 0.292*** 0.139*** 0.163*** (0.0528) (0.0273) (0.0185) (0.0344) (0.0490) (0.0517) Profitability 1.562* 5.114*** 5.110*** 5.367*** 2.677** 2.262** (0.837) (0.606) (0.330) (0.647) (1.242) (0.916) GDP 2.484*** 0.120 -0.208*** -0.260** -0.564** -0.134 (0.767) (0.152) (0.0771) (0.132) (0.239) (0.340) Inflation -2.508 -0.0993 0.0286 0.212 0.158 0.637 (2.944) (0.447) (0.236) (0.406) (0.699) (1.090) Constant 50.85 -4.885 -54.36*** -46.63*** 10.37 -23.52 (59.01) (11.89) (6.094) (17.22) (31.09) (34.10) Observations 784 14,786 46,179 16,918 5,738 5,611 R-squared 0.029 0.010 0.014 0.015 0.006 0.005 Number of banks 344 4,199 8,352 4,425 1,685 1,167
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
4.4. Validity of results
By performing multiple robustness analysis, I pursue validity of my results. My main robustness analysis is the Z-score to potentially capture inconsistencies in the measure. The results of the regressions regarding the Z-score are found in appendix A.2. Interestingly, strikingly different results are found among measures. The short and long-term interest rate are significant and have a positive coefficient, while the slope emerges to be negative. Potentially, the Z-score is a broader measure since this captures the risk of insolvency which is affected by a wide array of risks and not exclusively credit risk. For instance, Köhler (2012) shows the Z-score is
affected by abnormal loan growth, which might indicates the possible distortion of the Z-score in a risk-taking setting. The inclusion of abnormal loan growth is significant and yields to a change in sign for the long-term interest rate as is shown in appendix A.11. Furthermore the results of the GMM suggests high persistence of the
Z-27
score is rigid and is unresponsive to changes in the short-term interest rate and this could potentially devaluate the power of the results for the short-term interest rate.
The focus of this dissertation is the change in perception and the lowering of lending standards. Köhler (2012) argues that the abnormal loan growth is a good approximation of the perception of risk. Therefore, the results for abnormal loan growth outweigh the results of the Z-score. However, the controversy suggest caution while interpreting the results.
5. Conclusion
In this paper, I investigate the presence of a risk-taking channel for a
representative sample of US banks for the period of 1995-2014. The main focus is the impact of the short-term (hypothesis 1), long-term interest rate (hypothesis 2), and the slope of the yield curve (hypothesis 3) on the individual perception of risk. In addition, I provide a quick glance at the mediating role of leverage on risk-taking (hypothesis 4).
First, as for hypothesis 1, the short-term interest rate, which is most
28
and Marquez (2014) is given empirical support by regressing different levels of leverage on risk-taking. The asymmetrical response suggests that the amount of leverage matters for the influence on risk-taking.
This dissertation is limited in several ways. In first instance, it is unable to directly link risk-taking to monetary policy due inaccessibility of national credit registers and is therefore slightly distorted. Secondly, among the two employed measures some discrepancies exist which need to be taken into account by
interpreting the results. Abnormal loan growth embodies credit risk while the Z-score symbolizes solvency risk and in hindsight might not be a good measure. Thirdly, the generality of the sample has its drawbacks since the representativeness of specific bank specializations might distort the results of this dissertation. Fourthly, a proper assessment of the mediation between leverage and risk-taking transpired to be unachievable, since the influence of interest-rate on leverage is not researched in this dissertation and therefore the power of inferences regarding leverage is limited. Fifthly, the exclusion of potential regulatory time-varying state effects could distort the results of this dissertation.
The main concern of regulators is the balance between employment and inflation. This paper suggests that regulators should incorporate the destabilizing influence of interest rate in setting monetary policy, while accounting for a bank’s leverage, other bank characteristics, and market concentration, since these
characteristics influence the extent of risk-taking. In addition, daily supervisors should monitor banks which satisfy the conditions for high exposure to risk-taking. Nevertheless, a linear relationship between risk-taking and the interest rate environment is too good to be true. Dell’Ariccia, Laeven, and Marquez (2014) argue that the risk-taking is procyclical and therefore the effect varies in different periods and this also seen in appendix A.9. The aftermath of the financial crises does not exhibit additional risk-taking for a low interest rate environment. Berger and Udell (2004) confirm the institutional memory and procyclical lending behavior
29
implement capital buffers in times of excessive credit growth, since the last financial crisis is characterized by a significant increase in credit.
