UNIVERSITEIT VAN AMSTERDAM
MSc. in Economics
Specialization Monetary Policy and Banking Master Thesis
“Bank Risk taking and credit risk at the zero lower bound: evidence for the United States”
Supervisor: Dr. Christian A. Stoltenberg Second Reader: Dr. Ward Romp
14th October, 2013
ii
“Bank Risk taking and credit risk at the zero lower bound: evidence for the United States”
Bianca Andreini October, 2013
Abstract
This study examines the influence of monetary policy on bank risk taking from 2009 until 2012 in the United States. I estimate a panel regression using a broad sample of banks, measuring risk taking by a credit risk indicator and calculating a specific Taylor rule for the country. Thereby, bank risk taking results higher for expansionary monetary-policy surprises, i.e., for interest rates that are lower than the ones expected from a Taylor rule even if interest rates are reaching the zero lower bound. Persistent deviations of interest rates from a benchmark influences higher risk-taking; while short-term deviations from the current level of the monetary policy stance affect the riskiness of outstanding loans.
JEL Classification: E43, E44, E52, G21.
iii Table of contents
1. Introduction ... - 4 -
2. The risk-taking channel of monetary policy ... - 5 -
2.1. Monetary transmission and the risk-taking channel: ... - 5 -
2.2. Functioning of the risk-taking channel ... - 7 -
2.3. Empirical evidence ... - 10 -
3. Data description ... - 14 -
4. Empirical Strategy ... - 15 -
4.1. Interest rate and the risk-taking channel ... - 16 -
4.2. Risk-taking and the zero lower bound: ... - 18 -
4.3. Interaction with banks’ characteristics ... - 19 -
4.4. Monetary response during the crisis: unconventional monetary policy ... - 21 -
5. Empirical Results: ... - 23 -
5.1. Interest rate and the risk-taking channel ... - 23 -
5.2. Risk-taking and the zero lower bound: ... - 24 -
5.3. Interaction with banks’ characteristics ... - 26 -
5.4. Monetary response during the crisis: unconventional monetary policy ... - 29 -
6. Conclusions ... - 30 -
Appendix ... - 32 -
- 4 - 1. Introduction
The recent financial crisis put in evidence several weaknesses in the financial system. The fragility of the financial sector was mainly related to prudential regulation of a more complex and integrated financial sector, and also to the extent to what monetary policy contributes to the build-up of risk. Monetary policy interest rates were unusually low compared to historical levels (Taylor, 2009a) and were maintained at that level for a long period. Accordingly, accommodative monetary policy seemed to have contributed to higher risk-taking in the financial sector.
Whether monetary policy still influences higher risk-taking by banks when interest rates are near the zero lower bound is a matter of investigation. In this study, using a broad sample of banks I estimate an econometric panel data for the United States from 2009 until 2012 to measure the impact of monetary policy on bank risk-taking. Hence, I study the effect on credit risk of interest rates that are lower than the ones expected from a Taylor rule; and, the functioning of the risk-taking channel of monetary policy when the monetary policy interest rate is close to the zero lower bound.
The build-up of systemic risk and mispricing of risk that recently led to a global financial crisis forced a change on prudential regulation towards a macroprudential approach. This framework aims at limiting risk episodes of financial systemic distress that have a negative impact on the whole economy (Borio, 2003). In this view, monetary policy is not exempt of contributing to the build-up of risk. Changes in monetary policy alter risk perceptions or risk-tolerance, driving risk taking incentives (Borio and Zhu, 2008). Additionally, long periods of accommodative monetary policy can turn into instability when interest rates are raised again. Then, in this study, I will argue that monetary policy do play a role on the build-up of risk even when interest rates are reaching the zero lower bound.
The outline of the rest of this study is as following. The first section examines the academic literature on the risk-taking channel of monetary policy considering the recent growing empirical evidence. In this part, I review the concept of the risk taking channel, its
- 5 -
functioning, and I discuss some empirical findings on the academic literature. In the following section I present a brief description of the data in order to in the third section describe the empirical strategy. Consequently, in the fourth section, I analyze the main econometric results. Both sections, the empirical strategy and the econometric results, present the models and findings allowing for the effect of the stance of monetary policy on banks’ risk-taking, including additional measures to account for the zero lower bound, the effect of banks’ characteristics, and unconventional monetary policy. Finally, I summarize the main conclusions.
2. The risk-taking channel of monetary policy
Monetary policy transmits its impulses to the economy through different channels. In this section, I describe the traditional interest rate channel as well as the balance sheet and credit channel; and, set my focus on the risk-taking channel. Consequently, I explain how low interest rates affect risk-taking in order to finally summarize the empirical literature relevant for this study.
2.1. Monetary transmission and the risk-taking channel:
Economists have tried to explain the causes of the 2008 global financial crisis. Among others1, the looseness of monetary policy seemed to have contributed to higher risk taking in the banking sector. A growing literature developed under this theory, attaching more complexity to the transmission mechanism of monetary policy. As a result, the so called risk-taking channel theory emerged.
In spite of the new prominent research on this field, the risk taking channel has always been present. Risk is part of the nature of financial intermediaries. Banks borrow short and lend long incurring on interest rate risks. At the same time, credit risks and liquidity risks are also part of banks business. In order to dissipate these risks, banks monitor and
1 Other causes of the financial crisis: regulatory and supervisory frameworks (Blundell-Wignall, Atkinson and
Hoon Lee, 2008), development of complex credit market instruments (securitization Maddaloni and Peydró, 2010), savings glut hypothesis (Bernanke, 2005).
- 6 -
screen borrowers, and manage liquidity risks investing on liquid assets and diminishing withdrawal risks. However, despite these efforts, banks may accept higher levels of risks when monetary policy is expansionary, and recent financial developments and reforms may have intensified this effect.
Financial liberalization, technological change and the introduction of new financial products and strategies such as securitization, have added complexity to the transmission of monetary policy. The financial sector moved away from the traditional “originate-to-hold” to an “originate-to-distribute” model. In this framework, banks originate, repackage, and then sell their exposures in the financial markets. Accordingly, policy interest rates are interacting with more factors adding complexity to the transmission mechanism of monetary policy. As Gambacorta and Marques Ibañez (2011, p. 12) point out, ‘the same instruments that are used to hedge risks also have the potential to undermine financial stability [sic] by facilitating the leveraging of risk’.
Monetary policy transmission is now more complex. The macro literature about the transmission of monetary policy has mainly focused on the effect of the level of interest rates on investment and consumption. In this view, a change in the nominal interest rate leads under sticky prices to a change on real interest rates, which affects households and firms’ consumption and investment decisions. The role of financial intermediaries is ignored2. Nevertheless, financial and credit conditions are also important in the propagation of the business cycle. Under imperfect information a change on interest rates is amplified by the effect it has on borrowers and banks’ balance sheets, and the supply of credit (Bernanke and Gertler, 1995). The first effect is usually referred as financial accelerator or balance sheet channel of monetary policy, which stresses the potential impact of a change on interest rates on borrowers’ balance sheets or income statements and thus on their creditworthiness. Lower interest rates increase borrowers’ balance sheets via increasing borrowers’ financial position or net cash flows; at the same time, falling interest rates increase asset prices which in turn strengthen collateral values.
2
- 7 -
Thereby, financial health of borrowers facilitates credit extension and reduces lender’s risks.
The latter effect makes reference to the bank lending channel, or also called narrow credit channel. In this view, monetary policy tightening alters the cost of holding deposits which leads banks to react reducing its lending portfolio. The concept of external finance premium defined as the difference between the cost of funds raised externally (by issuing equity or debt) and the cost of funds generated internally (by retaining earnings) (Bernanke and Gertler, 1995) links both channels. In the financial accelerator theory, a change on interest rates modifies the external finance premium of borrowers firms and households, while in the bank lending view the effect is on banks’ balance sheets.
The bank lending channel and balance sheet channel effects are impossible to disentangle (Bernanke, 2007) being their focus on the quantitative impact of monetary policy on lending and its propagation to the business cycle.
This risk-taking channel on the contrary reflects the changes in the propensity of the banking sector to take on more risk due to low policy interest rates. In contrast to the other channels, its focus is on the quality of lending and the riskiness of banks activities. It can be defined as ‘the impact of changes in policy rates on either risk perceptions or risk-tolerance and hence on the degree of risk in the portfolios, on the pricing of assets, and on the price and non-price terms of the extension of funding’ (Borio and Zhu; 2008, p. 9).
2.2. Functioning of the risk-taking channel
The risk-taking channel encompasses different mechanisms. It operates boosting assets and collateral values, as well as through the relation between low interest rates and the rate of returns. In addition, some features of the banking sector such as limited liability, asymmetric information and the degree of leverage do also influence risk-taking.
Low interest rates have a direct effect on valuations, for the same cash flow or expected income, the net present value is higher when interest rates are low. Not only does this
- 8 -
boost in assets and collateral values impact through the financial accelerator as previously explained, but it also alters risk perceptions. An interest rate reduction affects the probability of default, loss given default, and regulatory measures such as expected loss and risk weighted assets, which in turn affect the propensity of banks to take on more risk. Therefore, the risk taking channel operates by amplifying the financial accelerator effect as a “persistence-enhancing” mechanism (Borio and Zhu, 2008)
This effect of interest rates on valuations and asset prices is related to the procyclicality of risk in regulatory frameworks. Risk measures tend to vary with business cycle fluctuations, and under the Basel II Accord capital measures are procyclical (Repullo, Saurina and Trucharte, 2009). When the economy is booming, more investment projects are profitable; thus, borrowers are less risky. At the same time low interest rates thrive collateral assets prices. Hence, risks are perceived as lower. This procyclicatlity of risk measures influences bank’s capital ratios. Borio and Zhu (2008) called this effect “capital framework effect” which makes reference to how interest rates operates influencing the way in which the bank perceives, manages and prices risks inducing a change on the risk level of the whole bank portfolio.
The second mechanism through which the risk-taking channel operates is the relation between low interest rates and rate of returns. When interest rates are low, compensation incentives might alter risk taking tolerance leading to higher risk taking. Financial institutions which have long term commitments need to obtain a fixed rate of return. Since investing in risk free assets when interest rates are low does not allow financial institutions to match the yield they promised on their liabilities with the one they obtain on their assets, the “search for yield” (Rajan, 2005) acts as a risk-taking incentive. This operates especially when monetary policy is accommodative for a long period of time.
Risk taking incentives arising due to contractual terms of financial institutions, can also go along with managerial incentives. Sometimes investment managers’ rewards are linked to the performance among their peers. This can be a perverse mechanism if as a result,
- 9 -
managers invest on strategies that involve high risks with low probability and high returns the rest of the time. Furthermore, if managers herd together in order not to underperform compared to the peers, a misalignment of assets prices respect to their fundamentals may arise. As pointed out by Rajan (2005) tail risk strategies and herding behavior are mutually reinforcing.
The search for yield effect is related to the asset substitution effect (De Nicolo et al., 2010). Asset substitution refers to portfolio reallocation as a result of a change on interest rates. Low yields on safe assets pushes risk neutral banks to increase their demand for risky assets. This effect could be intensified if economic agent´s risk perceptions or risk tolerance are affected by very low level of interest rates. In addition if the difference between market and target or contractual rates is unusually large, the search for yield can be stronger (Borio and Zhu, 2008).
The degree of financial leverage is also a variable that influences bank’s riskiness. Leverage exhibits a procyclical behavior, in the sense that it is increasing when balance sheets are increasing (Adrian and Shin, 2009a). As equity financing is more costly, when banks face shocks to their portfolios or profits, instead of distributing dividends or raising new capital they adjust their balance sheets. In this regard, a cut on interest rates alters the profitability of bank lending and the present value of banks income3; thereby, increasing forward-looking measures of capital. This feature enlarges banks’ capacity to lend. Banks responds to the fall in leverage by increasing their demand for assets. Accordingly, banks’ risk-bearing capacity increases boosting bank lending and risk-taking.
3 Adrian and Shin (2009b) make use of the concept of net interest margin (NIM) to explain how financial
intermediaries drive financial cycle adjusting leverage and risk-taking. The NIM is the difference between the total interest income on the asset side of the balance sheet and the interest expense on the liabilities side of the balance sheet. While a change on interest rates alters the term spread or the profitability of the marginal loan, it also boosts the present value of income; thus, increasing the incentives to expand the size of balance sheets.
- 10 -
Since banks are protected by limited liability, once leverage is optimally chosen, banks will unambiguously lower bank monitoring or increase risk-taking (Dell Ariccia, Laeven and Marquez, 2010).
Finally, financial regulation has also some implications for risk-taking incentives. The government safety net ensures that depositors and creditors will not suffer losses if a financial institution fails. This protection mechanism has the main drawback that lead to moral hazard introducing incentives to take risks than otherwise would have been taken (Mishkin, 2010). Thus, risk neutral leveraged banks operating under asymmetric information, limited liability and protection of deposit insurance, behave as risk seeking agents not internalizing the losses they impose on depositors and bondholders.
2.3. Empirical evidence
Turning to the empirical evidence on the risk-taking channel of monetary policy, the literature is quite recent. On the one hand some studies make use of confidential credit information of financial institutions to test the influence of interest rates on the riskiness of loans (Jiménez et al., 2008; Ioannidou, Ongena and Peydró, 2009; Dell’Ariccia, Laeven and Suarez, 2013) . On the other hand, some research use bank level data to test the influence of interest rates on different bank risk indicators (Altunbas, Marqués Ibañez and Gambacorta, 2010; Ozsuca and Akbostanci, 2012; Köhler, 2012).