New research could provide additional insights in the risk-taking channel of monetary policy. The influence of leverage is relatively under researched and this would complete the nexus of risk-taking and monetary policy. Moreover, the
30 6. References
Adrian, Tobias and Hyun Song Shin. 2009. Money, Liquidity, and Monetary Policy. The American Economic Review 99, 600.
Alessandri, Piergiorgio and Benjamin D. Nelson. 2015. Simple Banking: Profitability and the Yield Curve. Journal of Money, Credit and Banking 47, 143-175.
Altunbas, Yener; Leonardo Gambacorta and David Marques-Ibàñez. 2014. Does Monetary Policy Affect Bank Risk? International Journal of Central Banking.
Arellano, Manuel and Stephen Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. The review of economic studies 58, 277-297.
Berger, Allen N. and Gregory F. Udell. 2004. The Institutional Memory Hypothesis and the Procyclicality of Bank Lending Behavior. Journal of financial intermediation 13, 458-495.
Bernanke, Ben S. and Alan S. Blinder. 1992. The Federal Funds Rate and the Channels of Monetary Transmission. The American Economic Review 901-921.
Bernanke, Ben S. and Alan S. Blinder. 1988. Credit, Money, and Aggregate Demand. American Economic Review 78, 435-439.
Bernanke, Ben S. and Mark Gertler. 1995. Inside the Black Box: The Credit Channel of Monetary Policy. The Journal of Economic Perspectives 9, 27-48.
Blundell, Richard and Stephen Bond. 1998. Initial Conditions and Moment
Restrictions in Dynamic Panel Data Models. Journal of Econometrics 87, 115-143.
31
Borio, Claudio and Haibin Zhu. 2012. Capital Regulation, Risk-Taking and Monetary Policy: A Missing Link in the Transmission Mechanism? Journal of Financial
Stability 8, 236-251.
Boyd, John H. and Gianni De Nicoló. 2005. The Theory of Bank Risk Taking and Competition Revisited. The Journal of Finance 60, 1329-1343.
Brunnermeier, Markus K. 2009. Deciphering the Liquidity and Credit Crunch 2007-2008. Journal of Economic Perspectives 23, 77-100.
Bruno, Valentina and Hyun Song Shin. 2015. Capital Flows and the Risk-Taking Channel of Monetary Policy. Journal of Monetary Economics 71, 119-132.
Čihák, Martin; Asli Demirgüç-Kunt; Erik Feyen and Ross Levine. 2012.
Benchmarking Financial Systems Around the World. World Bank Policy Research Working Paper.
Danıelsson, Jón; Hyun Song Shin and Jean-Pierre Zigrand. 2004. The Impact of Risk Regulation on Price Dynamics. Journal of Banking & Finance 28, 1069-1087.
Delis, Manthos D. and Georgios P. Kouretas. 2011. Interest Rates and Bank Risk-Taking. Journal of Banking & Finance 35, 840-855.
De Graeve, Ferre. 2008. The External Finance Premium and the Macroeconomy: US Post-WWII Evidence. Journal of Economic Dynamics and Control 32, 3415-3440.
Demirgüç-Kunt, Aslı; Enrica Detragiache and Thierry Tressel. 2008. Banking on the Principles: Compliance with Basel Core Principles and Bank Soundness. Journal of Financial Intermediation 17, 511-542.
De Nicolo, Gianni. 2001. Size, Charter Value and Risk in Banking: An International Perspective.
De Nicolò, Gianni; Giovanni Dell'Ariccia; Luc Laeven and Fabian Valencia. 2010. Monetary Policy and Bank Risk Taking. Available at SSRN 1654582 .
32
Dell'Ariccia, Giovanni and Robert Marquez. 2006. Lending Booms and Lending Standards. The Journal of Finance 61, 2511-2546.
Dell'Ariccia, Giovanni; Luc Laeven and Gustavo Suarez. 2014. Bank Leverage and Monetary Policy's Risk-Taking Channel: Evidence from the United States.