Among the studies which make use of internal bank data, Jiménez et al. (2008) by means of monthly information of credit institutions in Spain find that lower short-term interest rates prior to loan origination result in riskier loans being granted. Furthermore, lower interest rates reduce the risk of outstanding loans. The results suggest that accommodative monetary policy lead to lower credit risk in the very short run due to an immediate interest rate effect on outstanding loans, but worsen the riskiness of the portfolio in the medium run because of credit standards and monitoring are lessened.
- 11 -
Similar approach is followed by Ioannidou, Ongena and Peydró (2009) who also estimate the effect of monetary policy on the time to individual loan defaults. Using credit information for Bolivia, the authors develop a quasi-natural experiment in which the US federal funds rates results the appropriate measure of monetary policy in an almost fully dollarized economy. Likewise, low interest rates lead to the origination of more loans with a higher hazard rate. The negative effect on the probability of default of outstanding loans is also found when interest rates are lower. Risk is also negatively priced, having subprime borrowers more access to loans while risk premiums are not augmented.
In the same line of research using micro level data, Dell’Ariccia, Laeven and Suarez (2013) investigate the risk-taking channel for the U.S. banking system by means of confidential data on internal ratings of banks on loans. In this case, they perform a regression of an ex-ante risk indicator measured by the risk rating of the loan portfolio on the short term rate (proxied by the fed funds rate) and an interaction term between interest rates and bank capital. Accordingly, they find risk-taking negatively associated with increases on interest rates; being this relationship more pronounced for highly capitalized banks.
Applying a different approach, Altunbas, Marqués Ibañez and Gambacorta (2010) measure bank risk-taking using a market based indicator. In a panel data regression the authors assess the effect of low and for a long time interest rates on the expected default frequency (EDF). This is a forward-looking indicator of credit risk which measures the expected probability of default of a bank within the period of one year. As mentioned in previous papers, if interest rates are lowered the overall quality of the portfolio increases, which is in line with the balance sheet channel of monetary policy. The empirical strategy includes the difference between the short term interest rate and a benchmark rate calculated by a “Taylor Rule”. Accordingly, they found that if interest rates are below the benchmark, banks do take more risks.
With a similar methodology, Ozsuca and Akbostanci (2012) study the risk-taking channel in the Turkish banking sector. In addition to the EDF, this paper makes use of accounting
- 12 -
based risk indicators. Estimating a dynamic panel regression, the authors find that low short term interest rates reduce the risk of outstanding loans. However, if short term interest rates are below a theoretical benchmark, risk-taking by banks increases.
The empirical literature provides strong evidence of the existence of a risk-taking channel of monetary policy. Monetary easing conditions for a long time make banks more prone to take higher risks and suffer a significant deterioration of solvency.
The interaction of monetary policy with the characteristics of the banking sector it is also an important factor to understand the functioning of the risk-taking channel. Concerning this, Altunbas et al. (2012) study monetary policy looseness and the interaction with bank characteristics modeling the probability of a bank to become risky during the crisis. Banks with different characteristics adopted different risk positions in the pre-crisis period when monetary policy was highly accommodative. As a result, liquid and well-capitalized banks suffered less during 2007-2009 financial crises. This finding is consistent with other empirical studies in the literature, in this sense well-capitalized banks are considered less risky by the market (Altunbas, Marques Ibañez and Gambacorta, 2010) and are less willing to take on more risks (Ioannidou, Ongena and Peydró, 2009). The exception is the paper by De Nicolò et al. (2010) who find that banks with lower level of capital take less risk since they cannot bear up as high risk as well-capitalized banks. The rationale behind the empirical finding that well-capitalized banks take less risk is in line with the “skin-in-the-game” effect. Under limited liability protection, banks have more incentives to take on more risks; while, the level of capital acts in the opposite direction. Therefore, well-capitalized banks will behave like institutions without limited liability protection taking lower risks when monetary policy is loose.
In the case of liquidity, Altunbas, Marques Ibañez and Gambacorta, (2010), Ozsuca and Akbostanci (2012) find that banks holding more liquid assets are perceived by the market as safer and have less non-performing loans. In this case, large liquid assets holdings are related to risk adverse banks which attempt to be more protected in case of unexpected
- 13 -
withdrawals or deposit outflows. On the contrary, Ioannidou, Ongena and Peydró (2009) find a positive relation between liquidity and bank risk. More liquid assets may allow banks to take more risks since they act as a cushion or a higher protection in case of a negative shock. Additionally, holding more liquid assets with lower yield involves a higher opportunity cost, which in turn may influence banks to take higher risks.
Liquidity and risk-taking can be tightly connected reinforcing monetary policy transmission (Borio and Zhu, 2008). The concept of liquidity includes funding cash liquidity and the ability to realize value from assets. If an increase of the monetary interest rate raises risk perceptions; then, risk tolerance diminishes and liquidity conditions in the market tighten. Monetary policy can influence liquidity conditions in the interbank market, but risk perceptions do also play a role. In the financial turmoil of 2007, liquidity in the interbank market dried up abruptly mainly due to a counterparty risk issue (Taylor, 2008). As a result, central banks responded easing monetary policy even including unconventional monetary measures. Nevertheless, all central banks´ massive injections of liquidity did little to restart interbank lending, and precautionary liquidity hoarding characterized the crisis and post-crisis period (Berrospide, 2012). Risk perceptions such as a higher perceived counterparty credit risk, affect market and liquidity funding weakening monetary policy transmission. Simultaneously, monetary policy may alter repricing or misperceptions of risk which have an effect on bank lending activities including also bank lending in the interbank market. In this sense, this study will attempt to analyze liquidity funding, a distinction that has not been done yet in the empirical literature.
The empirical evidence broadly supports the existence of the risk-taking channel as another channel through which monetary policy is transmitted to the economy. As emphasized by Borio and Zhu (2008) the risk-taking channel is not the most important channel of monetary policy but its exploration is important to understand the fully transmission of monetary policy impulses to the economy, especially in periods of changes in financial markets and prudential regulation. The risk taking channel in principle may act as a benign “persistence-enhancing” mechanism (Borio and Zhu; 2008) similar to the
- 14 -
financial accelerator. Its importance in terms of monetary policy arises when especial markets conditions make the build-up of risk and financial imbalances more likely, resulting in this way monetary easing in a kind of fuel to future financial distress. Macroeconomic conditions do also matter for the transmission of the monetary policy and for its interaction with risk perceptions, especially since risks tend to be underestimated in booms and overestimated in recessions (Borio, Furfine and Lowe; 2001). Finally, bank characteristics soften or intensify the effect of the monetary policy stance on bank lending and bank risk taking.
3. Data description
The sample was prepared considering the best sample in terms of quality and availability of data. Thus, the sample comprises quarterly observations for 7.982 banks of the United States in the period from 2009 until 2012. Table 1 provides a description of the dataset.
Listed and unlisted banks are included since, as noticed by Köhler (2012), unlisted banks are usually smaller and follow more traditional commercial banking activities. Therefore including unlisted banks provides a better sample when considering credit risk indicators.
Quarterly bank data are collected from Bankscope which provides information of banks' financial statements. Data was extracted on April and July 2013. The money market rate was obtained from Datastream (April,2013). In this case, the overnight interest rate or the fed funds rates is used. Finally, to estimate a Taylor rule for United States, nominal and real GDP, and inflation are obtained from the OECD Economic outlook (April, 2013). For the estimation of the Taylor rule, I use quarterly data and apply the Hodrick-Prescott filter with a smoothing parameter sets at 1600 (Graph 1). The estimation is done from 2002 Q2 to 2007 Q4, since later the interest rate is reaching the zero lower bound.
Following the existing literature about the risk-taking channel of monetary policy, I would prefer to employ a market-risk based indicator. However, due to data restrictions I use an accounting measure of risk. Precisely, in this study, risk-taking is measured using the ratio
- 15 -
of non-performing loans to total loans. This measure quantifies credit risk in the loan portfolio providing a signal of asset quality. High levels of this ratio mean high delinquencies or defaults in the loan portfolio which might reflect a general over-indebtedness of households and corporates fed by a higher propensity of banks to take on more risks. The ratio of impaired loans to total gross loans is on average equal to 3.27 for the whole period, but the sample includes banks with a large part of its portfolio on default (Graph 2 and 3).
To measure monetary policy stance I consider the federal funds rate. This period has the special feature that the money market interest rate reaches the zero lower bound (Graph 4). From the beginning of 2007 to the end of 2008, the Federal Open Market Committee (FOMC) lowered the target federal funds rate by more than 400 basis points from 4.75% to 0-0.25% trying to address the risks of slowing economic growth, inflationary pressures, and financial market disruptions4. At this level the target rate is set until now.
4. Empirical Strategy
In this section, I describe the empirical strategy followed to test the impact of low interest rates on bank risk-taking. I will divide the study into four topics. The first one attempts to assess the effect of changes in the money market interest rate on the percentage of non-performing loans as well as the effect of the level of interest rates relative to a benchmark rate given by a Taylor rule. The second considers alternative measures to account for monetary policy when interest rates are near the zero lower bound. In this case, I include the nominal interest rate and replace the benchmark rate by the interest rate sample mean and the annual mean. Then, I include banks’ characteristics to control for other factors that may alter the transmission of monetary policy, and finally, I test unconventional monetary policy using the change of the Federal Reserve balance sheet assets.
4
- 16 - 4.1. Interest rate and the risk-taking channel
In order to assess the effect of the stance of monetary policy on banks’ risk taking, I estimate a panel data regression. The first limitation in a model that attempts to measure the effect of accommodative monetary policy on risk taking is separating the effect of low interest rates on the riskiness of outstanding loans from the effect on incentives to take on more risk caused by misperceptions of risk or search for yield (Altunbas, Marqués Ibañez and Gambacorta, 2010). Additionally, for the correct identification of the model monetary policy must be exogenous to bank risk-taking.
To address the first limitation, I include the change of the fed funds rate with respect to the previous period and also the deviations of the interest rate from a benchmark rate following Altunbas, Marqués Ibañez and Gambacorta, (2010). This strategy allows to independently measure the effect of low interest rates on risk-taking since low rates are defined as negative deviations from a benchmark rate.
Regarding the second problem, it results very difficult to deny possible responses of monetary policy to financial distress during the crisis, but it is more appropriate to assume that the response of monetary policy does not take into account financial institutions’ incentives to take on more risks. In this sense, monetary policy objectives in the underlying sample period continue to be price stability and economic growth. As it is stated in the Federal Reserve Monetary report (2009), the FOMC participants agreed to maintain a low level of short-term interest rates for some time and to use balance sheet policies and communications in light of the sharp deterioration of the economy and the easing of inflationary pressures. Unconventional policy measures were directed to support the correct functioning of financial markets and facilitate the transmission of the monetary policy stance to the economy. The caveat in the model specification still exits, because in the first place it is very difficult to prove causality between monetary policy and bank risk due to the possible endogeneity of the monetary policy, and in second place,
- 17 -
it is difficult to find out how much risk-taking can be attributed to monetary policy (Altunbas, Marqués Ibañez and Gambacorta, 2012).
Despite the limitations, assuming that conventional monetary policy does not respond to financial stability incentives, I will test empirically the effect of low interest rates on bank risk-taking using the following econometrical specification:
(1)
with , t , where N is the number of banks, t represents the quarters and T is the final quarter. is the dependent variable measured as the
percentage of the total portfolio that was impaired for bank j in period t. In the right-hand side, the second term shows the response of the credit risk indicator to a
change in the short-term interest rate. The third term denotes the deviation
of the interest rate from the benchmark . The last regressor is the nominal GDP growth rate which is included to control for macroeconomic conditions. Better economic conditions increase the number of projects becoming profitable, thereby reducing credit risk (Kashyap and Stein, 1995); but also banks may become riskier if they reduce their screening activity and lending standards during expansions.
Rewriting Equation (1) :
(2)
I estimate the model in Stata using an unbalanced panel with fixed effects to control for unobserved variables that differ across banks, but do not change over time.5
For the benchmark monetary policy rate I estimate a Taylor rule6 through the following regression:
5
Due to the short period of the sample, I assume that bank business characteristics did not change throughout the time.
- 18 -
(3)
Where is the long term interest rate, measures inflation and the output
gap. Potential output is estimated using the Hodrick-Prescott (HP) filter. The inflation trend and the long term interest rate are also estimated using the HP filter. Alternatively, I implement a policy rule using a rule as originally proposed by John Taylor (1993):
(4)
In this case the rule does not include interest rate smoothing and inflation target was set at 2%.
4.2. Risk-taking and the zero lower bound:
In the period under analysis the zero lower bound is constraining the ability of the central bank to further lower the fed funds rate in the face of a weak economy and low inflation. Constructing the Taylor gap where interest rates are at the zero lower bound gives as a result negative deviations from the benchmark for the whole sample. To tackle this problem and also improve the robustness of the results I will estimate the following alternatives models:
(5)
(6) (7)
Firstly in equation (5), I include the nominal interest rate to account for the distance of the interest rate to the zero lower bound. Secondly, I calculate the difference between the level of the interest rate for each quarter with the annual mean and with the sample mean (Equations 6 and 7). These alternative approaches have the purpose of defining 6 A Taylor rule is a monetary policy rule that links systematically the level of the policy rate to deviations of
inflation above a target or deviations of the real GDP above its trend (Taylor; 1993). Policy rules would thus inform policy decisions; not a mechanical formula (Taylor and Williams, 2010).