Diamond, Douglas W. and Raghuram G. Rajan. 2009. The Credit Crisis: Conjectures about Causes and Remedies. The American Economic Review 99, 606-610.
Fama, Eugene F. 1980. Banking in the Theory of Finance. Journal of Monetary Economics 6, 39-57.
Foos, Daniel; Lars Norden and Martin Weber. 2010. Loan Growth and Riskiness of Banks. Journal of Banking & Finance 34, 2929-2940.
Friedman, Milton and Anna Schwartz. 1963. A Monetary History of the United States: 1867-1960.
Gambacorta, Leonardo. 2009. Monetary Policy and the Risk-Taking Channel. BIS Quarterly Review December.
Gambacorta, Leonardo and Paolo Emilio Mistrulli. 2014. Bank Heterogeneity and Interest Rate Setting: What Lessons have we Learned since Lehman Brothers? Journal of Money, Credit and Banking 46, 753-778.
Gujarati, Damoder N. 2009. Basic Econometrics.
Hess, Kurt; Arthur Grimes and Mark Holmes. 2009. Credit Losses in Australasian Banking. Economic Record 85, 331-343.
Ioannidou, Vasso; Steven Ongena and José-Luis Peydró. 2015. Monetary Policy, Risk-Taking, and Pricing: Evidence from a Quasi-Natural Experiment. Review of Finance 19, 95-144.
33
Kashyap, Anil K. and Jeremy C. Stein. 2000. What do a Million Observations on Banks Say about the Transmission of Monetary Policy? American Economic Review 407-428.
Kiyotaki, Nobuhiro and John Moore. 1997. Credit Cycles. The Journal of Political Economy 105, 211-248.
Konishi, Masaru and Yukihiro Yasuda. 2004. Factors Affecting Bank Risk Taking: Evidence from Japan. Journal of Banking & Finance 28, 215-232.
Köhler, Matthias. 2012. Which Banks are More Risky? The Impact of Loan Growth and Business Model on Bank Risk-Taking.
Laeven, Luc and Ross Levine. 2009. Bank Governance, Regulation and Risk Taking. Journal of Financial Economics 93, 259-275.
Maddaloni, Angela and José-Luis Peydró. 2011. Bank Risk-Taking, Securitization, Supervision, and Low Interest Rates: Evidence from the Euro-Area and the US Lending Standards. Review of Financial Studies 24, 2121-2165.
Mink, Mark. 2011. Procyclical Bank Risk-Taking and the Lender of Last Resort. SSRN Working Paper Series.
Modigliani, Franco and Merton H. Miller. 1958. The Cost of Capital, Corporation Finance and the Theory of Investment. The American Economic Review 48, 261-297.
Nickell, Stephen. 1981. Biases in Dynamic Models with Fixed Effects. Econometrica: Journal of the Econometric Society 1417-1426.
Okina, Kunio and Shigenori Shiratsuka. 2004. Policy Commitment and Expectation Formation: Japan’s Experience Under Zero Interest Rates. The North American Journal of Economics and Finance 15, 75-100.
34
Sinkey Jr, Joseph F. and Mary Brady Greenawalt. 1991. Loan-Loss Experience and Risk-Taking Behavior at Large Commercial Banks. Journal of Financial Services Research 5, 43-59.
35
Appendix A.1. Overview of variables mentioned in this dissertation Table 6 Used variables in this dissertation including, definition, source, and assumption
Variable Definition Source Assumption
Dependent
Abnormal loan growth Annual loan growthi,t – Median loan growthc,t Bankscope Endogenous Z-score Return on Assets,it− Capital Adequacy Ratioi,t
Standard Deviation of Return on Assetsi
Bankscope Endogenous
Independent
Short-term rate Annual average of the 3-month LIBOR in USD / Federal funds rate*
Federal Reserve of St. Louis
Exogenous
Long-term Annual average of the 10 year US government bond Federal
Reserve of St. Louis
Exogenous
Slope Annual average of the 10 year US government bond
minus Annual average of the 3-month LIBOR in USD Federal Reserve of St. Louis Exogenous Bank control
Size Natural logarithm of total assets Bankscope Predetermined
Profitability Pretax operating income divided by total assets Bankscope Endogenous
Capitalisation Equity over assets Bankscope Endogenous
Efficiency Cost-to-income ratio Bankscope Endogenous
Macroeconomic control
36
Appendix A.2. Results regarding the Z-score as main dependent variable Table 7 Fixed effects regression with the z-score as dependent variable
This table depicts the results of the fixed effects regression regarding the Z-score. The full sample contains balance sheet info of 10,808 banks for the period 1995-2014 while the active sample consists of 6,298 banks. The Z-score represents the solvency risk associated for a bank, the solvency is the capital to asset ratio of bank i minus the return on assets for the full sample divided by the standard deviation of the return on assets for the whole sample. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3) (4) (5) (6)
Variables Z-score Z-score Z-score Z-score Z-score Z-score
Short-term rate 0.200*** 0.164*** (0.0182) (0.0192) Long-term rate 0.170*** 0.00209 (0.0483) (0.0443) Slope -0.267*** -0.295*** (0.0206) (0.0251) Size -0.206 -0.194 -0.499*** -0.520*** -0.700*** -0.821*** (0.133) (0.170) (0.125) (0.120) (0.152) (0.115) Profitability 0.0352 0.0382 0.0323 0.0180 0.0194 0.0153 (0.0315) (0.0328) (0.0306) (0.0249) (0.0255) (0.0244) Capitalisation -0.435*** -0.434*** -0.447*** -0.353*** -0.361*** -0.365*** (0.0302) (0.0298) (0.0303) (0.0363) (0.0363) (0.0364) Efficiency 0.00887*** 0.00940*** 0.00646** 0.0202*** 0.0193*** 0.0175*** (0.00316) (0.00325) (0.00314) (0.00391) (0.00397) (0.00389) Economic growth -0.00388 -0.0608*** -0.0145 -0.0266 -0.0962*** -0.0157 (0.0198) (0.0186) (0.0196) (0.0232) (0.0212) (0.0238) Inflation 0.0458 0.360*** 0.0913 0.0724 0.439*** -0.00246 (0.0571) (0.0469) (0.0601) (0.0683) (0.0523) (0.0750) Constant 5.952*** 5.009** 10.61*** 8.274*** 10.31*** 13.18*** (1.564) (2.178) (1.483) (1.474) (1.998) (1.444) Observations 114,981 114,981 114,981 81,917 81,917 81,917 R-squared 0.014 0.013 0.014 0.009 0.009 0.009 Number of banks 10,808 10,808 10,808 6,298 6,298 6,298
37
Table 8 Dynamic GMM estimation with the Z-score as dependent variable
This table depicts the GMM estimation results concerning the Z-score. The full sample contains balance sheet info of 10,823 banks for the period 1995-2014 while the active sample consists of 6,305 banks. The Z-score represents the solvency risk associated for a bank, the solvency is the capital to asset ratio of bank i minus the return on assets for the full sample divided by the standard deviation of the return on assets for the whole sample. The lagged Z-score is equal to Z-score t-1. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Lagged profitability represents the pretax operating income divided by total assets at t-1. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3) (4) (5) (6)
VARIABLES Z-score Z-score Z-score Z-score Z-score Z-score