- 19 -
better measures to assess the impact of low interest rates on bank risk taking. The hypothesis behind is that if interest rates have been accompanied by extraordinary policy measures for too long a time, in presence of slow recovery on the functioning of the banking sector, two possibilities may arise. On the one hand the anticipation of persistently low short-term interest rates can lead to socially excessive short-term leverage and incentives to hold excessively illiquid assets (Diamond and Rajan, 2011). On the other hand, it might be the case that banks do not respond by taking more risk; on the contrary, they might maintain or reduce the riskiness of their portfolios so as not to have higher losses if the central bank reverts the current policy of low interest rates or if extra liquidity injected in the system is not transform in higher lending but on liquidity hoarding (Berrospide, 2013).
4.3. Interaction with banks’ characteristics
In the following part of the analysis I modify the baseline equation (Equation 1) to control for bank characteristics and analyze their interaction with the monetary policy stance. Following the literature, I include the capital to assets ratio (Kishan and Opiela, 2000), and the liquidity ratio7 (Stein, 1998). Capital is also measured using the regulatory capital (Equation 9). In order to assess the interaction of bank characteristics with monetary policy, an interaction term between the interest rate gap and capital is included (Equation 10).
(8)
(9)
7
The liquidity ratio is defined as the percentage of liquid assets over deposits and short-term funding. “This is a deposit run off ratio and looks at what percentage of customer and short term funds could be met if they were withdrawn suddenly, the higher this percentage the more liquid the bank is and less vulnerable to a classic run on the bank” (Bankscope - Bureau van Dijk, 2012).
- 20 -
(10)
Additionally, I include the ratio of short-term funding over total-funding where short-term funding comprises customer deposits, money market funding and other short-term funding8. To my knowledge this aspect is innovative in the literature since I do not recall a paper that makes a distinction of liquidity and funding. Borio and Zhu (2008) calls for a better understanding of the relation between risk-taking and liquidity. In this sense, they relate two different notions of liquidity. In first place “funding (cash) liquidity”, or the ability to realize value from liquid assets either via the sale of an asset or access to external funding. The other is “market liquidity”, or the ability to trade an asset in short notice without an impact on its price. For the purpose of investigating more deeply the link between liquidity and risk-taking, I will make a distinction between liquid assets and funding or banks liquid liabilities9. Liquid assets makes reference to bank’s holdings of cash, securities or bonds with the possibility to liquidate some of them at the market value in case of a negative liquidity shock. The second one considers the liabilities side including the possibility of a bank to raise easily liquid funds or be affected by a shortage of liquidity as result of a negative shock such as a bank run or a disturbance in the interbank market. The last one is interesting at the light of the recent episodes in the interbank market, and as a mechanism that can strengthen the transmission of monetary policy. Consequently, I estimate the following models:
(11) (12)
8 On the contrary long term funding will include loans from banks and unsubordinated debt securities with a
maturity of more than one year.
9
- 21 -
4.4. Monetary response during the crisis: unconventional monetary policy
Finally, to ensure robustness in the model I will consider more broadly particular features of the monetary policy after the financial turmoil of 2007-2008. As developed countries were in need of expansionary policies to move the economies out of recession, central banks responded cutting policy rates aggressively. In the presence of a decline in output and low inflation, when interest rates reached the zero lower bound conventional policy rules implied negative nominal policy rates; thus, several central banks decided to carry out unconventional monetary policies. The distinguishing feature of this policies is that regardless of the level of the interest rates, central banks engage on an active usage of their balance sheet to affect directly market prices and conditions beyond the short-term interest rate. These “balance sheet policies” consist of managing the size and structure of central banks’ balance sheets separately from the policy rate (Borio and Disyatat, 2009).
The implementation of unconventional policy measures has the purpose of restoring the functioning of financial markets and intermediation and, provides further monetary policy accommodation at the zero lower bound (Habermeier et al., 2013). For this purpose, the Federal Reserve implemented different credit easing policies. In first place, the Federal Reserve used traditional tools such as the discount window, and implemented other credit facilities (Term Auction Facility, Primary Dealer Credit Facility, and Term Securities Lending Facility) to provide short-term liquidity to banks and other depository or financial institutions.
Second, through the purchases of assets, mainly mortgage-backed securities, the Federal Reserve provided liquidity to the market that had dried up in the wake of the financial crisis (Joyce et al.; 2012 ). Finally, it also implemented Operation Twist to lower long-term interest rates through the sells of short-term government bonds and purchases of long-term bonds. These programs include targets and a conditional commitment to maintain high reserve levels. The commitment is particularly important in the case of unconventional policies at the zero lower bound when a time inconsistency problem
- 22 -
undermines the credibility of the central bank, since it has incentives to not holding the interest rates at low levels after the economy starts to recover. (Habermeier et al., 2013).
Unconventional monetary policies have an effect on financial intermediaries. Measuring its impact is empirically difficult, but considering the characteristics of the period under analysis, I will attempt to include unconventional monetary policies in the regression. In the spirit of McCallum (1988) when reaching the lower bound of policy rates, interest rate rules are implicitly replaced by quantitative reaction functions, where the main policy instrument is a quantitative aggregate10. Thus, following Gambacorta, Hofmann and Peersman (2013) I will include the growth of the Federal Reserve’ balance sheets assets in the model.
The fact that the Federal Reserve extended lending to banks when the interbank market froze; and, thereby reduced the spread between term inter-bank lending rates and the overnight rate (Taylor and William, 2008) is important for the correct identification of the model. Managing the size and structure of central banks’ balance sheets separately from the targeted policy rate requires the application of the decoupling principle (Borio and Disyatat, 2009). According to this, a central bank engages in transactions that sterilize the impact of the operations on the amount of reserve balances. Otherwise, the central bank makes sure that any induced changes in the amount of bank reserve holdings do not have an impact on the market reference interest rate. The application of this principle in the analyzed period would be a strong assumption (Graph 5). Therefore, in order to avoid identification problems in the specification of the model I will consider balance sheets
10
Interest rate is still a policy instrument but cannot be lowered further once it reached the zero level. Taylor and Williams (2010) provide a simple monetary policy rule modified to account for the zero lower bound including a lower threshold as well as taking into account that denotes the preferred setting of
the interest rate in the previous period that would occur absent the zero lower bound. The rule is as following where ρ incorporates the inertia in
the behavior of the interest rate, denotes the equilibrium real interest rate, denotes the inflation rate in period t, is target inflation rate, and denotes the output gap. Taylor (2009, p.2) points out that ‘as long as the overnight interest rate is at zero, the monetary policy framework needs to focus on the level or the growth rate of the quantity of money. As soon as conditions warrant, the policy framework should again focus on systematic procedures for setting the overnight interest rate—a policy which works well, as has been demonstrated during the great moderation period’.
- 23 -
assets as the monetary instrument instead of interest rates, excluding from the model the interest rate:
(13)
Central bank’s balance sheets assets do not capture the effect of policies on the long term interest rates. Therefore, I will include the slope of the yield curve, a variable that has also been taken into account in the literature of the risk-taking channel (Altunbas, Marqués Ibañez and Gambacorta, 2010).
(14)
5. Empirical Results:
In this section I present the main empirical findings. First, in subsection 5.1 I analyze the effect of changes on the money market interest rate, and of deviations of the interest rate from a benchmark on non-performing loans (Equation 1). Consequently, in subsection 5.2 accounting for the low level of the fed funds rates, I estimate the impact of deviations of the interest rate from the annual mean and from the sample mean (Equations 5 to 7). In the following subsection, I control for bank characteristics including the effect of liquidity, capital, funding and their interaction with monetary policy (Equations 8 to 12). In adittion I also include the slope of the yield curve (Equation 14). Finally, in subsection 5.4 I test the effect of unconventional monetary policies.
5.1. Interest rate and the risk-taking channel
The estimation of the baseline model (Equation 1) is in line with the empirical evidence of the risk taking channel of monetary policy. The results are reported in the second column of Table 2. The effect of changes on the policy interest rate in the percentage of non-performing loans is positive and significant. A positive change of 1% on the money market interest rates leads to an increase of 1.62% on the risk of outstanding loans. A higher
- 24 -
interest rate reduces the profitability of current projects, thus undermining borrowers’ repayment capacity.
The coefficient for the Taylor rule gap is negative11. This coefficient measures the effect of low interest rates on risk-taking by banks (measured in this model by a higher percentage of non-performing loans). When the difference between the money market interest rate and the benchmark rate is positive and large (the market interest rate is higher than the value given by a Taylor rule), the percentage of non-performing loans to total loans is lower. If the interest rate is 100 basis points higher than the value given by the Taylor rule, the percentage of non-performing loans to total loans decreases in a quarter by 1.32%. Hence, the evidence suggests that monetary policy influences bank risk-taking whenever interest rates are below the benchmark. The fact that banks take on more risk when interest rates are very low could be the result of optimizing behavior (Dell Ariccia, Laeven and Marquez; 2010). Risk taking inclination can also be a social optimal outcome of monetary policy during recessions (De Nicolo et al., 2010).
The coefficient related to the GDP is positive suggesting that bank lending is riskier despite of good economic conditions. Since in the period under analysis economic growth rates are very low (Graph 6), the effect of the interest rate on higher bank risk taking prevails over the effect of more projects becoming profitable due to better economic conditions. This means that low interest rates influence riskier lending and that the economic conditions do not counteract this effect. Ioannidou, Ongena and Peydró (2009)find also a positive impact of GDP growth on the hazard rate of bank lending in Bolivia.
5.2. Risk-taking and the zero lower bound:
In the period under analysis the federal funds rate was at the zero lower bound. Consequently, the Taylor gap, the difference between the market FED fund rate12 and the
11
The Taylor rule used is estimated using Equation (3), results are provided in Table 5.
12 Recall that variations on the interest rate affect borrower’s repayment capacity and the risk of
outstanding loans, while a negative gap between the market level of the interest rate and the level given by a Taylor rule increases risk-taking incentives by banks.
- 25 -
benchmark rate, is always negative. For that reason and, in order to separate the effect of monetary policy on the risk of outstanding loans from the effect on risk-taking incentives, additional measures are considered. The first approach is to measure the distance of the interest rate to the zero lower bound (the nominal interest rate). The results are shown in Table 2, Equation (5).
As expected the coefficient of the nominal interest rate is significant and positive, suggesting that the larger the spread between the level of the fed funds rate and zero, the higher is the percentage of non-performing loans. Since this indicator does not allow to completely disentangle bank’s incentives to take on more risk from the effect of the level of interest rates on risk of outstanding loans, the difference between the quarterly interest rate and the annual mean is included (Equation 6). Contrary to expectations the coefficient for this variable is both positive and strongly significant (at 1% level); thus positive interest rate deviations from the mean increase the percentage of non-performing loans by more than 5%. While the reduction of the interest rate below the benchmark causes an effect on banks’ incentives to search for yield or take on more risk (Equation 1), the reduction of the interest rate below the annual mean does not reflect an increase on bank risk-taking; on the contrary, it strongly reduces the percentage of non-performing loans.
Short-term deviations from the mean might affect more the profitability of current projects or cash flows, which is reflected on a higher ratio of delinquencies, than bank’s incentives to take on more risk. This effect can be influenced by the low rates of economic growth during the period under analysis, which may also have a negative impact on the rate of return of projects.
The empirical finding that interest rates above the mean increase the ratio of non-performing loans may also be related to the backward looking nature of this indicator.
In order to better test the effect on credit risk of deviations of interest rate from the benchmark, I also include the difference between the quarter interest rate and the mean
- 26 -
for the whole sample period (Equation 7). In this case, the sign of the coefficient for this variable is negative and significant at the 1% level. Thus, the interest rate lower than the sample mean increases risk-taking in the banking sector.
The finding that the coefficient for deviations of the interest rate from the sample mean is negative as it is the coefficient for the Taylor gap, in conjunction with a positive coefficient for the interest rate deviations from the annual mean, might provide a new insight on the functioning of the risk-taking channel. Banks risk taking incentives seem to be influenced deviations of the interest rate from a benchmark being this given by the level suggested by a Taylor rule or by the mean of the interest rate across many periods, while short term deviations from the current level of the monetary policy stance affect mainly borrower’s repayment capacity and the risk of outstanding loans.
The incentive to search for yield is especially strong when interest rates are too low compared to an expected rate of return. In this case the Taylor gap can be considered a proxy of a measure for the risk premium a bank may require to receive in excess of the interest rate when the monetary policy interest rate is very low. Assuming a bank which has rational expectations and which fixes the expected return or the interest rate for its contracts in line with a Taylor rule; an accommodative monetary policy might increase incentives to search for yield. As a result, the bank may lower lending standards incurring on higher risk taking (Graph 7).
5.3. Interaction with banks’ characteristics
In the previous section, I found evidence that when interest rates fall below the level suggested by a Taylor rule, they influence banks’ incentives to take on more risk. Even when interest rates are at the zero lower bound, the coefficient of the Taylor gap results significant. Thus, risk tolerance or risk incentives to search for yield are negatively affecting the quality of the loan portfolio.
In the rest of the analysis, I include balance sheet characteristics to account for other variables that might influence the relation between too low interest rates and risk-taking.
- 27 -
In this case, I will consider the effect of liquidity, capital, funding and the slope of the yield curve. Results are presented in Table 3.