Lagged Z-score 0.956*** 0.952*** 0.957*** 0.948*** 0.946*** 0.946*** (0.0108) (0.0107) (0.0108) (0.00869) (0.00869) (0.00863) Short-term rate 0.190*** 0.232*** (0.0198) (0.0200) Long-term rate 0.262*** 0.266*** (0.0331) (0.0361) Slope -0.180*** -0.233*** (0.0260) (0.0261) Size -0.315*** -0.308*** -0.330*** -0.249*** -0.247*** -0.262*** (0.0516) (0.0519) (0.0512) (0.0562) (0.0567) (0.0566) Lagged profitability -0.00943 -0.00598 -0.00572 0.00860 0.0124 0.00984 (0.0414) (0.0388) (0.0383) (0.0247) (0.0222) (0.0235) Capitalisation 0.203*** 0.312*** 0.199*** 0.0222 0.105* -0.00726 (0.0602) (0.0553) (0.0644) (0.0615) (0.0610) (0.0632) ROAA 0.0138*** 0.0170*** 0.0101** 0.0427*** 0.0466*** 0.0363*** (0.00468) (0.00481) (0.00457) (0.00718) (0.00751) (0.00695) Economic growth -0.0176 -0.0629*** -0.0387* 0.0135 -0.0455** -0.0230 (0.0206) (0.0186) (0.0218) (0.0249) (0.0228) (0.0256) Inflation 0.170*** 0.405*** 0.329*** 0.0286 0.354*** 0.231*** (0.0640) (0.0461) (0.0718) (0.0678) (0.0500) (0.0773) Constant -2.133** -4.853*** -1.226 -2.783*** -5.189*** -1.389 (0.931) (0.988) (1.075) (0.947) (1.071) (1.037) Observations 115,087 115,087 115,087 81,991 81,991 81,991 Number of banks 10,823 10,823 10,823 6,305 6,305 6,305 Wald score 14977.63 14304.12 14832.21 19775.80 19609.58 19690.32 AR (1) 0.002 0.002 0.002 0.001 0.001 0.001 AR (2) 0.188 0.140 0.163 0.163 0.147 0.151 J-test 0.000 0.000 0.000 0.000 0.000 0.000
38
Appendix A.3. Cumulative distribution of observations per year over sample period Graph 2 Cumulative distribution of observations over sample period
39
Appendix A.4. Fixed effects regression per specialisation regarding abnormal loan growth
Table 9 Abnormal loan growth fixed effect regression regarding bank holding companies
This table depicts the results of the fixed effects regression regarding abnormal loan growth. The full sample contains balance sheet info of 2,408 banks for the period 1995-2014 Loan growth describes the abnormal loan growth of bank i which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3)
Variables Loan growth Loan Growth Loan Growth
Short-term rate -0.193* (0.111) Long-term rate -0.353 (0.259) Slope 0.208 (0.143) Size 0.888 0.656 1.200** (0.658) (0.878) (0.573) Profitability 0.993** 0.974** 1.009** (0.470) (0.470) (0.466) Capitalisation 0.0560 0.0514 0.0678 (0.129) (0.130) (0.128) Efficiency 0.0748*** 0.0722*** 0.0777*** (0.0267) (0.0267) (0.0266) Economic growth 0.711*** 0.724*** 0.724*** (0.128) (0.127) (0.123) Inflation -0.553* -0.680** -0.638** (0.295) (0.278) (0.275) Constant -19.75** -15.18 -24.96*** (9.120) (12.89) (7.996) Observations 16,049 16,049 16,049 R-squared 0.007 0.007 0.007 Number of banks 2,408 2,408 2,408
40
Table 10 Abnormal loan growth fixed effect regression regarding commercial banks
This table depicts the results of the fixed effects regression regarding abnormal loan growth. The full sample contains balance sheet info of 8,441 banks for the period 1995-2014 Loan growth describes the abnormal loan growth of bank i which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3)
Variables Loan growth Loan Growth Loan Growth
Short-term rate -0.0261 (0.0639) Long-term rate -0.422*** (0.109) Slope -0.205** (0.0817) Size 2.304*** 1.754*** 2.231*** (0.290) (0.353) (0.288) Profitability 0.594 0.592 0.591 (0.406) (0.405) (0.404) Capitalisation 0.153** 0.133** 0.151** (0.0614) (0.0623) (0.0610) Efficiency 0.0918*** 0.0874*** 0.0911*** (0.0217) (0.0218) (0.0217) Economic growth 0.278*** 0.222*** 0.356*** (0.0707) (0.0614) (0.0674) Inflation 0.0888 0.342** -0.293* (0.193) (0.144) (0.173) Constant -38.55*** -30.33*** -36.67*** (4.339) (5.326) (4.492) Observations 93,884 93,884 93,884 R-squared 0.005 0.005 0.005 Number of banks 8,441 8,441 8,441
41
Table 11 Abnormal loan growth fixed effect regression regarding saving banks
This table depicts the results of the fixed effects regression regarding abnormal loan growth. The full sample contains balance sheet info of 699 banks for the period 1995-2014 Loan growth describes the abnormal loan growth of bank i which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3)
Variables Loan growth Loan Growth Loan Growth