The first model includes the effect of liquidity and capital on the ratio of non-performing loans (Equation 8).
The coefficient for capital is negative, hence banks with higher levels of equity as a percentage of assets have less non-performing loans in their portfolios. The “skin in the game effect” attenuates risk-taking. In this point, I test also the model including regulatory capital instead of equity over total assets (Equation 9). As pointed out by Altunbas (2012) if there are agency problems between shareholders and managers that create incentives to “search for yield” pushing investors to require higher capital ratios, or if supervisors force banks to build up capital, a positive relation between capital and risk may arise. The inclusion of Tier 1 regulatory capital shows no changes regarding the sign of the coefficient compared to previous results. The coefficient for capital measured as equity over assets( -0.091) results higher than the one when using regulatory capital (-0.015). Therefore, banks holding more equity as percentage of risk weighted assets are more risk adverse to credit losses. On the contrary, banks with relatively low capital levels have higher loan losses since they respond to moral hazard incentives by increasing the riskiness of their loan portfolio (Keeton and Morris, 1987).
In the same model estimation, the liquidity coefficient results positive and significant. Therefore the hypothesis, that banks holding more liquid assets take less risks, can be rejected. Banks with a liquidity ratio that is one percent higher are estimated to have a ratio of non-performing loans one percent higher on average. This suggests that in the United States for the period under analysis, banks with more liquid assets tend to take more risks to compensate for the lower yield of maintaining liquid positions.
Subsequently, I explore the relation between risk-taking, bank capital and interest rates (Equation 10), though it does not result significant. The interaction term between capital and the Taylor gap or the nominal interest rate is not significant. Therefore, in this
- 28 -
empirical exercise highly capitalized banks have less percentage of non-performing loans. Furthermore, this result is not influenced by monetary policy.
Another variable of interest is one that attempts to measure liquidity financing. Borio and Zhu (2008) state that a link between liquidity and risk-taking can strengthen the transmission of monetary policy. In order to test for liquidity financing and risk-taking, I include the ratio of short term funding to total funding (Equation 11). The coefficient for this variable is positive and significant; banks that finance their activities mainly with liquid deposits, money market and short-term funding have a higher percentage of risky loans. Having more access to short term funding increases the percentage of non-performing loans by more than 2%. The interaction term of short term financing and liquid assets is also positive and significant. Liquid asset holdings and short-term funding influence the riskiness of the bank’s credit portfolio.
When interest rates are above the benchmark, banks do have a lower percentage of non-performing loans measured by the negative coefficient of the Taylor gap (-1.195). However, the interaction term between the Taylor gap and the variable that measures short-term funding (Equation 12) shows that short-term financing have a higher influence on credit risk. Accordingly, banks which have more access to short-term financing options have a higher percentage of non-performing loans. Maddaloni and Peydró (2010) also highlight the link between low interest rates, short term-funding and risk-taking. They point out that funding liquidity for banks is mostly short-term which in combination with too low interest rates may have turned the abundant liquidity into an excessive softening of lending standards.
Finally, since banks make profits from maturity transformation, the slope of the yield curve provides an indication of future expectations on the interest rates and future profits for banks. Due to the particularity of the period under analysis when the money market interest rate is reaching the zero lower bound, the slope of the yield curve is always positive. The coefficient in the model (Equation 14) for this variable is negative; thus, an
- 29 -
upward-sloping term structure means a lower percentage of non-performing loans13. The risk of outstanding loans decreases with monetary easing,14 and if the expected long term interest rate is higher than the short-term interest rate.
In terms of completeness, the model estimated in Equation (14) represents better the transmission of monetary policy through the risk-taking channel. In this sense, the coefficient of the Taylor gap is negative and larger than 1, thus when the money market interest rate is above the benchmark, the percentage of non-performing loans decrease by more than 1%. In this case, the coefficient for the effect of a change on the interest rate on the ratio of non-performing loans is lower than in Equation (1). Therefore, controlling for the effect of bank´s characteristics is important to better measure the effect of the monetary policy stance on risk-taking.
5.4. Monetary response during the crisis: unconventional monetary policy
Finally, I estimate the effect of balance sheet policies replacing the interest rate as the main monetary policy instrument by the Federal Reserve’s balance sheets assets. As shown in Graph 8, the Federal Reserve’s balance sheets assets increased considerably since 2009. However, the coefficient for this variable results to be not significant. The results are provided in the second column of Table 3. As explained in the previous section endogeneity problems may arise if interest rates and the change on balance sheets assets are included together, in this sense, a further exploration of the effect of unconventional monetary policies would be better approached using a different methodology. A panel data model, as the one employed in this study, does not allow to fully analyze the effect of unconventional policies on risk-taking.
The robustness of the results was tested using an alternative Taylor rule and a different risk-taking indicator.
13 Altunbas, Marqués Ibañez and Gambacorta, (2010) also found a negative relation between risk-taking (a
decrease in the EDF) and the slope of the yield curve.
14
- 30 -
The results of the model estimated using an alternative Taylor Rule as the one originally proposed by John Taylor (1993) (See 4.1 Equation 4.) are presented on Table 4. In this case the sign of the coefficient for the variable that measures the interest rate gap is still negative, but its value (-0.421) is lower than the one using the estimated Taylor rule15 (-1.2 when including bank characteristics). The results are consistent since the estimated Taylor Rule gives a higher value for the interest rate for a given GDP and inflation rate.
The second robustness analysis uses the percentage of reserves provisions for loans losses to total loans as dependent variable. This ratio indicates how much of the total portfolio has been provided for but not charged off. The higher the ratio the poorer the quality of the loan portfolio will be (Bankscope - Bureau van Dijk, 2012). The results are in line with the previous estimated models (Table 4). The larger and negative the gap between the interest rate and the benchmark, the larger is the percentage of reserves for losses of the total loan portfolio. Therefore, it can be infer that banks expect a higher hazard rate and risk-taking is greater.
6. Conclusions
The recent period of low interest rates put in evidence that monetary policy influences banks’ incentives to take on more risk. The risk-taking channel (Borio and Zhu, 2008) of monetary policy implies that low interest rates alter risk perceptions or risk-tolerance, increasing the degree of risk on banks’ portfolios. Through the effect of interest rates on valuations and perception of risks, portfolio reallocation and the search for yield (Rajan, 2005); monetary policy may reinforce the build-up of risks in the financial system. In this sense, this study finds that this mechanism operates even when interest rates are near the zero lower bound.
Deviations of interest rates from a level implied by a Taylor rule induce banks to take on higher risks. This effect remains even when interest rates are near the zero lower bound.
15
- 31 -
However, short-term variations on the level of interest rates do not cause higher risk taking while they affect more the riskiness of outstanding loans. This last finding is important since current interest rates are near the zero lower bound. Consequently, the central bank will reach a point at which an increase in the Fed funds rate will be the optimal policy which in turn may have an impact on credit impairments and solvency.