Short-term rate -0.235* (0.122) Long-term rate 0.0191 (0.209) Slope 0.500*** (0.186) Size -0.129 0.205 0.0951 (0.718) (0.775) (0.670) Profitability 0.768 0.760 0.758 (0.551) (0.552) (0.551) Capitalisation -0.105 -0.0893 -0.0834 (0.0931) (0.0958) (0.0923) Efficiency 0.0906*** 0.0903*** 0.0945*** (0.0268) (0.0269) (0.0269) Economic growth 0.120 0.168 0.0976 (0.149) (0.146) (0.148) Inflation 0.795** 0.356 0.955*** (0.373) (0.326) (0.358) Constant -6.737 -10.85 -11.64 (9.046) (10.31) (8.419) Observations 10,743 10,743 10,743 R-squared 0.004 0.004 0.004 Number of banks 920 920 920
42
Appendix A.5. Fixed effects regression regarding defaulted banks during the sample period
Table 12 Abnormal loan growth fixed effects regression regarding inactive banks
This table depicts the results of the fixed effects regression regarding abnormal loan growth. The full sample contains balance sheet info of 5,478 banks for the period 1995-2014 Loan growth describes the abnormal loan growth of bank i which is calculated by the loan growth of bank i at time t minus the median loan growth of time t. Short-term rate is the 3-month LIBOR (USD). The long-term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2) (3)
Variables Loan growth Loan Growth Loan Growth
Short-term rate -0.0633 (0.0958) Long-term rate -0.276 (0.210) Slope 0.00309 (0.125) Size 2.554*** 2.339*** 2.614*** (0.524) (0.597) (0.502) Profitability 1.693*** 1.690*** 1.693*** (0.359) (0.361) (0.358) Capitalisation -0.0718 -0.0798 -0.0703 (0.0875) (0.0886) (0.0874) Efficiency 0.131*** 0.128*** 0.131*** (0.0200) (0.0200) (0.0200) Economic growth 0.800*** 0.798*** 0.822*** (0.121) (0.115) (0.118) Inflation -0.454 -0.432* -0.571** (0.286) (0.238) (0.282) Constant -45.32*** -41.48*** -46.07*** (6.775) (8.148) (6.570) Observations 37,962 37,962 37,962 R-squared 0.009 0.009 0.009 Number of banks 5,478 5,478 5,478
43
Appendix A.6. Fixed effects regression regarding federal funds rate as short-term interest rate instrument
Table 13 Abnormal loan growth fixed effects regression regarding Federal Funds rate
This table depicts the results of the robustness test for the short-term interest rate. The full sample contains balance sheet info of maximum 11,836 banks for the period 1995-2014 while the active sample consists of 6,358 banks. Loan growth describes the abnormal loan growth of bank i which is calculated by the loan growth of bank i at time t (%) minus the median loan growth of time t. Short-term rate is the Federal funds rate The term rate is the 10 year US government bond. The slope illustrates the long-term interest rate minus the short-long-term. Size is the natural logarithm of total assets. Profitability represents the pretax operating income divided by total assets. Capitalisation is equity over assets. Efficiency is represented by the cost-to-income ratio. GDP is the annual growth of the real gross domestic product. Lastly, Inflation depicts the annual rise in consumer prices as measured by the CPI.
(1) (2)
Variables Loan growth Loan growth
Short-term rate -0.0549 -0.0300 (0.0488) (0.0555) Size 1.805*** 1.893*** (0.267) (0.279) Profitability 0.594* 0.391 (0.352) (0.254) Capitalisation 0.119** 0.195*** (0.0507) (0.0543) Efficiency 0.0851*** 0.0898*** (0.0186) (0.0145) Economic growth 0.344*** 0.146*** (0.0551) (0.0552) Inflation 0.0142 0.134 (0.139) (0.157) Constant -32.01*** -33.13*** (4.091) (3.791) Observations 121,169 83,207 R-squared 0.004 0.004 Number of banks 11,836 6,358
44
Appendix A.7. Developments of the slope and interest rates over the sample period
Graph 3 The development of the interest rates and slope over time
45
Appendix A.8. First differencing the dependent variable for the fixed effects regression
Graph 4 Residual plot of first difference method.
This graph depicts the residuals of the standard fixed effect regression and the fixed effect regression with the first difference of abnormal loan growth. The y-axis denotes the residuals while the x-axis shows the year of the