This investigation also highlights the link between short-term funding and the transmission of monetary policy. Liquidity and short-term funding strengthen the impact of the monetary policy stance on risk-taking. Thus, the risk-taking channel acts as a “persistence-enhancing” mechanism (Borio and Zhu, 2008) in interaction with the characteristics of the financial sector and the business cycle. Large short-term funding can also leads to higher risk-taking even when interest rates are above the benchmark.
Central banks would need to be aware of the possible effects of increasing interest rates while credit quality is low and economic conditions affect the profitability of projects, since this policy may have a negative effect on outstanding loans and undesirable effects from a financial stability perspective. In addition, this study suggests that a better analysis of unconventional monetary policies would be recommended. Short-term funding is an important variable influencing bank risk-taking; thereby the effect of unconventional monetary policies on providing collateralized funding after the disruption of the interbank market may have important consequences.
- 32 - Appendix
Table 1: Description of the sample
Variable Obs Mean Std. Dev. Min Max
Non-Performing loans 107207 3.270351 4.141841 0 100 Capital 107210 11.0336 4.661441 -10.910 96.830 Liquidity 107180 12.68233 15.18775 0.000 873.800 Funding 107186 0.9672787 0.0525747 0.241 1.000 Tier 1 106874 16.88434 13.97922 -14.340 538.560 i 111747 0.1492851 0.0446356 0.080 0.220 GDP nominal 111747 1.47E+07 567238.7 1.390 1.560 Fed Assets 111747 0.0257742 0.0379297 -0.028 0.095
The sample includes high stressed banks and even banks which went bankrupt.
- 33 -
Graph 2: Non-performing loans
Example for some banks of the sample
Source: Bankscope
Graph 3: Non-performing loans (all banks)
Source: Federal Reserve of St. Louis
0.00 1.00 2.00 3.00 4.00 5.00 6.00 Perc e n t Month-year Sample period
- 34 -
Graph 4: Fed funds rate and quarterly change
Graph 5: Monetary Base components
- 35 -
Graph 6: Nominal GPD growth rate
Graph 7: Lending Standards
Net percentage of banks tightening lending standards.
Source: Senior Loan Officer Opinion Survey on Bank Lending Practices
-1.1% -0.3% 0.5% 1.3% 1.0% 1.0% 1.1% 1.1% 0.5% 1.3% 1.1% 1.0% 1.0% 0.7% -1.5% -1.0% -0.5% 0.0% 0.5% 1.0% 1.5% GDP growth -20 -10 0 10 20 30 40 50 60 70 Per cen t year-quarter
- 36 -
Graph 8: Federal Reserve Balance Sheet Assets
- 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 3,500,000 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 M ill io n s o f D o llar s Per ce n t year-month
- 37 -
Table 2: Empirical Results
*** Significant at 1% level. ** Significant at 5% level. * Significant at 10% level.
Dependent variable: ratio of non-performing loans to total loans OLS Fixed Effects Baseline Model Equation (1) Nominal interest rate Equation (5)
Deviation from the annual mean
Equation (6)
Deviation from the sample mean
Equation (7) Independent variable: Coeff. S. error Coeff. S. error Coeff. S. error Coeff. S. error Coeff. S. error
1.285*** 0.357 1.623*** 0.153 1.075*** 0.19 1.152*** 0.154 1.075*** 0.192 -1.478*** 0.08 -1.327*** 0.071 -1.52*** 0.074 -1.239*** 0.069 -1.52*** 0.742 0.199*** 0.032 0.332*** 0.019 0.356*** 2135 0.516*** 0.023 0.356*** 0.021 1.52*** 0.074 5.88*** 0.286 -1.521*** 0.223 -5.924*** 0.476 -5.104*** 0.43 -6.53*** 0.461 -4.70*** 0.42 -6.303*** 0.45
- 38 -
Table 3: Empirical Results Dependent variable: Ratio of non-performing loans to total loans Fixed Effects Bank Characteristics:
Liquidity and Capital
Equation (8) Bank Characteristics: Tier 1 Regulatory capital Equation (9) Bank Characteristics: Capital interaction term Equation (10) Bank Characteristics: Funding Equation (11) Bank Characteristics: Funding and interaction term Equation (12)
Slope of the yield curve
Equation (14) Independent variable: Coeff. S. error Coeff. S. error Coeff. S. error Coeff. S. error Coeff. S. error Coeff. S. error
1.507*** 0.154 1.394*** 0.159 1.510*** 0.159 1.356*** 0.159 1.353*** 0.159 0.800*** 0.138 -1.253*** 0.069 -1.222*** 0.072 -1.290*** 0.072 -1.198*** 0.071 -1.195*** 0.071 -1.079*** 0.529 0.32*** 0.020 0.340*** 0.020 0.317*** 0.019 0.323*** 0.02 0.318*** 0.02 0.384*** 0.017 0.017*** 0.003 0.015*** 0.003 0.018*** 0.003 0.018*** 0.003 -0.091*** 0.012 -0.066*** 0.013 -0.094*** 0.012 -0.094*** 0.012 -0.066*** 0.011 2.237*** 0.793 2.456*** 0.796 2.636*** 0.786 2.492*** 0.779 2.076*** 0.079 -1.092*** 0.015 0.0911** 0.051 -0.015*** 0.003 5.251** 2.700 0.019 0.048 -3.86*** 0.431 -6.582*** 0.817 -6.523*** 0.814 -6.055*** 0.798 -5.888*** 0.802 -4.613*** 0.780
- 39 -
Table 4: Empirical Results
*** Significant at 1% level. ** Significant at 5% level. * Significant at 10% level.
Dependent variable: ratio of non-performing loans to
total loans
Fixed Effect Robustness tests Alternative Taylor Rule FED Balance Sheet Assets
Equation (13)
Dependent variable: Reserves to total loans
Independent variable: Coeff. S. error Coeff. S. error Coeff. S. error
0.774*** 0.168 0.674*** 0.020 0.521*** 0.4254 -0.421*** 0.131 -0.443*** 0.028 0.642*** 0.021 0.037*** 0.005 0.020*** 0.003 0.020*** 0.003 0.006*** 0.001 -0.097*** 0.012 -0.093*** 0.012 -0.036*** 0.004 3.998*** 0.801 1.463*** 0.305 -0.210 0.128 -1.970*** 0.883 3.502*** 0.137 -2.073*** 0.296
- 40 -
Table 5: Estimation Taylor Rule
. *** Significant at 1% level. ** Significant at 5% level. * Significant at 10% level. Linear Regression: Number of Observation: 30, R-Squared: 0.9321.
Dependent variable:
Taylor Rule
Equation (3)
Independent variable: Coeff. S. error
0.718*** 0.917
0.116 0.113
